Continuing a discussion on Uncommon Descent I am posting here to allow more participants, if they wish.
The discussion stems from this comment by Gpuccio.
Let’s suppose that species “b” derives from species “a” through darwinian mechanisms.
Let’s suppose, for simplicity, that the main difference between the two is the emergence of a new protein, “B”, form a pre-existing protein, “A”.
Let’s say that “B” is very different from “A” (maybe a new protein domain), so that the transition from one to the other qualifies as a dFSCI transition. For instance, the transition could require 200 bits of functional information.
Such a transition cannot happen through mere RV.
Therefore, our darwinian friends have to deconstruct the transition into intermediate functional steps, just to be credible, so that NS cam enter the scenario.
Let’s pretend they succeed (they never have, but just for discussion…).
So, the transition from “A” to “B” has been deconstructed into, say, 20 intermediates, each of them functional and selectable. No single intermediate transition is so complex as to configure dFSCI at a 150 bit threshold, so in theory each transition could have happened by chance, and then been selected.
Well, my simple question is:
Why today, in the existing proteome, we do find protein “A” (in the progenitor species) and protein “B” (in the derived species), but no instance of any of the 20 intermediates?
Rather than repeat what is written on UD I will summarise my position (in any case some of my UD comments were written in a hurry and are unclear or wrong ).
- I am not a biologist or biochemist. I only got involved with this because it interests me. I have no doubt there are others much better qualified – but they probably also have much better things to do with their time.
- I dispute some of Gpuccio’s premises. There are all sorts of problems with the concept of dFSCI as discussed elsewhere. It may be true that if you changed 20 amino acids at random the chances of getting protein B from protein A are incredibly small. But it does not follow that each step has to have a selective advantage. This is for simple reason that if P(X|Y) is very low it does not follow that P(Y|X) is very low.
- However, it is still reasonable to ask why there are no intermediaries between two proteins A and B that are very different but similar enough to be supposed to have a common ancestor.
- There are two possibilities. One is that there were no intermediaries. The common ancestor jumped straight to A and B through just a few mutations that included at least one large change e.g. insertion, deletion, translocation. There is an interesting discussion as to whether this is feasible but rather beyond my competence to assess.
- The other is that there were intermediaries but they were removed. This is very similar to the discussion of why there are no living intermediaries between two present day species with a common ancestor e.g. chimpanzees and gorillas. The steps from the common ancestor to each species must have been fairly small – animals do not give birth to viable young that are dramatically different from their parents – but these steps no longer exist. As the two branches diverge the later representatives of each branch supersede the earlier members of each branch which were closer to each other.
- There are a number of theories of how this happens (these theories are not meant to be mutually exclusive). One is of the most popular is allopatric speciation and I don’t see why something similar should not happen in the case of two proteins or protein families derived from a common ancestor.
- The steps are fairly straightforward.
- Imagine a population with the ancestor protein X.
- A small subset of the population is isolated so there is no possibility of gene flow between the subset and the main population (and very likely the environment is significantly different).
- Both parts of the population continue to evolve and many proteins will be subject to small changes but in different directions (the smaller population will change more quickly). There may be some selection pressure owing to the different environments or it may be nothing but genetic drift. The variants may become widely established or be fixed at relatively low frequencies in the population.
- Some of those proteins will have variants of the variants and so on. The chances of a protein having many variations will be very small – but some will. This is more likely in a small population.
- At some stage one of these proteins with multiple variants hits on a real bonanza and eliminates all the other variants of that protein in that population. This is protein B – most likely in the small isolated population.
- Protein A in the other population may be identical to or very similar to the original protein X – or maybe that population got lucky too.
- As far as I can see there is nothing particularly ad hoc or implausible about this process. We know that small populations do get isolated start to vary genetically from the main population. You can see that in human communities. It would be difficult to bet on which protein would diverge before the process started. But it seems quite a good bet that at least one protein will undergo this process (just as there is very good chance there will be a lottery winner – but a very small chance it is you). The winner is the one that we retrospectively call the start of new protein family.
I will leave it at that for the moment.
In Moderation 
Mark:
Just to start the discussion, could you please detail better this statement of yours:
“It may be true that if you changed 20 amino acids at random the chances of getting protein B from protein A are incredibly small. But it does not follow that each step has to have a selective advantage. This is for simple reason that if P(X|Y) is very low it does not follow that P(Y|X) is very low.”
Thank you.
Sure – but it is elementary stuff already much discussed in the ID debate in general and between us in particular. First let me clarify – I am assuming that the probability of getting from A to B is lower if there is no selective advantage.
(1) As I have described before, any outcome can be described in a way to make it as improbable as you like. So do you look at the probability of getting to protein B exactly, or a protein with any kind of fitness advantage, or a protein that is viable, or even a gene that is not expressed but capable of creating a protein that is 20 aas different?
(2) Even if you settle on one of these specifications you have to have an alternative hypothesis to compare it to and there must be evidence for this hypothesis other than the observed outcome. Sober explains this very well in Evidence and Evolution.
Mark:
I have always been very clear as to how the outcome has to be described. I am surprised that I have not yet been able to clarify that with you after all our discussions about dFSCI.
The first moment in analyzing dFSCI is to have an outcome. Given my definition, the outcome must be readable as a digital string of values.
Both proteins and protein coding genes satisfy that requirement.
I will go on with proteins, for simplicity, considering their primary structure as a sequence in base 20 and length equal to the number of AAs.
Are you OK with that?
Then, an observer, any observer, must define the function for that protein. That can be done in different ways, but it doesn’t matter. Any definition can be chosen, provided that it is objective and that it includes an objective way to measure the defined function in any output, so that we can clearly judge if it is present or absent in that output (usually through a conventional threshold of functionality, that too defined by the observer).
So, for an enzyme, the simplest way to define the function is: “any protein which can accelerate reaction such of at least such in these defined lab conditions”
Each different definition of the function conditions what follows. Therefore, each measurement of dFSCI in an output is linked to the definition of the function.
A good definition of the function will yeld the most relevant value of dFSCI and will make the following measurements easier.
So, as discussed recently with Neil, I can define the function of my old computer as “being useful as a doorstop”, but its complexity for that function will be low, and not interesting. But if I define the function as being able to compute at a certain level of efficiency, than everything changes.
So, for an enzyme, what we are interested in is that it can perform its biochemical task at least at a minimum level of efficiency. And how much complex information is necessary for that (in bits).
Please, note that this has nothin to do with the function being selectable in a darwinian context. I am defining the function much more simply, at the biochemical level. A good biochemical function is not in itself selectable. And I am not discussing of the gene is expressed or not (IOWs, I am not discussing the gene regulation, but only the biochemical function of the output).
For a transition form A to B which gives to B a new function not present in A, the output is essentially the necessary changes in primary sequence to pass from A to B. If the change is of 50 AAs, we will have to try to compute the functional space (the number of combinations of those AAs which bring to the functional island of B (that is, to any protein with the function of B), and the search space (that is easier: all possible combinations of those 50 AAs). We are assuming here vthat the rest of the molecule remains the same because it is fixed by NS (a very generous assumption in favour of the darwinian model, and not a realistic one).
As you can see, while there are certainly difficulties in the calculations, defining the function is not one of them.
It is certainly possible to define the function beared by a mutation in terms analogue to NS: that’s what has been done in that recent paper which was discussed at UD about mutations created in two ribosomal proteins, where almost all mutations decreased slighty fitness (even synonimous ones), and none increased it. In that case, we need an experimental setting to measure reproductive fitness with precision. The concept is different, because in that case we measure the global effect of a mutation on fitness, which includes all indirect effects. That is different form my concept for dFSCI, where the direct function of the protein is measured.
The second scenario is rather a good way to test the effects of NS, but it seems that the results are not very encouraging for darwinists, up to now. However, it is always a measurement of dFSCI, where the function is defined as “an increase in fitness of at least such in this specific fitness measuring setting”.
Gpuccio
I was mainly interested in the removal of intermediates – but I guess this is good a place as any to discuss the problem of the description of function.
Yes you have been clear about the description you would like to use – performing its biochemical task. But of course there are many other descriptions – some with higher probabilities and some with lower – from being a viable protein through to performing this task at a given speed. AS far as I can remember you have never justified your particular choice other than it being “relevant”, “clear” or “simple”. This hardly seems an adequate justification. This is not me word smithing. We are concerned with how new protein families are derived. The first member of a new protein family could be any protein that is viable – it might not even be active. At some later stage it could be involved in some biochemical function or simply mutate into closely related proteins in the same family wich become active. There is nothing about a specific biochemical function here.
gpuccio,
So we are defining specified complexity based on a given level of function, and the percentage of sequences which can provide at least that level of function? Is that correct?
gpuccio: It is certainly possible to define the function beared by a mutation in terms analogue to NS: that’s what has been done in that recent paper which was discussed at UD about mutations created in two ribosomal proteins, where almost all mutations decreased slighty fitness (even synonimous ones), and none increased it.
Yes. If a function is optimized, then any change will tend to reduce the level of function. If we then define functional specificity as being at a singular peak, then the specified complexity will be a simple exponent based on the length of the sequence. But if we relax our specification, then the specified complexity will decrease.
I would like to point out that gpuccio has never provided a detailed, worked example of how to calculate CSI (or any variant thereof) for a real world biological system or component. He is happy to go on at length about his conclusions, but never has he shown how he arrives at them.
So, gpuccio, I challenge you yet again to simply show your work. I suggest the evolution of the ability to digest citrate from Lenski’s experiment. Please demonstrate exactly how to calculate the CSI of that newly evolved function.
Mark:
First of all, I hope you will pardon me if I ignore Maya’s interventions. I have already answered extensively all her questions elsewhere, and frankly I cannot waste any more time.
Regarding the definition of function, what is the problem? dFSCI is a tool for interpretation in some model. The definition must be appropriate for the general model we are using.
For the emergence of protein superfamilies, my model is very simple: a protein which has not achieved the folding and active site indispensable for their specific biochemical function is useless. I don’t see how it would be selectable or other.
New superfamilies, in my model, emerge because there is a new biochemical function which is necessary in a more general context. This is a very reasonable model, and the definition of the basic biochemical function is absolutely appropriate for that model.
There is really nothing special in this reasoning. I suppose that if scientists had your kind of doubts, they would never build explanatory models for reality. If you have a better model, which explains things, and in which function can be defined so that it emerges without the levels of complexity implied in my model, please expose it and I will seriously take it in consideration. You can specify the model, define your function, and compute dFSCI for your function.
I am waiting.
Gpuccio
I am not talking about explanatory models. I am not denying that a protein superfamily fulfils a specific function (what would I know?). I am asking how you justify one specification which a protein matches as opposed to another when doing a probability calculation.
Let me try an analogy.
An organism starts with protein A. After many millions of years this protein is transformed to protein B which is so different it counts as the basis of a superfamily (we will get back to this missing intermediates later). But it could have been transformed to a very large number of other conceivable proteins which were equally different, some of which might have brought different unknown fitness advantages to that organism. So if we regard every conceivable transformation as a lottery ticket the organism was lucky enough to stumble across a winning ticket. However, there were an unknown number of other winning tickets and we have little idea the chances were of getting a winning one (not all transformations are equally likely). You seem to be saying “how extraordinary I got this ticket”. I am saying should you not be taking into account the probability of getting a winning ticket or the the chances of getting a ticket at all? I should then add that every protein is candidate for such a transformation so actually the organism has many hundreds of thousands of chances to draw a winning ticket.
Mark:
You say:
” there must be evidence for this hypothesis other than the observed outcome”
What evidence is there for the hypothesis of macroevolution by darwinian mechanisms, which has been accepted for more than a century? Just to know…
Plenty. The mechanisms – RM+NS etc – have been observed. You may dispute whether they account for large scale change but at least there are mechanisms to assess. No one has even begun to observe the mechanism of the designer.
gpuccio,
That is completely untrue. If you have provided a detailed, worked example of how to calculate CSI for a real world biological system or component, just point to it. You have never done so.
You continue to waste time bloviating about your dFSCI hobbyhorse without ever rigorously defining your terms or demonstrating how to calculate it for a real biological artifact.
It is well past time for you to put up or shut up.
Maya
I really would like this discussion to take place on a basis of mutual respect. I think Gpuccio is deeply mistaken but I respect his honesty, intelligence and openness to other views. This is what makes the discussion interesting.
I also think he has made a honest attempt to estimate something call dFSCI in some proteins – for example here. I think these attempts are fallacious but we are the stage of pointing out what is wrong with the attempts rather than accusing him of not doing it.
gpuccio, see our question above.
gpuccio: For the emergence of protein superfamilies, my model is very simple: a protein which has not achieved the folding and active site indispensable for their specific biochemical function is useless. I don’t see how it would be selectable or other.
That’s not necessarily correct. Simple peptides may have biological function, and primitive life may have been full of such activity.
The evidence indicates that the most recent common ancestor already had many of the protein domains that make up modern organisms. Perhaps they arose spontaneously, or maybe they evolved from primordial ancestors, or both. Even if we assume the former, that doesn’t imply design, as many random amino acid sequences can apparently fold into biologically useful proteins.
Mark,
I am paying gpuccio more respect than he has earned by taking his arguments seriously enough to address them. I read through the thread you referenced and I must disagree with your assessment; at no point did gpuccio provide a mathematically rigorous definition of dFSCI, nor did he demonstrate how to calculate dFSCI for any biological system, nor did he address any of the substantive points raised by Petrushka, MathGrrl, or warehuff.
Anyone serious about science must understand that clearly defining one’s terms and providing the calculations underlying one’s conclusions is essential. When I challenged gpuccio to do so here, he again failed to do so. My final comment on that thread is still applicable here:
Unless and until gpuccio addresses this issue, his claims are baseless. Intellectual integrity demands he stop making them until he can support them.
My way of showing gpuccio’s ideas respect is far less combative than what he would experience were he to try to publish in the peer reviewed literature. Nonetheless, if you wish me to discontinue posting here I will do so. That won’t make gpuccio’s failure to support his claims go away, though.
Maya – I am not going to ask anyone to discontinue posting unless it is very extreme. That’s the whole point of moving the discussion here. I just think it is more productive to proceed on the basis of mutual respect – whatever might happen in the academic publishing world!
Zachriel’s point is the important one. Gpuccio feels he has provided a sufficiently rigorous definition. He has certainly put quite a lot of work into it. It seems more constructive to explain why it is not rigorous than just repeat that he has failed to define his terms. Otherwise we risk ending up like Behe continually asking for more and more detailed accounts of how natural selection worked 600 million years ago without beginning to provide his own account of common descent.
Thanks for providing a forum where the comments of one side aren’t held back until they are buried.
The basic question posted by ID is whether conditions whose antecedents have been erased by time are best considered to have natural (uniformitarian) antecedents or supernatural antecedents.
One will never know the TRVTH, but I find it interesting that supernatural causation is seldom invoked in courts of law, even though most juries have been predominantly composed of believers.
Regarding Durston and Axe, I think it is relatively easy to construct experiments and scenarios in which causes are not found. Somewhat more difficult to find naturalistic scenarios.
I would ask the same question of biologists as I would of police investigating a crime scene. What sort of hypothesis is most likely to be productive in the long run: one that assumes natural causes, or one that invokes unseen, undefined vapors and emanations?
Maya: I read through the thread you referenced and I must disagree with your assessment; at no point did gpuccio provide a mathematically rigorous definition of dFSCI, nor did he demonstrate how to calculate dFSCI for any biological system, nor did he address any of the substantive points raised by Petrushka, MathGrrl, or warehuff.
Ah, but he thinks he has! Here’s a few problems with gpuccio’s definition:
gpuccio: You ask why the non compressibility is important for dFSCI.
Unfortunately, compressibility doesn’t have a unique solution. A sequence may appear non-compressible, unless you know the key.
gpuccio: Therfore, it is important that the string we analyze be a truly pseudo-random string, otherwise a necessity mechanism cannot be ruled out.
Therefore, if you require knowing whether it is “truly” pseudo-random, then your calculation may have already failed.
gpuccio: So, the threshold must be high enough to empirically exclude the random emergence of the functional sequence in a random model (or of the random transition, in a model with random steps).
That’s a problem, too. Random emergence, then optimization through random mutation and selection is not only possible, but easy to test and replicate. The function should be definable as the weakest selectable function to comport with the model.
Another problem is that the same structure may have more than one function, perhaps one optimized, another weakly functioning. And that means it may have more than one measured specified complexity.
Mark,
Thank you for your very hinest intervention about Maya. It is sincerely appreciated.
gpuccio,
You wouldn’t need to hide behind other people if you’d support your ideas yourself.
Clearly define your terms and provide your calculations or stop making your claims. To do otherwise is dishonest.
Zachriel:
Thank you for coming here to discuss.
You brief post sums up many of the basic differences between your position and mine. While respecting yours, I will try to make my position more clear.
You say:
“Simple peptides may have biological function, and primitive life may have been full of such activity.”
Many “mays” here. OK, I agree with you that simple peptides can have function, there is some evidence for that, but it is usually a regulatory function. Complex enzymatic functions require longer molecules.
And anyway, when I evaluate dFSCI in a protein, I refer to the specific fucntion of that protein. So, for an enzyme, the function is the catalytic activity of that enzyme. Therefore, unless you can show that short peptides can perform that catalytic function, your note about short peptides is of no interest to the analysis of that protein.
The “may” about primitive life, IMO, remains a myth and nothing more. You may believe that kind of things, but I don’t feel in any way inclined to take them seriously.
You say:
“The evidence indicates that the most recent common ancestor already had many of the protein domains that make up modern organisms.”
That’s perfectly correct. About half of basic protein domains were already present in LUCA. See also my post here:
http://www.uncommondescent.com/intelligent-design/the-darwinian-basis-of-the-prokaryote-to-eukaryote-transition-collapses/#comment-367664
But the other half developed in the course of natural history, and in all forms of life.
See also this important reference, especially Table 1:
http://www.plosone.org/article/fetchObjectAttachment.action;jsessionid=278A0A6D05C24BABC4253F231CF7EAAB.ambra02?uri=info%3Adoi%2F10.1371%2Fjournal.pone.0008378&representation=PDF
You say:
“Perhaps they arose spontaneously, or maybe they evolved from primordial ancestors, or both.”
I am always surprised at how many darwinists are ready to “bypass” the problem of OOL, even defining it beyond the boundaries of science.
For me, it is rather obvious that the mechanism which was able to find half of the basic protein domains in a few hundred million years, from OOL to LUCA, is the best candidate to explain the emergence of the other half in the course of natural history. I believe this is the most parsimonious way of explaining facts. And the only candidate I am aware of is design.
Do you want to discuss OOL and the evolution to LUCA?
You say:
“Even if we assume the former, that doesn’t imply design, as many random amino acid sequences can apparently fold into biologically useful proteins.”
Here I completely disagree with you. If you base this statement on the Szostak paper, you may know that I have analyzed it in detail at UD, and that I find it deeply biased. If you are interested, I can try to find the link to the long discussion I had there with some very good biologist. I remain convinced that the Szostak paper is a very biased artyicle.
Axe is correct in his view of this problem, and the coming research will certainly make light on these aspects. I have all the reasons to believe that existing knowledge already points to functional proteins as extremely isolated islands of functionality.
I will close with a question to you. I understand that much of your faith in the darwinist mechanism of evolution is base on the conviction that functional, selected intermediaries of functional proteins must exist. How do you explain that no functional intermediary of the basic protein domains is known? And according to which strange principle do you believ that such a thing should be true?
IOWs, why should comple protein fucntions (for instance, the functional seqwuence of an enzyme) be deconstructable into simpler functional intermediaries? That is certyainly not true of complex functions we find, for instance, in computer programs. Why should it be true of protein sequences?
Beware, I am not speaking here of modular systems, where different modules, retaining their functionality, may contribute to a more complex entity and therefore to a new complex function. I am speaking of basic protein domains, not of multidomain proteins or of exon shuffling.
“I believe this is the most parsimonious way of explaining facts. And the only candidate I am aware of is design.”
———————
I seem to recall asking you to explain why design is superior.
My intention was to gain some understanding of what process you believe underlies design.
If the designer is omniscient, there is no problem. But if the designer is finite and intelligent in a way that is analogous to humans being intelligent, there is a problem. There is no known way to anticipate the effects of protein coding strings. It is extremely difficult to compute protein folding. Chemistry is faster than any known process of emulating chemistry, in the absence of quantum computers.
I think if you compound the problem of computing the effects of evolution with the problem of survival in a changing and competitive ecosystem, you have a computational problem that is beyond any known means of solution, other than evolution itself.
So I think it is fair to ask what attributes and capabilities you assign to the designer.
Zachriel:
some comments on your comments:
“Unfortunately, compressibility doesn’t have a unique solution. A sequence may appear non-compressible, unless you know the key.”
I am well aware of that. But here the only meaning of that requirement is to exclude sequences which can be generated by some known algorithm. In that case, the algorithm would be a necessity explanation of the sequence, or at least of part of its information. You must understand that my definiton is empirical. If no known algorithm can generate that sequence, that’s enough for me. We are dealing with competing explanatory models here, not with absolute truth. This is empirical science, not philosophy. I have always been very clear on that point.
“Therefore, if you require knowing whether it is “truly” pseudo-random, then your calculation may have already failed.”
I have already answered that in the previous point.
“That’s a problem, too. Random emergence, then optimization through random mutation and selection is not only possible, but easy to test and replicate. The function should be definable as the weakest selectable function to comport with the model. ”
It’s not a problem at all. You are right. The function must be defined as a function which is required to emerge in a random way in the model.
I will be more clear. Let’s say that we choose protein A, which is a basic protein domain which emerges at some point of natural history. We know no ancestor, we know no functional intermediary in the proteome for that domain. So, the only model is to explain its whole functional information in the context where the protein first arises.
By the Durston method, we can compute the functiona information for that protein family. If it is higher than some reasonable threshold for random biological systems (I have suggested 150 bits, which, believe me, is still extremely generous), I conclude that a random roigin is not the best explanation for that protein, and I adopt design as the best explanation.
But obviosuly, as soon as you can show me any functional intermediary in the proteome, or even in the lab, (let’s call it pre-A) which could have been selected before and copuld have helped the emergence of that protein by providing an intermediate functional state, I will accept that, and repeat my calculation separately for:
1) The emergence of pre-A
2) The transition from pre-A to A
that is, for tha two events for which, in your new model, there is no evidence of a necessity explanation.
I will not accept as a credible model the “theoretical possibility” that such intermediaries may exist. There is no reason to believe that. There is no evidence that they may exist, even less that they ever existed.
You must forgive me: in science, I am not a man of faith.
Finally, you say:
“Another problem is that the same structure may have more than one function, perhaps one optimized, another weakly functioning. And that means it may have more than one measured specified complexity.”
Again, that’s no problem at all. I have been very clear that each calculation of dFSCI is relative to a specifically defined function. Therefore, any string, or molecule, can have as many values of dFSCI as we can define functions for it.
The utility of a dFSCI value, as I believe I have repeated many times, is only in the context of an explicit model. So, the defined fuinction must be appropriate for that model, otherwise computing dFSCI has no particular meaning.
One very general request: it would be useful if my interlocutors here (except Maya, to whom I will not answer) could be understanding, and tried to keep the discussion focused on a limited number of arguments, possibly related. I am one, and while I appreciate all your questions, I cannot answer everything at the same time. And I would like to keep some time for posting at UD, too…
Mark:
About your post starting with “I am not talking about explanatory models.”
Explanatory models are what science is all about. I would add, competing explanatory models.
But I understand your point. Anyway, I don’t believe it is relevant.
Your point is another way of expressing the concept which I call “any possible function”. Simply put, many darwinists, realizing the extremely huge search space for sequences of humdreds of aminoacids, comfort themselves saying:
“But we must not necessarily get this specific protein. That is only the result which occurred. But evolution can proced thorugh “any possible function”, and therefore the probabilities of success are much higher than it seems”.
I don’t believe that is a good reasoning.
First of all, it is not true that evolution can proced through “any possible function”. Darwinian evolution can only use “any function which can be selcted by NS in a specific context”. Have you idea of how restrictive this is, compared to the myth of “any possible function”?
I have always stated that a compelx envirinment is a dramatic constraint of what can be functional in that environment (I am referring here to the biological environment already existing in the cell, not to the outer environment). And NS is an extreme constrint too, because “naturally” selectable functions are a truly minuscule subset of all “artificially” selectable functions. This fundamental difference is often overlloked by darwinists.
Moreover, even if there are a number of selectable functions at a certain stage, tjheir probability must be summed. The result will not be specially comforting in a search space of hundreds of orders of magnitude.
The requirement for specific functions becomes almost exact, finally, when those functions must be integrated at higher level, in irreducilbly complex systems, as it is almost always the case.
So, I am afraid that I am nor impressed by that kind of objections. It is perfectly reasonable to go on analyzing the probability of a specific function which did arise, whitout being sidetracked by vague philosophical arguments which have no empirical or mathemathical support from reality.
Gpuccio
There are two issues here. One is – to what extent does expanding the target space to “any possible function” increase the probability. You write how the context and the need to be subject to natural selection limits this target space. I am not qualified to comment but it seems to me that noone (including you) really has much idea of the target space. Remember that the context which you find so constraining is not static. A mutation can enable a change in context. And natural selection is not always required to fix a mutation as an allele. A mutation which is not expressed may lie dormant for many generations.
The second point is more philosophical. I intended “any possible function” as an example of an alternative specification. You prefer the more limited specification of “fulfils this function”. They have different probabilities. There may be other specifications with yet again different probabilities. How are you to justify your preferred choice?
gpuccio: Many “mays” here.
You had said “I don’t see how it would be selectable or other.” Hence, maybes, especially bolstered with observational evidence, would be relevant. We know peptides can have biological functions, and can even form bilayer membranes.
gpuccio: Complex enzymatic functions require longer molecules.
Yes, but complex structures can evolve from simpler ones, which was the nature of your comment.
gpuccio: Therefore, unless you can show that short peptides can perform that catalytic function, your note about short peptides is of no interest to the analysis of that protein.
That wasn’t the question raised by your comment above.
Zachriel: Perhaps they arose spontaneously, or maybe they evolved from primordial ancestors, or both.
gpuccio: I am always surprised at how many darwinists are ready to “bypass” the problem of OOL, even defining it beyond the boundaries of science.
Abiogenesis is not beyond the bounds of science, but the details are beyond what we have been able to discern with any certainty.
For me, it is rather obvious that the mechanism which was able to find half of the basic protein domains in a few hundred million years, from OOL to LUCA, is the best candidate to explain the emergence of the other half in the course of natural history. I believe this is the most parsimonious way of explaining facts. And the only candidate I am aware of is design.
Seriously? You’re not aware of any other candidate?
gpuccio: Do you want to discuss OOL and the evolution to LUCA?
Only when shown relevant to the topic.
Zachriel: Even if we assume the former, that doesn’t imply design, as many random amino acid sequences can apparently fold into biologically useful proteins.
gpuccio: Here I completely disagree with you. If you base this statement on the Szostak paper, you may know that I have analyzed it in detail at UD, and that I find it deeply biased.
Huh? The experiments have been repeated and extended many times.
gpuccio: If you are interested, I can try to find the link to the long discussion I had there with some very good biologist.
Punt.
gpuccio: I understand that much of your faith in the darwinist mechanism of evolution is base on the conviction that functional, selected intermediaries of functional proteins must exist.
How would get from random sequence libraries forming functional proteins to the “faith” that all functional proteins must have precursors?
gpuccio: How do you explain that no functional intermediary of the basic protein domains is known?
Some domains may have been formed de novo, others from simpler common ancestors. But even if de novo, it doesn’t necessarily mean design. It could just mean random assembly or some other principle not yet understood. In order to support design, you have to be able to propose a clear hypothesis with specific entailments.
gpuccio: But here the only meaning of that requirement is to exclude sequences which can be generated by some known algorithm.
That’s not what you said, “it is important that the string we analyze be a truly pseudo-random string, otherwise a necessity mechanism cannot be ruled out. ” That’s right. You can’t rule out a “necessity mechanism” unless you have explored every possible (not just every known) algorithm. That’s an inherent problem, where you have defined your results in terms of your ignorance. The more ignorant you are, the more likely you are to conclude design.
gpuccio: I have always been very clear on that point.
That’s fine. Every known algorithm.
gpuccio: I will be more clear. Let’s say that we choose protein A, which is a basic protein domain which emerges at some point of natural history. We know no ancestor, we know no functional intermediary in the proteome for that domain. So, the only model is to explain its whole functional information in the context where the protein first arises.
We already have an explanation for the high degree of optimization. All we have to explain is the origin of some degree of function.
gpuccio: I will not accept as a credible model the “theoretical possibility” that such intermediaries may exist. There is no reason to believe that. There is no evidence that they may exist, even less that they ever existed.
In fact, you are the one making the claim that such precursors are not possible. They are certainly plausible, as we have countless examples of protein evolution. Otherwise, we could simply say we don’t know at this time.
Petrushka:
An old argument. I don’r consider design supernatural. Indeed, I rarely, almost never, use the word “supernatural”. My map of reality does not allow such silly distinctions. What is, is.
Design is the product of conscious representations and conscious intent in a conscious being. That is a very simple, purely empirical definition.
Consciousness is not supernatural, for me. Conscious beings exist, and there is no evidence that cosnciousness is created, or is restricted to, a physical body. Indeed, most human maps of reality have thought differently.
If you like to call anything that is beyond your personal map “supernatural”, that’s your choice. But I will not follow you on that path.
Petrushka:
But it is very obvious why design is superior: design creates dFSCI, every day, everywhere. How can such a simple fact elude the intelligent minds of so many darwinists?
We have a property which comes out in tons at each moment from conscious intelligent beings, and which has never been observed coming out from any system where no conscious intelligent being has been involved. And you ask why design is superior in explaining dFSCI? Are you kidding?
“Consciousness is not supernatural, for me. Conscious beings exist, and there is no evidence that consciousness is created, or is restricted to, a physical body.”
————————–
No evidence except that no example of consciousness apart from a human body exists. That and the fact that physical defects in the brain are reliably correlate with characteristic changes in behavior.
My main point, however, is that any entity or agency that is not omnipotent or omniscient will encounter the same problem with large numbers that ID proponents use as evidence against evolution. The simple fact is that evolutionary algorithms are the only known effective class of solutions for large search spaces.
Mark:
“I am not qualified to comment but it seems to me that noone (including you) really has much idea of the target space.”
Not completely true. The target space of specific proteins can be approximated, either following Axe’s line, or the more precise Durston method. It is certainly big, but it leaves a huge ratio to the search space in all cases (in orders of magnitude).
Summing a finite, and certainly not very big, number of functional domains which could be useful in a context will not help.
Just think: all evolution, in 4 billion years, has found only about 4000 basic domains (in terms of families). Half of them were necessary for the first life. 500 of them were necessary to create the eukaryotes. the others were required for all other diversifications.
Do you suggest that a much bigger number of functionally useful domains exists (let’s say 10^100) and that only a minuscule subset has been explored by evolution, that wonderful evolution which can exploit “any possible function” and which is continuously stimulated by changing fitness functions in the outer environment? Is that what you really believe?
I don’t. I believe that the 4000 basci domains we observe are probably most of what is useful in the living context, out of the huge search space of proteins. I believe that the reason why half of them have been discovered before LUCA is that, without them, LUCA simply could not exist, because OOL, the transition from inanimate matter to living beings, is certainly the biggest informational leap we can imagine.
I believe that the reason why only a few new domains have been discovered in mammals, and maybe none in humans, is that recent evolution, while stunning in its results, is obtained mainly by intelligent programming at the regulation level (a level whose biological basis we still understand only in little bits), and requires a higher level of information, while the basic biochemical information is already available and can be reutilized.
A simple question: if the natural search space of proteins is so rich in functiona as you and Zachriel seem to believe, how is it that intelligent protein engineering, in many years, has not yet found one single new protein fold which may be said really useful to something?
Mark:
You say:
“The second point is more philosophical. I intended “any possible function” as an example of an alternative specification. You prefer the more limited specification of “fulfils this function”. They have different probabilities. There may be other specifications with yet again different probabilities. How are you to justify your preferred choice?”
Again, it’s a problem of models. We try to understand reality through models.
The only way to know the probability of “any possible function” would be to be able to detect “any possible function”. That is impossible, even in principle, because you can define as many functions as you want.
But you can ask: I want to know the probability of a well defined event. But you have to define it clearly.
I have defined my event, and my event is more than useful for my model: it is perfectly appropriate.
You could simple define: any possible variation which increases fitness in a specific fitness measuring system. That is fine. It has recently been done. You know the results.
So your justification for your particular specification is:
“my event is more than useful for my model: it is perfectly appropriate”
Well my specification is “any protein”. I find it more useful for my model and perfectly appropriate.
Where do we go from here? Don’t you feel the need a further justification?
gpuccio,
I think what Petrushka is trying to say is that any designer like us, that has restrictions, that is fallible, would have the same problems trying to design something that will give us the result we want.
Also, any future goal we are targetting for, would reside in a future environment that may be completely unforseen by us unless, we had the ability to see that future first.
A designer like us, would have the same problems we would have trying to design for an unknown environment.
He would end up with more misses than hits, a lot more misses.
Zachriel:
I must say that I don’t find in your last comments many things on which I could really discuss. Maybe I miss something in your logic.
I will try anyway, at least for some points.
“We know peptides can have biological functions, and can even form bilayer membranes.”
I have never denied that peptides can have biological functions. I have denied that they may be intermediaries to more complex functions.
“Yes, but complex structures can evolve from simpler ones,”
That’s exactly what I deny.
“Abiogenesis is not beyond the bounds of science, but the details are beyond what we have been able to discern with any certainty.”
That’s funny. It’s not the details that are missing. It’s the full substance.
The only “viable” theories of abiogenesis today are, as far as I can say, those based on the myth of the “RNA world”: we have absolutely no empirical data which suggest that such a thing ever existed. I could define it “wishful thinking”, but I will call it for what it is: a mere myth, created to defend an undefendable theory. This is not science.
“Seriously? You’re not aware of any other candidate?”
No. None serious.
“Huh? The experiments have been repeated and extended many times.”
References, please.
“Punt.”
I don’t understand what that means.
“Some domains may have been formed de novo, others from simpler common ancestors.”
Simpler common ancestors which have never been observed. Like RNA based beings. What about being empirical sometimes?
” But even if de novo, it doesn’t necessarily mean design. It could just mean random assembly or some other principle not yet understood. In order to support design, you have to be able to propose a clear hypothesis with specific entailments.”
Given that random assembly would violate all known laws of probability, I am happy that you are proposing a clear hypothesis with specific entailments: “some other principle not yet understood”.
“The more ignorant you are, the more likely you are to conclude design.”
Then I am happy that I am very ignorant.
Anyway, I maintain that the non compressibility has only one purpose: to rule out explicitly known algorithms which could in theory work in the biological environment. All we know of the biochemistry of life exludes that hypothesis. Therefore, I will go on pursuing what facts point to, instead of wasting time with completely unjustified theoretical possibility. But if you want to waste your time, that’s your choice. If and when you find a biological algorithm which can explain the generation of protein functional sequences, please let me know.
“We already have an explanation for the high degree of optimization. All we have to explain is the origin of some degree of function.”
What explanation?
“In fact, you are the one making the claim that such precursors are not possible. They are certainly plausible, as we have countless examples of protein evolution. Otherwise, we could simply say we don’t know at this time.”
First, I am making the claim that no such precursors are known or have ever been observed. From that I make the empirical inference that they probably don’t exist.
Second, they are not plausible at all, because we have no example of macroevolution by darwinian mechanism.
Third, saying that we don’t know at this time is fine for me. That’s why we build explanatory models. The design model works. Yours doesn’t.
Mark:
No. I only feel the curiosity of knowing what your model is.
gpuccio: Do you suggest that a much bigger number of functionally useful domains exists (let’s say 10^100) and that only a minuscule subset has been explored by evolution,
You’re conflating domains with protein sequences. A single domain may represent a large number of possible sequences. That’s why proteins form families and superfamilies, which probably represent only a tiny proportion of all possible sequences.
gpuccio: A simple question: if the natural search space of proteins is so rich in functiona as you and Zachriel seem to believe, how is it that intelligent protein engineering, in many years, has not yet found one single new protein fold which may be said really useful to something?
Again, that’s irrelevant. There may be only a few thousand domains, but they may still be evolvable from more primitive structures, or available to random processes. It’s doubtful, however, if nature has tested every possible domain.
Surdo1, Walsh & Sollazzo, A novel ADP- and zinc-binding fold from function-directed in vitro evolution, Nature 2004.
In any case, the evidence sugggests that functional proteins are reasonably common among random sequences, and even more common when using primitive amino acids.
Toronto:
I am happy to meet you again.
I agree with you that a designer “like us” will have problems and will be restricted by the context.
That’s exactly what I think of the designer of biological beings.
Many aspects of what we observe, regularly used by darwinists, for mysterious metaphysical reasons, to counter the design hypothesis, are of that kind: “an omnipotent designer would never have done that; an omnipotent designer would have done things as I would do them.”
Who has ever said that the designer is omnipotent? I have always thought that the designer acts in a context, and is limited by it. And that explains exactly what we observe.
But just a comment on the concept of “like us”. There is only one general and necessary implication of “like us” for a designer to be a designer: he must be a conscious intelligent beings, like we are.
But that does not in any way mean that he must have exactly the limitations we have. That is not true even of any two human beings.
So, conscious and intelligent, with conscious representations similar to ours (intent, cognition, and so on). But for the rest, he could certainly be different.
For example, he could, I hope, be smarter…
It’s good to talk to you again too!
Exactly. We observe trial and error methodology at work which looks a lot like what we consider evolution to be.
If there was a designer, that had the ability to create function that was applicable to an unknown future at the time of his design, we would see more of what Behe expects, “irreducible complexity”, a direct hit smack in the middle of a massive search space. What we see instead, is convergence to a function, where two or three different stabs at a lifeform compete, with none of them functioning better than any other.
Zachriel:
“You’re conflating domains with protein sequences. A single domain may represent a large number of possible sequences. That’s why proteins form families and superfamilies, which probably represent only a tiny proportion of all possible sequences.”
I am conflating nothing. My argument is about the origin of protein domains. The target space of each family has obviously to be computed, and the Durston method is an indirect way to do that.
If you read again my definition of dFSCI, you will see that this concept is expressed very clearly. Nobody in ID believes that the target space is made of one sequence.
“There may be only a few thousand domains, but they may still be evolvable from more primitive structures, or available to random processes.”
Both statements are not supported by facts. I don’t believe them.
” It’s doubtful, however, if nature has tested every possible domain.”
I have never said that. I have said that it is completely unreasonable to assume that in 4 billion years it has tested only 4000, while 10^100 or something are still there, waiting. You can believe that, if you like. I don’t.
Thank you for the reference. I will read it and comment (when you guys leave me the time 🙂 ).
Petrushka:
“No evidence except that no example of consciousness apart from a human body exists. That and the fact that physical defects in the brain are reliably correlate with characteristic changes in behavior.”
Well, first of all I think it exists in animal bodies. And I believe there are a lot of other evidences, but not scientific ones, that it exists independently. But that’s not what I would discuss here.
My point is only that billions of people have not found that belief unreasonable, and I don’t think that the scientist dogma has changed that simple fact.
Petrushka:
“My main point, however, is that any entity or agency that is not omnipotent or omniscient will encounter the same problem with large numbers that ID proponents use as evidence against evolution. The simple fact is that evolutionary algorithms are the only known effective class of solutions for large search spaces.”
I don’t understand. I am an entity, and I have no special reason of believing that I am omnipotent or omniscient. But I do solve the problems with large numbers all the time. Indeed, I output dFSCI all the time (and a little more, since I have come to this blog 🙂 ).
ID proponents use the “large numbers” argument against “undirected evolution”. Directed designed outputs solve those problems all the time. Because they come from conscious intelligent beings.
Evolutionary algorithms are an intelligent output of intelligent beings. That’s why they can solve some special problems, even some of those that intelligent beings cannot solve directly. That’s what intelligent beings do. They build machines to reach their purposes.
Toronto:
I don’t agree with you.
Trial and error methodology is a design methodology. Non conscious systems make no trial, and are aware of no error. Trial is a form of intent, and error o cognitive judgement.
“If there was a designer, that had the ability to create function that was applicable to an unknown future at the time of his design, we would see more of what Behe expects, “irreducible complexity”, a direct hit smack in the middle of a massive search space.”
First of all, all the “unknown future” staff is only your assumption. I have never said anything like that. My simple statement here has been that a designer is necessary to find the basic protein domains out of the search space.
Second, we see a lot of “irreducible complexity”. Almost everything in a cell is irreducibly complex. I stick to protein domains just because the evidence is more detailed, and I love details. But when we will be able to analyze in detail higher order functional systems in the cell, you will see…
Third, if it is true (and I believe it is) that basic protein domains have no intermediaries as precursors, then I would say they are “a direct hit smack in the middle of a massive search space”. Exactly what you asked for.
gpuccio: I have defined my event, and my event is more than useful for my model: it is perfectly appropriate.
Model? The only model that is apparent is evolution or random assembly. Did you propose a model?
Zachriel: Yes, but complex structures can evolve from simpler ones,
gpuccio: That’s exactly what I deny.
Oh.
We have substantial evidence of the evolution of complex structures. We might consider a simple and well-substantiated case, the mammalian middle ear. We can also show how complex structures evolve in silico so we can better understand the process and limitations.
gpuccio: That’s funny. It’s not the details that are missing. It’s the full substance.
It’s being approach from both ends, as it were; from chemistry and the origin of biological components, and from biology and the most primitive processes. There is no complete theory, but abiogenetics keeps pushing the curtain back. Maybe, they’ll find the Wizard if they keep looking, but that’s not what the evidence indicates at this time.
Zachriel: Huh? The experiments have been repeated and extended many times.
gpuccio: References, please.
Davidson & Sauer, Folded proteins occur frequently in libraries of random amino acid sequences, Biochemistry 1994.
Doi et al., High solubility of random-sequence proteins consisting of five kinds of primitive amino acids, PEDS 2005.
Tanaka et al., Comparative characterization of random-sequence proteins consisting of 5, 12, and 20 kinds of amino acids, Protein Science 2010.
gpuccio: If you are interested, I can try to find the link to the long discussion I had there with some very good biologist.
Zachriel: Punt.
gpuccio: I don’t understand what that means.
It means the dog ate your homework.
gpuccio: Like RNA based beings. What about being empirical sometimes?
We have ample evidence that life has evolved, and that this process extends back as far as we can discern. As our technical abilities increase, we continue to find more evolution at work.
There’s little point in discussing protein evolution at the cusp of life if you are not familiar with the evidence for evolution throughout the known course of organic history.
gpuccio: Like RNA based beings. What about being empirical sometimes?
RNA World exists. It’s in your cells.
gpuccio: Given that random assembly would violate all known laws of probability, I am happy that you are proposing a clear hypothesis with specific entailments: “some other principle not yet understood”.
As for “some other principle not understood,” then the proper position would be we just don’t know. Meanwhile, we know that random amino acid sequences can fold into functional proteins.
gpuccio: First, I am making the claim that no such precursors are known or have ever been observed. From that I make the empirical inference that they probably don’t exist.
You do realize that they keep finding intermediates, both macroscopic and molecular. When they do, they fill a Gap. But they create two new Gaps!
gpuccio: Second, they are not plausible at all, because we have no example of macroevolution by darwinian mechanism.
Okay. As we pointed out, there is no point discussing the most ancient transitions with the least evidence when you reject the well-established transitions with the most evidence. We should start there.
gpuccio: Third, saying that we don’t know at this time is fine for me. That’s why we build explanatory models. The design model works. Yours doesn’t.
What model? Have you proposed a model?
Zostak:
Unfortunately, the paper you reference is not freely available.
But I hope you have read it: from the abstract, I have impression that it is just a structural studt of the Zostak protein, or of one of its derivatives. Is that so?
Regarding the Zostak paper, here is the link to my discussion of it:
http://www.uncommondescent.com/intelligent-design/proteins-fold-as-darwin-crumbles/#comment-358394
If you like, we can discuss it, but I woul ask that you read first at least my posts 81 and 84, and possibly the following discussion with “rna”, starting at 138.
Zachriel:
“Model? The only model that is apparent is evolution or random assembly. Did you propose a model?”
Yes, that new protein domains emerge rather suddenly by design, without any related precursor or intermediary. That is a model.
“We have substantial evidence of the evolution of complex structures. We might consider a simple and well-substantiated case, the mammalian middle ear. We can also show how complex structures evolve in silico so we can better understand the process and limitations.”
I thought we were speaking of molecular evolution, and of examples of macroevolution of genetic information by darwinian mechanisms. Why are you changing subject?
“Maybe, they’ll find the Wizard if they keep looking, but that’s not what the evidence indicates at this time.”
The evidence at this time indicates exactly nothing. There is no relevant evidence.
For the references, again give me some time. Some of them I know, some not.
“It means the dog ate your homework.”
Again I don’t understand. Anyway, I have given you the link.
“There’s little point in discussing protein evolution at the cusp of life if you are not familiar with the evidence for evolution throughout the known course of organic history.”
Please, make me familiar. I must have missed the evidence.
“RNA World exists. It’s in your cells.”
You mean my cells work with rna only, without dna and proteins? I must really have missed a lot.
“then the proper position would be we just don’t know. Meanwhile, we know that random amino acid sequences can fold into functional proteins.”
I doubt. Anyway, we do know that designer ouptut dFSCI.
“You do realize that they keep finding intermediates, both macroscopic and molecular. When they do, they fill a Gap. But they create two new Gaps!”
I am not interested in macroscopic intermediates. I am discussing molecular events here.
And they are not finding any molecular intermediate to basic protein domains. Filling a gap to create two minor gaps is exactly what you need to solve the problem of dFSCI (well, maybe tens of gap fillings for each case). But nobody is finding anything.
” As we pointed out, there is no point discussing the most ancient transitions with the least evidence when you reject the well-established transitions with the most evidence. We should start there.”
Where’s the problem? Give me a well established recent transition to a basic protein domain, or any other well stablished macroevolutionary event whose molecular darwinian mechanism is known. I don’t pretend all the details, jusy a credible and realistically detailed model.
“What model? Have you proposed a model?”
Yes.
Zachriel, listen. I am not here to confront my dialectic ability with yours. I have great respect for your intelligence, and if we can discuss about interesting aspects of our views, even if retaining our personal convictions, I am very happy. But this is not a fight, for me. A fight can be funny, but in the long range it is not useful. So, let me know what you think.
gpuccio: If you like, we can discuss it, but I woul ask that you read first at least my posts 81 and 84, and possibly the following discussion with “rna”, starting at 138.
Nothing you said in your comment changes the basic fact. Random sequences folded into functional proteins that binded to a very specific molecule. That they were amplified is irrelevant to that point. They amplified sequences had to start with function in order to be subject to selection.
gpuccio: Yes, that new protein domains emerge rather suddenly by design, without any related precursor or intermediary. That is a model.
That’s not a model. A model has to account for how the design was implemented. Otherwise, you may as well say “Abracadabra!”
gpuccio: I thought we were speaking of molecular evolution, and of examples of macroevolution of genetic information by darwinian mechanisms. Why are you changing subject?
Because you said you deny that complex structures can evolve from simpler ones. That means we should look at well-established cases before considering those from the dawn of life.
gpuccio: But this is not a fight, for me. A fight can be funny, but in the long range it is not useful. So, let me know what you think.
It’s doubtful a discussion of the origin of protein domains can be productive while you deny evolution generally. When first engaging this thread, the extent of your denial wasn’t apparent. You had, after all, pointed to protein families and superfamilies. Again, that means we should look at well-established cases before considering those from the dawn of life.
Meanwhile, it may be productive to resolve the issue of random sequence proteins, something that’s well-established in protein engineering. It’s not a mystery.
gpuccio: I am not here to confront my dialectic ability with yours. I have great respect for your intelligence, and if we can discuss about interesting aspects of our views, even if retaining our personal convictions, I am very happy.
Our only goal is a civil and reasoned discussion. We are certain you hold your beliefs honestly, and we don’t challenge your sincerity or intelligence. However, your position is unsupportable, and we intend to point out the reasons why when given the opportunity.
gpuccio:
This is actually Toronto logged in with my blog name.
Without knowing what the future holds, how do you know your design will be valid in that environment?
That’s the problem with being a constrained by nature designer. If you’re going to design something as complex as life, it’s got to work generations into the future.
In the middle of Antartica, there are tree stumps, remnants of an ancient forest that existed when the environment was completely different.
Unless the designer is omnipotent,and can see the future, the idea of conscious design, has very significant drawbacks compared to a non-thinking feedback oriented system like evolution.
Mark,
You seem to mean something different by “respect” than I do. A number of people have pointed out the flaws in gpuccio’s arguments in UD (until banned) and other forums. If by “respect” you mean that we should continue to avoid pointing out his intellectual dishonesty, then I must disagree with your definition.
As I said before, spending the time to engage his arguments, yet again, is more respect than he has earned.
I agree that Zachriel’s points are important, but they don’t obviate the need to point out that gpuccio has never defined his dFSCI with mathematical rigor nor has he demonstrated how to compute it for a real biological system. Unless and until he does, he is quite literally talking nonsense.
Actually I think Gpuccio has made a reasonable effort at defining dFSCI and for fun I will try and put in my own words. This is not mathematically rigorous but I think it could be made more rigorous quite easily. As a result I think the problems will become apparent. I have found it easier to introduce a few terms of my own.
1) A digital string is any object comprising an ordered set of elements where the elements can take a finite number of discrete values – examples include a gene which is a string of DNA bases and a protein which is a string of amino acids.
2) The raw probability of a particular value of a digital string is calculated by assuming:
(A) Each element of the set has the same probability of taking any one of the discrete values (this does not mean that each discrete value has the same probability, but the probabilityof that discrete value is the same whichever member of the set is being considered). So every base pair in a gene has the same probability of taking each of the four values ATCG – but it is not necessarily true that P(A) = P(T) = P(C) = P(G).
(B) The probability of each element of the set taking on a discrete value is independent of the value of other elements of the set
(3) Some values fulfil a specific function. In the case of a protein this is a precise biological function such as carrying oxygen in the blood which must work in the overall context of the rest of the organism.
(4) The raw probability of digital string fulfilling a function is calculated as the sum of the raw probabilities of all values which fulfil that function.
(5) It may be that some natural process is known or suspected which increases the probability of a creating a digital string fulfillng that function. Call this the adjusted probability.
(5) The digital functional information in a string (which is only defined is relative to a function) is the -log2 of the maximum of the raw probability and the adjusted probability.
(6) If the digital functional information exceeds a certain value e.g. 150 then the digital string is said to have digital functional specified complex information.
I would be interested to know if Gpuccio agrees so far.
I would point out:
(1) The dFSCI in a digital string is defined relative to a function. Every digital string has a wide range of raw probabilities depending on the function chosen.
(2) The calculation of the raw probability is based on assumptions which have no justification other than a rather brave application of the principle of indifference (which is know to be inconsistent)
(3) The assignment of dFSCI to any object can only be considered to be provisional as it can be removed by the discovery of a higher adjusted probability
In effect dFSCI is not a property of a digital string but a property of relationship between a digital string, a function and known natural processes for creating that string. It is wrong to treat dFSCI as a property of an object which can be created and correlated with intelligence.
gpuccio,
Ignoring me will not make the facts that you have not rigorously defined your terms nor have you demonstrated how to calculate CSI for a real biological system go away. Continuing to assert your desired conclusions without supporting them in that way is intellectually dishonest. Doesn’t your holy book say something about bearing false witness? As I remember, it was against it.
“the idea of conscious design, has very significant drawbacks compared to a non-thinking feedback oriented system like evolution.”
——————–
That’s pretty much the nub of my question. How is a feedback system not a species of intelligent designer? Evolution is not conceptually different from how brains work.
And how would a non-omniscient designer gain knowledge of what works and what doesn’t except via feedback?
If gpuccio is asserting that there is some set of first principles governing protein folding and protein function, he is asserting something not in evidence.
At the very least, mutation and selection is at face value capable of traversing a search space. The claim that function is a set of peaks with vertical slopes is contradicted by the fact that nearly every living individual is genetically unique.
Zachriel:
” That they were amplified is irrelevant to that point. They amplified sequences had to start with function in order to be subject to selection.”
Wrong. They were amplified and artificially selected. The function which was studied in the final protein was not the same as that present in the original random sequence. I have discussed that in detail already. You are free to believe differently.
“That’s not a model. A model has to account for how the design was implemented. Otherwise, you may as well say “Abracadabra!””
Wrong. It is a model. Design can be implemented in many different ways, still it retains the characteristics of design.
We can discuss models of how design was implemented, and I have indeed done that many times. Unfortunately, we need more data to choose the best explanation in that range of models. But all those models are potentially explanatory, while the darwinian model isn’t.
“Because you said you deny that complex structures can evolve from simpler ones. That means we should look at well-established cases before considering those from the dawn of life.”
I was speaking of molecular structures, or if you want a wider scenario, of digital ones (as should be very clear from all the context).
Again, your reference to the “dawn of life” is wrong and unjustifiable. I have never required an example form the dawn of life. Basic protein domains have emerged throughout all natural history. You are free to produce a well established case form any geological era you like, instead of evading the question.
“It’s doubtful a discussion of the origin of protein domains can be productive while you deny evolution generally. When first engaging this thread, the extent of your denial wasn’t apparent. You had, after all, pointed to protein families and superfamilies. Again, that means we should look at well-established cases before considering those from the dawn of life.”
This is nonsense. I don’t deny evolution. I deny the darwinian mechanism. Don’t tell me you didn’t know that before engaging in this thread!
What does the phrase about family and superfamilies mean? I do point to families and superfamilies as my main argument for design in the proteome. And so? Where is the denial you speak of? I am only denying the validity of many of your arguments, because they are bad arguments. You are free to do the same with mine. I will not complain of your “denial”. That is simply an unfair way to discuss.
“Meanwhile, it may be productive to resolve the issue of random sequence proteins, something that’s well-established in protein engineering. It’s not a mystery.”
There is no issue. No useful complex function, least of all a naturally selectable one, has ever been found in random libraries.
“However, your position is unsupportable, and we intend to point out the reasons why when given the opportunity.”
The darwinian position is equally unsupportable IMO. I equally try to explain why. But I don’t want to convince anyone, and I don’t complain of anybody’s “denial”. Moreover, there is a slight difference between a constructive discussion, which tries to understand the motives if the other’s position, and discussion which, however respectful and intelligent, has the only purpose to show the supposed faults of the interlocutor at all costs. I accept both attitudes, but I respect the first (of which Mark is a very good example) a lot more.
uncommondescentdissent / Toronto:
I don’t really understand your point.
Microsoft programmers do go on writing software, and yet I don’t think they know “what the future holds” more than you or me.
I don’t know how much the biological designer knew of “what the future holds” when he designed living beings. But why are you so convinced that he could not design them without knowing every detail from then to eternity? That seems a weird idea to me.
“In the middle of Antartica, there are tree stumps, remnants of an ancient forest that existed when the environment was completely different.”
And so? What about the first, huge, computers? When our culture was completely different?
“Unless the designer is omnipotent,and can see the future, the idea of conscious design, has very significant drawbacks compared to a non-thinking feedback oriented system like evolution.”
The only strange “drawback” I can see is that conscious design can create complex functions, and non thinking feedback cannot. Indeed, I don’t thing the word “feedback” has any meaning outside of consciousness. In non conscious reality, there are only, maybe, events, and not feedbacks.
Have you read the last Abel paper, on these general questions? Here is the link:
Click to access 14TOCSJ.pdf
I highly recommend it.
Petrushka:
“That’s pretty much the nub of my question. How is a feedback system not a species of intelligent designer? Evolution is not conceptually different from how brains work.”
See also my previous answer to Toronto.
A “feedback system” is not intelligent. It designs nothing. Design is the projection of conscious representations and intents on an external output (see for instance the concept of “configurable switches” in the Abel paper I linked to.
A feedback system is a designed system, where an intelligent designer has configured the system so that it can treat some kind of data as “feedback information”, and react to them according to the way it has been programmed by the designer.
In non conscious reality there are at most events which can be interpreted by intelligent conscious observers according to laws (of necessity or of probability).
“And how would a non-omniscient designer gain knowledge of what works and what doesn’t except via feedback?”
Why shouldn’t a non-omniscient designer gain knowledge of what works and what doesn’t except via feedback? We do that all the time.
Feedback is not the only factor, anyway. The intrinsic faculties of intelligence, cognition, feeling and choice are the most important part of understanding.
So, I don’t understand your point. Feedback is indeed a component of intelligence and design. It is not a new, non thinking deity which pervades the cosmos, as some of you seem to believe.
“If gpuccio is asserting that there is some set of first principles governing protein folding and protein function, he is asserting something not in evidence.”
There is certainly some set of principles overning protein folding and protein function: it’s the laws of biochemistry. The problem is that computing final results like folding in a “top down” procedure is so complex, that our present computational resources are not enough. But nobody doubts that it can be done. So again, I am surprised of your strange statements.
“At the very least, mutation and selection is at face value capable of traversing a search space.”
As I have often discussed with you, a guided random search, coupled to intelligent selection, can certainly traverse many search spaces. And so? That’s one of my models for how design was implemented in biological beings, as you should remember.
“The claim that function is a set of peaks with vertical slopes is contradicted by the fact that nearly every living individual is genetically unique.”
This statement, frankly, eludes my poor understanding.
Gpuccio
“No. I only feel the curiosity of knowing what your model is.”
I have no idea what my model is, how a model would justify my choice of specification. I was trying (rather poorly) to illustrate I don’t understand in general the role that models play in your argument or how they justify your choice of specification.
“I have said that it is completely unreasonable to assume that in 4 billion years it has tested only 4000, while 10^100 or something are still there, waiting.”
This is one your poorer arguments. What matters is the ratio of viable domains to candidate domains and how frequently new candidate domains are “tested”. Your initial argument was that essentially the number of viable domains was of the same order of magnitude as the number that actually exist – therefore the ratio of viable domains to candidates is mindboggling low and beyond the scope of any reasonable frequency of testing. The counter argument is that the number of viable domains is a lot higher so the chances of finding 4000 of them over the course of life is reasonable. The chances of finding a large proportion of them may well be mindbogglngly low.
Mark:
I am afraid I miss your point. What exactly do you mean by “viable domains” and “candidate domains”?
I am sorry. I always write these things in too much of a hurry. Here is a rather detailed account which may veer the other way and be too detailed.
I will define three quantities:
E = The number of protein domains that actually exist. As I understand it E is approximately 4000.
V = The number of protein domains that would prove viable if evolution were to stumble upon them. These are the ones that you describe as “still waiting£.
C = The number of protein domains that evolution might stumble upon through successive mutations that might or might not prove to be viable. This is a slightly vague concept. It is almost the same as the number of different proteins that evolution might stumble upon – but if two proteins are very similar then they could both be considered to be a trial of the same domain (which might then prove to be viable or not). If necessary we can simply say C = number of proteins that evolution might stumble upon. I believe you would assume that C = number of possible permutations of amino acids up to a certain length – I don’t necessarily agree – but it is certainly a very large number indeed.
Let also define another number
T = the maximum number of trials of different protein domains that life has attempted since its inception.
Your argument, I believe, is that
V is similar in value to E.
E is mindboggling smaller than C.
Darwinism works by selecting members of C at random with equal probability and independently of each other.
Therefore the probability of stumbling upon a member of V in one trial = V/C = E/C is mind boggling small – much smaller than 1/T.
So the expected value of E = TV/C is far, far less than 1 and nowhere near 4000.
Therefore, the members of V were not discovered through Darwinism.
There are many problems with this argument and we have discussed them all too often here and elsewhere.
However, one objection is that V may actually be very much larger than E. So V/C = the probability of stumbling upon a member of V in a trials is much larger than expected and TV/C may actually be approximately 4000.
You argue that if V is so large how come E is not larger? But all the counter argument is saying is that:
TV/C may be close to 4000
This does not imply at all that we should expect discover most of the members of V.
I hope this makes sense!
Mark,
Very concisely put. This is one of the reasons I continue to note that gpuccio has not defined his terms with any degree of mathematical rigor.
I’m going to go back to lurking for a while, since I can’t add anything significant to what you, Zachriel, and Petrushka are writing. I admire your patience, but I predict that you will demonstrate the myriad failings of gpuccio’s arguments, with far more detail than he himself provides, at which point he will scurry back to UD to hide behind Clive’s skirts again, all the while claiming to have proven his case to those “Darwinists” who are just too close minded to recognize his brilliance. At no point during the process, however, will he define his terms more rigorously nor will he demonstrate how to calculate CSI for a real biological system.
The bottom line is that gpuccio, like other intelligent design creationists, cannot allow for the possibility that he is wrong. Reality must conform to his religious beliefs. Actually doing the work to make his claims testable is not only difficult, but dangerous. He’s just not going to do it.
If my prediction is proven wrong, I will, of course, apologize profusely and publicly to gpuccio.
gpuccio: They were amplified and artificially selected.
Yes. “Starting from a library of 6*10^12 proteins each containing 80 contiguous random amino acids, we selected functional proteins by enriching for those that bind to ATP.”
gpuccio: The function which was studied in the final protein was not the same as that present in the original random sequence.
No. “Starting from a library of 6 X 10^12 proteins each containing 80 contiguous random amino acids, we selected functional proteins by enriching for those that bind to ATP… We therefore estimate that roughly 1 in 10^11 of all random sequence proteins have ATP-binding activity comparable to the proteins isolated in this study. The function, albeit weak, was already there.
gpuccio: I have discussed that in detail already.
You pointed to your previous comments elsewhere.
gpuccio: What is being selected, and how? In the beginning, it is only any possible binding to ATP, even the weakest.
That’s right. A small number of random sequences folded into functional proteins that bind to ATP.
Zachriel: That’s not a model. A model has to account for how the design was implemented. Otherwise, you may as well say “Abracadabra!”
gpuccio: Wrong. It is a model. Design can be implemented in many different ways, still it retains the characteristics of design.
The Abracadabra model makes as specific of predictions.
Zachriel: Because you said you deny that complex structures can evolve from simpler ones. That means we should look at well-established cases before considering those from the dawn of life.
gpuccio: I was speaking of molecular structures, or if you want a wider scenario, of digital ones (as should be very clear from all the context).
Then you agree that evolution explains the evolution of complex macroscopic structures? If not, then we need to start there. It’s important, because much of the evidence for molecular evolution is based on common descent, and a consilience of evidence from many different fields of study.
gpuccio: This is nonsense. I don’t deny evolution. I deny the darwinian mechanism. Don’t tell me you didn’t know that before engaging in this thread!
We simply read your words. Do you accept Common Descent? Adaptation? If not, then we need to look at a variety of evidence.
Mark Frank: I hope this makes sense!
Not sure. A domain can be formed from a very large number of sequences, so counting domains doesn’t give a very good idea of their proportion in the set of all possible sequences. We know that functional proteins occur in at least 1 of 10^11 random sequences. Of course, nature may not work this way. Some domains may have discovered as simpler peptides evolved.
Maybe we’re misunderstanding your point.
gpuccio:
But they do know what the future holds when they get a spec outlining what needs to be done to support planned changes for communication protocol changes or new video standards such as 1020p. If they didn’t know and woke up one morning with their systems non-compatible, they would disappear from the market/environment.
To be successful, they need to know what’s coming.
It’s worse for a designer of life, as his designs must be viable for generations after their design.
The designer of life must also not have been designed, since that would be proof that life does not require a designer.
The only way out out is to have a completely omnipotent being not constrained in any way by a nature that could not bring forth life without his help.
Sorry, that should read,
“The designer of life must also not have been alive, since that would be proof that life does not require a designer.”
gpuccio:
When a non-conscious transistor circuit is designed, negative feedback is built into the circuit to bias the transistor at a known value which may be half of the supply voltage. If this isn’t done, even minor changes in temperature can cause the circuit to go to min or max values or even oscillate between the two. Ask kairosfocus about this.
Why couldn’t the designer design a feedback oriented life-modifying system that doesn’t require his conscious and constant attention to respond to future needs?
You could have your designer of this “system”, and we could verify that it works scientifically without needing to know about the designer or his intentions at all.
We believe such a system works and it is called The Theory Of Evolution.
gpuccio: Again, your reference to the “dawn of life” is wrong and unjustifiable. I have never required an example form the dawn of life. Basic protein domains have emerged throughout all natural history. You are free to produce a well established case form any geological era you like, instead of evading the question.
The most widely held hypothesis is genomic and, in particular, exon shuffling.
Zachriel: It’s doubtful a discussion of the origin of protein domains can be productive while you deny evolution generally. When first engaging this thread, the extent of your denial wasn’t apparent. You had, after all, pointed to protein families and superfamilies. Again, that means we should look at well-established cases before considering those from the dawn of life.
gpuccio: What does the phrase about family and superfamilies mean? I do point to families and superfamilies as my main argument for design in the proteome.
The term families usually refers to being related by descent.
We suggested that we were confused on your position, and offered a way forward in the discussion — that you clarify your position on Common Descent and macroscopic adaptation by natural selection. The Theory of Evolution, including protein evolution, depends on a consilience of evidence. You seem to be saying that evolution works for macroscopic forms, but that there is an insurmountable Gap for evolution in the origin of protein domains.
Zachriel:
About the Szostaz paper:
A weak binding to ATP is all that was present in the initial library. There is no evidence of any specific folding. The original molecules were treated with cycles of mutational PCR (amplification and variation), and each time selectede for ATP binding. That is intelligent protein engineering.
The properties of the final protein, on which the authors base their conclusions, were not in the initial proteins. Nobody knows if the initial proteins folded in any way, because nobody has studied them. They simply felt an unnecessary necessity to modify them. The work is propaganda, and violates the rules of good research.
If your point is simply that some weak biochemical binding to ATP (or, for that, to any other molecule) can be found in a small number of proteins in a large enough random library, I can agree. But that’s all. Your bold statement that:
“A small number of random sequences folded into functional proteins that bind to ATP.”
is completely unsupported by the paper, and the fault is of the authors.
gpuccio: A weak binding to ATP is all that was present in the initial library.
That’s all that matters! It’s a functional protein.
(Of course, they fold, or they wouldn’t bind so specifically, nor would simple mutation result in a folded protein with improved functional specificity. Furthermore, even if it forms a naïve binding, then evolves into a complex, folded protein, the result for our discussion is the same: It’s an initially random structure with a minimal function that evolves improved functional specificity.)
Zachriel:
“Then you agree that evolution explains the evolution of complex macroscopic structures? If not, then we need to start there. It’s important, because much of the evidence for molecular evolution is based on common descent, and a consilience of evidence from many different fields of study. ”
The neo darwinian model is a molecular model of RV and NS. Macroscopic structures are only the effect of what happens at the molecular level.
I accept common descent, but not unguided common descent, nor gradual common descent. I believe in designed, “punctuated” common descent.
“We simply read your words. Do you accept Common Descent? Adaptation? If not, then we need to look at a variety of evidence.”
As already said, I do accept CD. I don’t know what you mean by “adaptation”. Please, be more specific.
“The most widely held hypothesis is genomic and, in particular, exon shuffling. ”
No. Exon shuffling is a valid hypothesis for the origin of multi-domain proteins. In no way it explains the origin of single domains.
“The term families usually refers to being related by descent. ”
Which I do accept.
“We suggested that we were confused on your position, and offered a way forward in the discussion — that you clarify your position on Common Descent and macroscopic adaptation by natural selection.”
You only needed to ask. For CD, I think I have answered.
Your concept of “macroscopic adaptation by natural selection” is rather obscure. The cause of variation, as I have already stated, is genomic variation. Random, and not adaptational, according to neo darwinian model. And it is the genomic asset, again, which is selected, fixed and expanded. What has “macroscopic adaptation” to do with that?
“The Theory of Evolution, including protein evolution, depends on a consilience of evidence. You seem to be saying that evolution works for macroscopic forms, but that there is an insurmountable Gap for evolution in the origin of protein domains.”
Not at all. I am saying that darwinian evolution does not work at all as an explanatory model. Macroscopic forms are only a consequence of genomic information, and the relationship between the two things is too poorly understood to build explanatory models on macroscopic considerations.
But if your point is that some macroscopic continuity is a good evidence of CD, on that I agree.
And yes, my main point in this discussion is exactly “that there is an insurmountable Gap for evolution in the origin of protein domains.”
Toronto:
(may I call you so? I am rather a sentimentalist).
“But they do know what the future holds when they get a spec outlining what needs to be done to support planned changes for communication protocol changes or new video standards such as 1020p. If they didn’t know and woke up one morning with their systems non-compatible, they would disappear from the market/environment.
To be successful, they need to know what’s coming.”
Are you not going from one extreme (the designer must know everything up to very distant times) to the other (the designer cannot even know what will happen tomorrow)? Why such an extreme attitude? Why cannot a designer, like Microsoft engineers, be aware of the future possibilities of what he is designing, but not necessarily of every future development?
“It’s worse for a designer of life, as his designs must be viable for generations after their design.”
Well, a good designer can certainly achieve good enough results.
“The designer of life must also not have been designed, since that would be proof that life does not require a designer.”
Well, not the infinite regress again! I believe that a transcendent God is the only final answer to the problem of infinite regress. But I am aware that others prefer different “solutions”.
But that needs not be the central theme for biological design. Here we are not discussing the origin of all that exists. We are discussing the origin and development of something which has a specific beginning in time, in a specific place in space (except possible panspermia). We are discussing a realistic intervention of some intelligence on the material world, in time and space. We need not go to final infinite regress to explain that.
Maybe God is omnipotent and omniscient. I believe that, and I have my good reasons. But here I am arguing for a reasonable explanation of the intelligent and complex order and plans we onserve in living beings. And my idea is that the designer, or the designers, of that has certainly acted in a context, with objective limitations, but with huge skills.
“The only way out out is to have a completely omnipotent being not constrained in any way by a nature that could not bring forth life without his help.”
IMO, we need a transcendent being to justify the existence of any phenomenal reality. But that is a philosophical point, not appropriate here.
To explain the information in biological beings, all we need is at least one conscious intelligent being with the necessary understanding, intention and power. That does not imply omniscience or omnipotence.
“Why couldn’t the designer design a feedback oriented life-modifying system that doesn’t require his conscious and constant attention to respond to future needs?”
I believe he can, and he has done that. I believe there are many adaptational intelligent algorithms in the genomes (or more generally in living beings) which provide very smart and efficient adaptation (sometimes even neo-Lamarckian). I believe that the plasmid system and HGT in prokaryotes, as well as the working of the immune system, and may the transpposon system, are among them.
But new dFSCI cannot come from existing algorithms. Existing algorithms can at most incorporate some information from the outer world, and elaborate it according to the existing procedures. That’s what antibody maturation is about. Maybe an existing adaptational algorithm can sometimes “tweak” an existing protein’s active site to adapt to some external stimulus (the emergence of nylonase could be an example). But that is more or less all adaptation can do.
For truly complex dFSCI, and especially for higher level organization (body plans, regulation networks, and so on), adaptation is powerless. True creative interventions are necessary.
Well, I have read only in the end the corrected phrase:
“The designer of life must also not have been alive, since that would be proof that life does not require a designer”
I have difficulties with the definiton of “life”. While I have an empirical definition for consciousness, I have non for “life”. For me, the only simple requirement is that the designer must be a conscious intelligent being. And to anticipate a frequent objection, I don’t believe that consciousness implies complexity. For me, consciousness is in itself simple, while its representations can certainly be very complex.
Zachriel:
“That’s all that matters! It’s a functional protein.”
It’s a protein for which you can define an extremely trivial function, whose complexity is minimal.
“Of course, they fold, or they wouldn’t bind so specifically,”
The original proteins don’t bind specifically at all.
“nor would simple mutation result in a folded protein with improved functional specificity.”
Why not? RV + Intelligent selection is very powerful.
“Furthermore, even if it forms a naïve binding, then evolves into a complex, folded protein,”
through intelligent selection; only intelligent selection can recognize a weal and trivial biochemical interaction.
“the result for our discussion is the same: It’s an initially random structure with a minimal function that evolves improved functional specificity”
again, through intelligent selection. And even the final protein would never be selected by NS.
gpuccio:
In addition, the designer cannot be a living being, since that is what his role is, to actually design life.
That is the question that needs to be addressed and is the key to ID theory.
Why does life, described as being to complex to arise without being designed, have a designer with a greater complexity than life itself. Why did the designer not have to be designed?
In this context, it is not philosophical at all to discuss the relationship between the design and its designer. It is the ID scientific claim that design was a historical event.
So scientifically, if a complexity of “X” is the threshold we apply to a probablity achievable by non-design, and life exceeds X, why does a designer, who exceeds “X” by magitudes more than life, not require design?
This is not rhetoric, it is something ID has to address because without a designer, ID has no case to present at all.
Your designer is the major part of your theory, without which, you don’t have a theory at all.
Your designer could not have been alive, otherwise you again don’t have a theory.
Mark:
I answer first your post about dFSCI, becasue it’s easier.
First of all, thank you for having tried so well to understand my concepts. You have done a good work.
Obviously, there is some point I would like to specify.
a) You say:
“A digital string is any object comprising an ordered set of elements where the elements can take a finite number of discrete values”.
I would change that to :
“A digital string is any object comprising an ordered set of elements where the elements can be read as a finite number of discrete values”
Indeed, it’s not the physical elements in themselves that “take a value”. It’s the information which derives from assigning a digital value to each element. I again suggest that you read the Abel paper about constraints and controls, and again I endorse the clear distinction he makes between what he calls “Prescribed Information” (the same conceot as my “functional information”) and its physical support (for instance, configurable switches).
b) About (A): indeed, it is true that the four nucleotides have not exactly the same probability in the existing genome, but that could be the consequence of restraints secondary to the functional design. It is not completely clear if they have the same probability in a random biochemical system. You are wrong on one point: if the probabilities are slightly different “a priori” (which is possible), then they are also slightly different for each position. But that’s not a problem. It’s not a difference in biochemical probabilities of the individual nucleotides which can explain a higher probability of functional sequences. There is no known biochemical principle which can connect the two things.
A common error I have found on the blogs is that many believe that for a system to be completely random, it must follow a uniform distribution. That’s not true. If the probabilities of the two values of a coin are 0.4 and 0.6, the system is random just the same. In no way it is more able to express dFSCI.
About (5): I have never used this concept of “adjusted probability”. I assume that the probability of the various sequences in a random system is more or less the same (allowing for possible differences in the individual probabilities of the nucleotides). If a necessity mechanism intervenes (like NS), at that point the system is no more random. The fixation and amplification generated by NS changes the probabilistic resource, as I have already debated. That’s why, as in the neo darwinist model RV and NS must act sequencially (NS can only select what RV has generated), my approach is to split any transition in random parts and necessity parts. So, the intervention (proved and detailed) of NS would transform an event into two different random events, joined by a necessity mechanism.
The rest is fine for me.
About your comments:
(1) is fine.
Not so (2). The assumption of general similar probability of all sequences (with the exceptions already considered) is very reasonable here. There is really no theoretical or empirical reason to think that a sequence which codifies for a functional protein has more probability to arise in DNA, in a random system, than any random sequence. That is not brave at all. It’s the only reasonable assumption we can make. Again, you seem to forget that this is empirical science. It’s all about competing explanatory models. You may already have heard this, but what is your alternative, competing model?
About (3): let’s say by any new data which allow to refine, or correct, the model.
Finally, you say:
“In effect dFSCI is not a property of a digital string but a property of relationship between a digital string, a function and known natural processes for creating that string. It is wrong to treat dFSCI as a property of an object which can be created and correlated with intelligence.”
I perfectly agree with that. And, as I have tried to clarify many many times, the association between dFSCI and design is purely empirical, derived from observation, and not essential or philosophical or logical. That is a very important point. That does not exclude, obviously, that I have also theoretical reasons to connect dFSCI and design, but they are not the important point.
Mark:
about your other post:
now I have not much time, but I will try a brief comment.
First of all, I agree with Zachriel that there are technical imperfections in your reasoning. But that’s not the important point. I have understood better your point.
So, let’s say that again it’s a question of competing models.
As I seem to be the only one here fond of the modelling approach, I wil try to sum up what I see:
a) My model is more or less: The functional domains which can really be useful in a biological context are extremely rare. The existing proteome exhibits approximately 4000 of them. I have all the reasons to believe that the total number of protein domains which can be really used in a context (without creating a completely new reality) is not much bigger. I have no idea exactly how much, but if I had to guess I would stay around 10^3 or, at most, 10^4. But may they are much less. The probability of finding one of these domains is therefore extremely low, even accounting for individual target spaces and for realistic probabilistic resources. The probability of finding 4000 of them is not only mindblogging, it is complete folly.
Therefore, I infer that protein domains are designed. They come into existence through a rather “acute” input of information, and through implementation procedures which must be detailed in the future, in a relatively “short” biological time.
Moreover. they come into existence because they are specifically needed for vaster plans, which require the cooperation of various basic biochemical functions to realize higher levels of organization.
With the present proteome. most of what is necessary in the biological context we know is already present at basic biochemical level. More recent inputs of information have been especially at higher, regulatory levels.
b) Your model is: though the search space is huge, functional selectable proteins are extremely common. They are potentially present in huge numbers, and that explains why 400 of them have been found by random processes.
c) Finally, Zachriel seems to confide more in the concept that protein domains can gradually and functionally derive form much simpler functional peptides or small proteins.
As you can see, it’s all a question of competing models. The important thing is that all of these three models (or other variants) are scientific: they can indeed be falsified (let me be popperian for once).
And I believe that some of them will be falsified, in the next few years.
I have other things to say, but now I am tired, and I have no more time.
Toronto:
In brief:
“Life”, as I have said, is not a well defined concept.
Consciousness is the correct concept.
While human consciousness needs a complex interface (the brain) to interact with the physical world, consciousness can be (and IMO is) simple. It is the manifestation of a transcendental self which relates to himself the various complex representations which derive form its interactions with outer reality.
A consciousness which can interact with physical reality could do that in other ways, which do not require a complex physical interface like the brain. Therefore, it would not require special complexity.
What I am saying is not strange or new. For centuries philosophers have believed that a complex reality can come from something simple.
God, for many religious positions, is simple.
By the way, the Big Bang theory is not so different in essence.
gpuccio:
But “life” is what ID is concerned with and it is their concept I am using. ID says that life, this lump of matter we call Bob or Jane, because of their complexity, cannot have arisen without being designed.
This designer cannot be an element of the set of things we claim to have the attribute the ID movement calls “life”, since this designer is charged with creating all the elements that belong to that set.
Do you see your scientific problem here? The designer cannot share the attribute “life” since the designer must exist before the life he is charged with creating.
In this post, you are arguing the case for a designer, using protein as a topic. Protein is a material component of “life” as ID defines it. Everyone from kairosfocus to Dembski are debating material combinations and labeling them as “life”, and calculating the improbablity of their arising without a designer.
The label “conscious” doesn’t help, since anything having only that attribute could not affect anything around it. You need a material agent of some type to render your designs into the material world.
In order to be able to manipulate the material world in a way that the material world on it’s own cannot, you need something more complex than the material world itself.
This thing cannot be alive and it must be very complex.
If something so complex doesn’t need to be designed, why do we?
gpuccio: The neo darwinian model is a molecular model of RV and NS.
The “neodarwinian model” integrates evidence from population genetics, as well as from other related fields, and includes Common Descent and Natural Selection. However, there have been many advances in evolutionary biology since the modern synthesis was proposed in the first half of the last century.
gpuccio: I accept common descent, but not unguided common descent, nor gradual common descent. I believe in designed, “punctuated” common descent.
Okay. Punctuated in scare-quotes presumably distinguishes the term from Eldredge and Gould’s Punctuated Equilibrium which is a type of gradualism.
gpuccio: I don’t know what you mean by “adaptation”.
Adaptation refers to the process of organisms becoming more adapted to their environment through natural selection. It is an observed phenomena.
gpuccio: Exon shuffling is a valid hypothesis for the origin of multi-domain proteins. In no way it explains the origin of single domains.
New exons are posited to be created by mutations to introns, which are then shuffled to create new domains. Here’s a review article:
Schmidt & Davies, The origins of polypeptide domains, Bioessays 2007.
gpuccio: What has “macroscopic adaptation” to do with that?
Referring to adaptation of macroscopic structures.
gpuccio: And yes, my main point in this discussion is exactly “that there is an insurmountable Gap for evolution in the origin of protein domains.”
In order to address that concern, we should look at the cases where we have strong evidence of evolutionary processes. And that certainly includes macroscopic structures that leave fossils.
gpuccio: algorithms can at most incorporate some information from the outer world, and elaborate it according to the existing procedures.
We have a simple solution string, and an evolutionary algorithm. The solution string evolves and becomes complex in response to experience with the environment. How is the dFSCI not higher in the evolved string than the primordial string?
I’d be interested in knowing why infinite regress is not a problem.
Or why anyone should be taken seriously who asserts the existence of an entity for which there is no evidence, but which conveniently has the attribute of halting the regress.
Evolution does not explain origins. It deals only with the observed processes of change in populations of replicators.
Evolution and its feedback mechanism are not inventions; they are attributes of replicators. And replication with variation and selection can happen with rather simple chemistry. It doesn’t require life.
Arguing that we don’t know exactly how chemistry transitions to biology is just a God of the gaps argument. The key question is whether those believe that chemistry cannot transition to biology without intervention are willing to design and perform experiments to settle the question.
It’s really about the functionality of the hypothesis. Does ID lead to useful hypotheses? If not, it’s completely worthless.
One could, of course, perform all kind of experiments that led nowhere. Kind of like getting grant money to demonstrate that a technology is impractical. It’s a cinch.
The people who get remembered in science are those who make the breakthroughs that connect phenomena to uniform processes in nature. There are no examples of breakthroughs leading to supernatural causes.
If the designer does not use iterative variation and feedback to accumulate the knowledge of how his materials work, then what is the source of information?
We have lots of examples of evolution finding variations that work. What ID needs in order to achieve minimum credibility is an example of design in which the knowledge of materials and their attributes does not rest on evolved knowledge.
gpuccio: It’s a protein for which you can define an extremely trivial function, whose complexity is minimal.
That’s called handwaving. It’s a functional protein.
gpuccio: again, through intelligent selection.
It was already there.
Here’s a similiar study using random-sequence RNA molecules.
Bartel & Szostak, Isolation of new ribozymes from a large pool of random sequences, Science 1993.
gpuccio,
Here is a great point from Petrushka and one that needs to be addressed by the ID side.
By saying ID is scientific, ID is claiming to be able to explain something that exists in the natural world, but for the first and only time, this explanation is rooted “outside” the natural world according to ID.
Why does ID claim, that of all the scientific breakthroughs and knowledge we have aquired, for the first time, we must look outside of nature for the answer?
No other scientific answer man has sought, has required us to look outside of the environment where the question was asked.
Why is this one different?
I think, at the very least, that ID proponents need to put forth some hypothesis regardings the designer’s source of knowledge or information.
Since they deny that information can accumulate through iterations of variation and feedback, then I would like to know its source.
It’s no problem if ID claims the designer is God, because by definition, God can poof anything into existence.
But in the absence of magic, how does the designer know?
Toronto:
Please, be more precise. ID is about biological information, not about life. Although biological information is found in living beings, it does not require life to exist. A protein is not alive, and still retains its information.
Consciousness certainly needs some interface to interact with the material world, but why should that interface be necessarily complex? That is only your personal assumption.
“In order to be able to manipulate the material world in a way that the material world on it’s own cannot, you need something more complex than the material world itself.”
Again, only your assumption. It’s not necessary. It’s not true. dFSCI comes from consciousness, not from complexity. Even the most complex computer cannot create new, original dFSCI. A human can. You may say that is because he is more complex. I say that is because he is conscious.We are both entitled to our opinion, but there is no reason that yours is anything more than an opinion.
Gpuccio – I am sorry but dFSCI is not stuff that can be created or correlated with intelligence. It is a number indicating a probability relationship between a digital string, a function that string might perform, and known explanations for the creation of that string. That number can change if any of those three factors change.
Petrushka:
“I’d be interested in knowing why infinite regress is not a problem.”
I have never said it’s not a problem. It is a well defined, but not solved, philosophical problem. It is not a scientific problem.
“Or why anyone should be taken seriously who asserts the existence of an entity for which there is no evidence, but which conveniently has the attribute of halting the regress.”
Because that’s the most common philosophical asnwer to the problem of infinite regress. And, IMO, the best one. You may not like it. You are not obliged to agree. But why would you state that nobody who asserts that should be “taken seriously”? So, according to Petrushka, none of the many philosophers who have believed that in centuries should be “taken seriosuly”? I am really amazed at what scientism has done to the human mind.
And by the way, that “there is no evidence” is your statement, and nothing else.
“Evolution does not explain origins. It deals only with the observed processes of change in populations of replicators.”
Science must explain origins. OOL and evolution are deeply connected problems.
“Evolution and its feedback mechanism are not inventions; they are attributes of replicators. And replication with variation and selection can happen with rather simple chemistry. It doesn’t require life.”
And it does not produce complex functional information.
“Arguing that we don’t know exactly how chemistry transitions to biology is just a God of the gaps argument. The key question is whether those believe that chemistry cannot transition to biology without intervention are willing to design and perform experiments to settle the question.”
No. And yes.
“It’s really about the functionality of the hypothesis. Does ID lead to useful hypotheses? ”
Yes. I have just described three different competing models for the origin of protein domains, each with different implications.
“The people who get remembered in science are those who make the breakthroughs that connect phenomena to uniform processes in nature. There are no examples of breakthroughs leading to supernatural causes.”
I don’t think I will comment anymore about any phrase with the word “supernatural” in it. There are limits.
“If the designer does not use iterative variation and feedback to accumulate the knowledge of how his materials work, then what is the source of information?”
It’s you who seem to have decided that he doesn’t. I just don’t know, and am willing to inquire.
“We have lots of examples of evolution finding variations that work. What ID needs in order to achieve minimum credibility is an example of design in which the knowledge of materials and their attributes does not rest on evolved knowledge.”
I am still waiting for a single example of molecular macroevolution well explained by an explicit darwinian model. From you, or Zachriel, or anyone else. ID is perfectly credible, and does not need to achieve anything. Darwinist need to achieve a minimum of unbiased, non fanatic attitude.
Toronto:
I have written about “natural” and “naturalism” for years at UD. You will forgive me if I am tired and bored of that, and don’t want to repeat everything here. I repeat what I have said to Petrushka: I refuse to use words which mean nothing, and to comment on them.
Zachriel:
For the nth time, I ask you one example of molecular macroevolution explained by the darwinian model, and for the nth time you shift to “macroscopic evolution” or to some very vague and generic paper which contains no real example. What’s the problem with the flaunted “overwhelming evidence”?
Zachriel:
“We have a simple solution string, and an evolutionary algorithm. The solution string evolves and becomes complex in response to experience with the environment. How is the dFSCI not higher in the evolved string than the primordial string?”
Please, refer to Dembski and Marks for hidden information in evolutionary algorithms.
Petrushka:
What’s wrong with you?:
“Since they deny that information can accumulate through iterations of variation and feedback, then I would like to know its source.”
Information can certainly accumulate through iterations of variation and feedback, and a conscious intelligent being who represents what is happening, understands it, makes choices, has purposes, and guides the process.
Clear?
Information can certainly accumulate through iterations of variation and feedback, and a conscious intelligent being who represents what is happening, understands it, makes choices, has purposes, and guides the process.
___________________________
Other than producing differential reproductive success, what exactly does your being do?
Can you provide even thought example?
Mark:
A brief remark for you, then I’ll go to sleep.
How do you explain, according to your model, the following two simple facts?
1) About 2000 protein families which emerged before LUCA
2) About 500 protein families which emerged at the transition from prokaryotes to eukatyotes
That, out of 4000 in total.
Just to have a real example, take one of the aminoacyl-trna synthetases, the proteins which are essential for translation (they couple each aminoacid to its correct trna, and therefore are the depositaries of the symbolic information in the code).
This is one of the 2000 domains which originated between OOL and LUCA, before the division of bacteria from archaea.
Why? It is rather simple. I have blasted the protein in E. coli against archaea. Obtaining a series of homologies around 40-45% of identities. In a protein of about 800 aminoacids (800!), that corresponds to an E value (improbability of random homology) so low that, in the first hits, it is given as 0, and then goes up to… 2e-178
IOWs, the bacterial protein and the archaea protein (which, let’s remeber, have the same function) are certainly related. That means that their common ancestor (which obviously had the same function) originated before the separation of bacteria and archea, usually put at something like 3.5 billion years ago.
We could repeat the same reasoning for about 2000 protein families.
Gpuccio
I hope you sleep well. You deserve it. This is a marathon effort in hostile territory.
That is indeed a better argument than just saying why haven’t we discovered more protein families. As I keep on emphasising I am not a biologist but a brief scan of the literature suggests:
The evolutionary history of archea and bacteria is still very uncertain. There are even some who believe that archea evolved from bacteria.
There is evidence for horizontal gene transfer between archea and bacteria. So a protein that originated quite late on in one kingdom could transfer to the other.
It also occurs to me that early life would allow for more options for discovering new protein families for reasons that you yourself have given. Once evolution has stumbled upon a protein family that provides additional fitness that discovery creates an environment that inhibits other options that might otherwise been available. The new species will have created complex relationships within the cell that another protein will find hard to exploit and different species based on different protein families will find it hard to get established in an environment where all the niches are already occupied. It is the molecular equivalent of the expansion of new species that takes place after mass extinctions. In the first billion years there were lots of molecular niches that gradually got filled.
I appreciate this is pure conjecture. But so is the proposal that some unknown force made it that way. However, the first conjecture has enough detail that one can begin to examine the evidence (as you have done). The second proposal is so lacking in detail that the only evidence for it is the evidence against alternatives.
gpuccio: For the nth time, I ask you one example of molecular macroevolution explained by the darwinian model, and for the nth time you shift to “macroscopic evolution” or to some very vague and generic paper which contains no real example. What’s the problem with the flaunted “overwhelming evidence”?
You made a claim that “there is an insurmountable Gap for evolution in the origin of protein domains.” Even if no one can explain the origin of protein domains, that doesn’t lend support to your position, nor would it impact the consilience of evidence in support of the Theory of Evolution.
Furthermore, we know that functional proteins occur in random amino acid sequences, so their distribution is not that rare. Given an environment rich in peptides, there is no reason to believe they couldn’t form naturally.
–
You’ve also claimed that “new dFSCI cannot come from existing algorithms.” And when asked, you said this,
gpuccio: Please, refer to Dembski and Marks for hidden information in evolutionary algorithms.
So you base your claim on a paper by Dembski and Marks? They don’t seem to have gained much traction in the mathematical community. If the paper is intrinsic to your claim, then we would be happy to engage that point.
gpuccio: For the nth time, I ask you one example of molecular macroevolution explained by the darwinian model, and for the nth time you shift to “macroscopic evolution” or to some very vague and generic paper which contains no real example.
Don’t the wide variations within protein families count as “macroscopic evolution”?
The problem is with your question. “Darwinian model” is vague. We know about Common Descent, which you accept. We also know of complex structures that have formed through natural selection, and we can model these processes. This you reject.
This evidence is crucial to understanding the evolution of protein families because we can’t always reconstruct the entire series of events. So we start with what we can establish with some certainty. You can’t just wave away the known evidence.
Don’t the wide variations within protein families count as “macroscopic evolution”?
Should read “molecular macroevolution”.
gpuccio:
Then let’s not use a word at all.
Let’s simply draw a circle around us and everything material.
Every single question science has ever answered, was answered with resources and information from within that circle.
Why do we now have to make an exception and look outside?
“Why do we now have to make an exception and look outside?”
_________________________________________________
I find it laughable that a whole chorus of people assert they are being scientific, but invoke magic every time they encounter something that is yet unexplained.
This kind of knee jerk assertion that the unexplained is unexplainable except by reference to gods or demiurges, is exactly what science was invented to exclude.
If you start with the assumption that something cannot be natural, you are effectively saying that research is futile. A case that has never been observed in the history of science.
I am in moderation at UD for saying this. I was a little more blunt. I asserted that one of the sacred cows there was ignorant of the history of science, because he misrepresented this point.
All I ask of Id is some scenario under which ID as understood by its proponents, is possible. If biological information is not the result of accumulated feedback, what is the source? If the designer is finite and not an omniscient god, what is the source of the designer’s information?
Even a hypothetical scenario would be interesting.
Mark:
I agree that “The evolutionary history of archea and bacteria is still very uncertain.” I have no pretence to solve the problems implied. I have just adopted what is the most accepted view, and I have made the point that a design model explain it better than the other models. Which is my reason to prefer it (or at least, one of the reasons).
Conjectures are the essence of science. Their first requisite is that they must explain. That’s where the darwwinian conjectures fail, unless one adopts a very vague concept of “explanation”. The design model is explanatory, because design can generate new dFSCI in great abundance.
You complain, not for the first time, that the design model “is so lacking in detail that the only evidence for it is the evidence against alternatives.” But I don’t think it’s true.
First of all, many different models based on design can be proposed. I have tried to propose mine for the emergence on protein domain, in one of my posts here. I paste it again:
“My model is more or less: The functional domains which can really be useful in a biological context are extremely rare. The existing proteome exhibits approximately 4000 of them. I have all the reasons to believe that the total number of protein domains which can be really used in a context (without creating a completely new reality) is not much bigger. I have no idea exactly how much, but if I had to guess I would stay around 10^3 or, at most, 10^4. But may they are much less. The probability of finding one of these domains is therefore extremely low, even accounting for individual target spaces and for realistic probabilistic resources. The probability of finding 4000 of them is not only mindblogging, it is complete folly.”
As you can see, this model has many implications which re definitely different from what one could expect from a darwinain model. I believe that the facts we already have support this model. I have tried to explain why. But, as facts continue to accumulate, they could as well falsify mu model. IOWs, my model is “popperianly correct” 😉 .
Gpuccio
Your “model” is not a model of how domains were created by design. It is simply a conjecture as to how many domains there are. As a model the only implication is that if you chose a string of amino acids at “random” the chances of finding one in a possible domain is mindboggling small. It is doing exactly what I said – offering evidence against an alternative.
Petrushka:
First your last post, then the previous one.
Now that I am less tired, I feel more inclined to explain again why I don’t accept the concept of “natural”.
Nobody can say what “nature” is.
Try a definition which is not in itself an “a priori” assumption of some specific world view.
IOWs, “nature” and “natural” are not scientific, sharable concepts. They are rather purely philosophical concepts, and each thinker is free to adopt the meaning he likes, or simply to reject them (as I do) as useless.
Let’s discuss it a little more clearly.
Science is a map of reality. It is not reality. Reality is the territory. The map is not the territory.
OK? Well, how can you define “nature”.
If you say that “nature” is all that exists, you are conflating nature with reality. I prefer to use ” reality”. If God exists, the, He is part of reality. And therefore of nature, in this wide sense.
Obviously, you can separate the concept of nature from the concept of reality, and define nature as a subset of reality which has to have other properties, beyond existing.
But what are those properties? You can verify that, whatever you choose, you are implying some specific general worldview, and very specific philosophical choices, that need not be shared by all. Indeed, that are not shared by all. At all.
Materialist reductionism could define “mature” as all that is material. But that is really wrong, because even the most extreme reductionist would admit that there are principles which are not matter (anything without a mass, I suppose). So, to do better, we could say “all that is physical”. But what does “physical” mean? Nothing. Its only meaning can be “anything which can be explained by physical laws”.
But what are physical laws? Again, you can’t define the concept, unless you say: “whatever is not in conflict with our current map of the universe”.
Well, this is the truly important point. The only meaning of “nature” in the reductionist position, is: anything which is not in essential contrast with out general map.
That does not sound good. To me, that is only an encouragement of intellectual conformism, and a truly anti-scientific concept.
That’s why I reject the concept of “nature”.
In the same way, you define “magic” all that is not included in your personal map. Well, in that sense I would accept the “magics” concept as a true compliment.
Petrushka:
Thought example.
The designer, being a conscious intelligent being, has conscious intelligent representations of reality. IOWs, he is a cognitive being. So, he is aware of what exists, of his possibilities of interaction with what exists, and probably of many of the laws governing what exists. That’s not essentially different form our condition, except that I assume that the interface is different, and does not require a complex brain.
Moreover, the designer has purposes and desires. He can plan and imagine and give judgements about possible plans. He can will to implement some specific plan. Just like us. IOWs, he is a purposeful being.
Finally, the designer can act. IOws, he is an agent. Just like us. Only, he acts through different channels. But the principle is the same: he can project his conscious plans into matter. He can set “configurable switches” to give form to matter.
How? I will accept, as a tentative model, Eccles’ideas about the mind-brain interaction: let’s say that the designer can interact with outer reality exactly as we interact with our brain: by a direct control exerted by consciousness at the quantum level.
We can do that only with our personal brain. He can do that with a wider physical scenario.
So, what does he do?
There are different possibilities.
Let’s try one.
First of all, he “conceives” life. In terms of what it could be, and of what purposes it could serve.
Then he “plans” life. He conceives some more detailed plan of how life could work. That plan need not include all the details.
Then, he “implements” life. That can be a long an patient task, but I suppose that time and hurry have not the same meaning for the designer as they have for us.
So, he intervenes on the reality of out planet at the best time he can see.
He can do many things.
He can set a special environment, separated from the general environment (a “lab”?).
He can set material conditions there.
He can find the information necessary at basic molecular level to implement the simplest life. There are two different possibilities for that, which are not mutually exclusive:
a) he can derive part of his information from his understanding of the general laws of the universe.
b) he can derive part of the information from experimentation in his “protected environment”: IOWs he can implement trial and error processes to perfect any initial implementation, observe the results, understand them, and use the gained feedback to implement new information. he can also, definitely, use intelligent selection to fix and expand the results which are in line with his expectations.
c) if he fails, he can try again.
Well, I think that could do for a start.
Zachriel:
“gpuccio: It’s a protein for which you can define an extremely trivial function, whose complexity is minimal.
That’s called handwaving. It’s a functional protein.”
No, that’s calling things for what they are. It’s you who are using the word “functional” as though it had in itself important implications. But that’s not the case.
If you go back to my definition of dFSCI, and to the following discussion, it should be clear that one can define some function (indeed, many functions) for anything.
For instance, I may need a molecule which is an organic acid. In that sense, any protein which is an organic acid would satisy the requirement, and would be a functional protein according to that definition.
IOWs, as should be clear after so many posts, “function” is a relative concept, not an absolute one (as Mark has very correctly remarked).
Therefore, just being “a functional protein” means nothing.
The only functions which are “interesting” for our debate here are functions which:
a) are useful in a definite biological context
b) are complex
The original ability to stick to ATP in the original molecules is useless and simple. It means nothing.
Your only argument could be that such a trivial function can be in some way a precursor of useful and complex functions through NS (I can’t see how, given that it is obviously non naturally selectable): but it is your burden to demonstrate such an odd thing.
From the paper, the only thing we know if that such a useless and simple function can be transformed into the same, equally useless, but slightly more complex function, through a few rounds of intelligent protein engineering.
That’s all.
Zachriel;
“gpuccio: For the nth time, I ask you one example of molecular macroevolution explained by the darwinian model, and for the nth time you shift to “macroscopic evolution” or to some very vague and generic paper which contains no real example. What’s the problem with the flaunted “overwhelming evidence”?
You made a claim that “there is an insurmountable Gap for evolution in the origin of protein domains.” Even if no one can explain the origin of protein domains, that doesn’t lend support to your position, nor would it impact the consilience of evidence in support of the Theory of Evolution.
Furthermore, we know that functional proteins occur in random amino acid sequences, so their distribution is not that rare. Given an environment rich in peptides, there is no reason to believe they couldn’t form naturally.”
IOWs, You cannot give any example of “molecular macroevolution explained by the darwinian model”.
Zachriel:
“So you base your claim on a paper by Dembski and Marks?”
Yes, in part.
“They don’t seem to have gained much traction in the mathematical community.”
That’s not a problem for me. I am a minority guy.
“If the paper is intrinsic to your claim, then we would be happy to engage that point.”
I would too, but at present I believe we already have too much at stake. And it’s not eaxctly my filed. Anyway, if you have some major comment on the subject, feel free to express it, and we’ll see.
Zachriel:
Correction:
“And it’s not exactly my field.”
Toronto:
“Why do we now have to make an exception and look outside?”
Please, refer to my answer to Petrushka for that.
Zachriel:
No. They keep the same basci function. They are probably the effect of neutral variation in the island of function, and possibly in some cases of tweaking if the protein for different molecular contexts via microevolutionary events (what we could call microevolutionary adaptation, which in principle could be achieved by darwinian mechanisms).
Even when different final functions are achieved in a superfamily through minor variations in the active site and in the affinity for ligands, as could be the case in the superfamily of nuclear receptors studied in a recent paper, and in the case of nylonase, he events are microevolutionary, and not complex. A case for design could be still made in these cases, but it would require an analysis of the complexity of the whole system, and not of the single molecule. I will abstain from that, for the moment.
For a very convincing model (IMO) of how protein domains appear suddenly, and then explore their island of function through neutral mutations throughout natural history, please see the following paper about the “big bang theory of protein evolution” (Sequence space and the ongoing expansion of the protein universe):
http://www.nature.com/nature/journal/v465/n7300/full/nature09105.html
Zachriel:
The previous post was an answer to you question:
“Don’t the wide variations within protein families count as “molecular macroevolution”?
Zachriel:
“The problem is with your question. “Darwinian model” is vague. We know about Common Descent, which you accept. We also know of complex structures that have formed through natural selection, and we can model these processes. This you reject. ”
I just say that you cannot use macroscopic data whoe molecular basis is unknown in favour of the neo darwininian molecular model.
The explanatory part in the neo-darwinian model is the RV + NS model. Genetic modifications are the cause of what we observe. It’s those genetic modifications which must be explained.
Zachriel:
“This evidence is crucial to understanding the evolution of protein families because we can’t always reconstruct the entire series of events. ”
It would be better to say that we can never reconstruct any part of the supposed sequence.
“So we start with what we can establish with some certainty. You can’t just wave away the known evidence.”
I have no intention to wave away any evidence. I deeply respect evidence. I am just saying that that evidence is not ecifence in favour of the darwinian molecular model, and therefore it is not causally explanatory. That is a comment about the explanation of evidence, not about evidence itself. Wrong explanations of evidence can certainly be waved away.
Mark:
“As a model the only implication is that if you chose a string of amino acids at “random” the chances of finding one in a possible domain is mindboggling small. It is doing exactly what I said – offering evidence against an alternative.”
All biological sciences work that way: they reject a null hypothesis, and affirm the best explanatory model available. It’s called fisher’s hypothesis testing procedure. There would be no biological empirical science without that procedure.
Edited this comment for greater clarify (I hope)
Gpuccio
First – note that you have not refuted my assertion that the only evidence for design is the evidence against alternatives. All you have done is assert that evidence against the alternatives counts as evidence for it.
However, your comment is confused on a number of counts. Some relatively unimportant, some fundamental.
Fisher’s procedure was simply to accept or reject a hypothesis based on extreme values – there is no concept of an alternative hypothesis. Although a breakthrough at the time it is now rejected by pretty much all statisticians and philosophers of science as inadequate. It was superceded by the Neyman-Pearson approach which is probably the one you are thinking of. This identifies a hypothesis which we have some other reason to suppose is true and opposes it to a null hypothesis. This is also controversial and has widely criticised and rejected in favour of Bayesian approaches (see for example http://onlinelibrary.wiley.com/doi/10.1002/wcs.72/abstract). However, whichever approach you favour there has to be some reason for believing the alternative hypothesis other than the rejection of alternatives. Otherwise I could use the rejection of “Darwinism” as support for any other alternative e.g. Larmarckism.
Toronto:
Just as an example of my point:
“Let’s simply draw a circle around us and everything material.”
You are simply implying that we are material. Do you think that is a light implication? One that must be shared by all?
gpuccio:
First of all, I’d like to thank you for hanging around and debating with us.
The material question asked by ID is, “Can the matter, organized as us, have come about without the help of something that does NOT consist of matter?”
That statement has no “natural”, “supernatural”, “magical” or “philosophical” reference at all.
If the answer is no, propose a likely non-matter process that guided the organization of the matter that ended up as us. We can all then look at it, evaluate, study and maybe accept it.
You would then be able to answer the point of this post, “Why no intermediates?”, because you could show, with your ID process, that they may not necessary.
That would make ID a solid contributor to our pool of knowledge.
Otherwise, ID as a theory, takes our efforts but never returns results.
Correction:
You would then be able to answer the point of this post, “Why no intermediates?”, because you could show, with your ID process, that they may not be necessary.
gpuccio: And yes, my main point in this discussion is exactly “that there is an insurmountable Gap for evolution in the origin of protein domains.”
Sometime later.
gpuccio: The functional domains which can really be useful in a biological context are extremely rare.
Do you realize that your goal posts have slipped?
gpuccio: It’s a protein for which you can define an extremely trivial function, whose complexity is minimal.
Zachriel: That’s called handwaving. It’s a functional protein.
gpuccio: No, that’s calling things for what they are. It’s you who are using the word “functional” as though it had in itself important implications.
Odd that. Most microbiologists consider binding an important protein function. Indeed, that’s the word used in the several papers cited above.
In any case, it is a protein domain — which was the question. Do you think this is the only type of function discoverable in random sequence proteins?
gpuccio: IOWs, You cannot give any example of “molecular macroevolution explained by the darwinian model”.
In other words, you can’t support your position which is that “there is an insurmountable Gap for evolution in the origin of protein domains,” but rely upon supposed Gaps in alternative mechanisms.
As we explained the “darwinian model” depends on a consilience of evidence from multiple fields of study, including evidence of adaptation in macroscopic structures which demonstrate that complex features can evolve in selectable fashion.
Finally, we point to the wide variations in structure and function within protein families as examples of “molecular macroevolution.”
Zachriel: So you base your claim on a paper by Dembski and Marks?
gpuccio: Yes, in part.
Zachriel: They don’t seem to have gained much traction in the mathematical community.
gpuccio: That’s not a problem for me. I am a minority guy
If the paper was widely accepted in the mathematical community, we might accept it at face value. However, it isn’t. That means the claim is contingent on arguing its conclusions on the merits.
gpuccio: please see the following paper about the “big bang theory of protein evolution” (Sequence space and the ongoing expansion of the protein universe):
“big bang theory of protein evolution”
gpuccio: http://www.nature.com/nature/journal/v465/n7300/full/nature09105.html
Nothing in Povolotskaya & Kondrashov’s paper seems to support your position. Indeed, they conclude “Our analysis shows that it is conceivable that many more proteins were present in LUCA but have since diverged beyond our ability to detect their homology, and that given enough time some of the currently identifiable orthologues among the major kingdoms of life will diverge beyond recognition. Finally, our observation of receding protein sequences provides novel evidence of the common ancestry of life.”
In other words, they explain why homologies are difficult to discern, and their results are consistent with a world of more primitive polypeptides that precedes the LUCA.
gpuccio: The explanatory part in the neo-darwinian model is the RV + NS model. Genetic modifications are the cause of what we observe. It’s those genetic modifications which must be explained.
Any theory has to be consistent with genetics, but neodarwinism isn’t just genetics, but the unification of data from disparate fields, including macroscopic evidence. That’s why it is also called the modern synthesis. (Though modern is relative. All its inventors have been dead for many years. We may use the term to refer to the current Theory of Evolution, but it has been substantially modified and expanded.) In any case, you can’t just ignore the evidence, which includes strong support for complex adaptation by natural selection. If you say there is no such evidence, then we have to grapple with that, because that evidence is crucial to understanding the molecular evidence.
This is the basic outline. We have evidence of genetics, the distinction between genotype and phenotype. We have evidence of common descent. We have evidence of adaptation by natural selection. We have evidence of common descent among protein families and superfamilies.We have evidence that at least some protein domains can be discovered by stochastic methods. We have tentative evidence of exonization leading to novel domains. And you yourself cited a paper that explained why we may have difficulty resolving the gross phylogeny of proteins.
In other words, the perceived Gaps are consistent with what we know of protein evolution.
uncommondescentdissent: First of all, I’d like to thank {gpuccio} for hanging around and debating with us.
We would like to echo that.
Dembski & Marks: Conservation of information theorems indicate that any search algorithm performs, on average, as well as random search without replacement unless it takes advantage of problem-specific information about the search target or the search-space structure.
The very first line of the paper is presumes someone attempting to match search algorithms with search spaces. No search algorithm will work better for all abstract landscapes than any other search algorithm. But the vast majority of abstract landscapes are highly chaotic and disordered. And the vast majority of search algorithms are likewise highly chaotic.
Saying we have to take advantage of problem-specific information is highly misleading in an evolutionary context. A particular search algorithm may work better in a particular landscape. A simple replicator taking advantage of local resources may be very successful in a world organized by locality. It doesn’t require “taking advantage of problem-specific information about the search target or the search-space structure.” It’s a consequence of features of the natural world.
It’s sort of like saying rivers can never find the ocean better than any other search algorithm. But the world is not mathematical chaos, but highly ordered. There’s apparently a lot of “active information” in the natural landscape.
There’s apparently a lot of “active information” in the natural landscape.
______________________________________
And a landscape that’s a lot flatter than imagined by ID.
The slope of functionality is actually the first and only assertion made by ID that is remotely scientific (in the sense that it is testable).
Unfortunately for ID the slope seems to be quite climbable, at least in the foothills. And if it steepens hyperbolically, that would be completely consistent with the observation that innovations thins out with time after mass extinctions.
Toronto:
“If the answer is no, propose a likely non-matter process that guided the organization of the matter that ended up as us.”
I have proposed it in my answer to Petrushka.
Zachriel:
“Do you realize that your goal posts have slipped?”
No. Please, explain.
“Do you think this is the only type of function discoverable in random sequence proteins?”
Yes. Trivial, useless functions of no real complexity.
And binding is useful only when it can determine functional outcomes in the cell context. Binding can be useful in different ways: the oxygen binding of myoglobin and hemoglobin is useful because it can store and release the ligand in different, specific conditions. The binding of an enzyme to its substrate is useful because it can catalyze a necessary reaction. The binding in a cascade can transmit a signal. The biinding of the right aminoacid to the right trna, catalyzed by the right aminoacyl trna synthetase, is the key to the translation process.
Binding in itself means nothing. When the final Szostak protein was added to a real biological system, the only result was to create ATP deprivation.
Zachriel:
“In other words, you can’t support your position which is that “there is an insurmountable Gap for evolution in the origin of protein domains,” but rely upon supposed Gaps in alternative mechanisms.”
This is really beyond my understanding. What do you mean? As scarce clarity is not usually a characteristic of yours, I suspect that it may be a consequence of scarce arguments 🙂 (just joking, no offence intended).
“As we explained the “darwinian model” depends on a consilience of evidence from multiple fields of study,”
except, it seems, molecular biology…
” including evidence of adaptation in macroscopic structures which demonstrate that complex features can evolve in selectable fashion.”
without explaining how.
“Finally, we point to the wide variations in structure and function within protein families as examples of “molecular macroevolution.””
Wrong. As I have already said, that’s mainly neutral variation, which does not modify folding and function. It is evidence of CD, but not of macroevolution. No new function, and above all no new complex function, emerges.
IOWs, all the “overwhelming evidence” for neo-darwinism is only evidence for CD.
Zachriel:
“If the paper was widely accepted in the mathematical community, we might accept it at face value. However, it isn’t. That means the claim is contingent on arguing its conclusions on the merits.”
And that we can do. But please, you start.
“In other words, they explain why homologies are difficult to discern, and their results are consistent with a world of more primitive polypeptides that precedes the LUCA.”
No. Their model implies an emergence of protein domain as single ancestors, and then a diversification due to neutral mutations which is kept under the strict boundaries of their functional island by negative NS. There is no explanation of how the single ancestors of each protein family arise. IOWs, the homologies which are the basis for the “overwhelming evidence” are indeed homologies between equivalent molecules, or only slightly different ones (functionally). THat explains why even different sequences can have the same folding and function: neutral mutations have changed up to a point the primary structure, but the folding and function have been preserved by negative NS.
Zachriel:
“This is the basic outline. We have evidence of genetics, the distinction between genotype and phenotype.”
What do you mean? That the only evidence form genetics is “the distinction between genotype and phenotype”. THat’s a strange concept. And in what way would that disctinction be evidence for the darwinian model (classical, neo or neo-neo)?
“We have evidence of common descent.”
True. Which in no way is evidence of the darwinian model (classical, etc).
“We have evidence of adaptation by natural selection.”
That is not very precise. We have certainly some limited evidence of adaptation. We have some evidence of negative NS, and very limited evidence of positive NS in rare cases of simple mutations and extreme selective pressure (what we usually call “microevolution”). I am not so sure that all cases of “adaptation” can be explained as RV + NS. The cause of adaptation, IMO, can probably be found, in many cases, in adaptational intelligent algorithms embedded in the existing genome. Epigenetics too? Maybe.
“We have evidence that at least some protein domains can be discovered by stochastic methods. ”
Already commented on that. Nothing to add.
“We have tentative evidence of exonization leading to novel domains.”
I can accept that with the “tentative”. Attempts are always welcome.
“And you yourself cited a paper that explained why we may have difficulty resolving the gross phylogeny of proteins.”
There are certainly difficulties. No reason to deny that. And great is the merit of people who work hard to gather new facts, whatever their ideology. I am grateful to them.
“In other words, the perceived Gaps are consistent with what we know of protein evolution.”
That is your sincere opinion, and I respect it. But I disagree.
Zachriel:
“uncommondescentdissent: First of all, I’d like to thank {gpuccio} for hanging around and debating with us.
We would like to echo that.”
Thank you Zachriel and Toronto. And, obviously, thank you Mark for having been such a perfect host.
I am very happy of the discussion here. I have enjoyed it very much. Certainly, it was hard work (for you too, obviously).
I must say that I have profited of a few days of not too much work, and of the relative lack of stimulating threads at UD (at least for me). I am afraid that this cannot go on forever, but as far as the discussion remains stimulating (and it is stimulating) I can try to endure a little more.
I want to say that it is really a pity that you guys cannot post at UD (at least Zachriel, if I understand correctly). I can see no justification for that.
gpuccio:
I sense a hesitancy on your part to engage, so I’ll leave you with one last question and many thanks.
I’d like a suggested process that meets this level of performance;
Q: “Suggest a process to turn paper white, something I can try.”
A: “I think immersing the paper in bleach would turn it white.”
So here’s the question.
Propose a real non-matter process that guided the organization of the matter that ended up as us, something we could try.
Zachriel:
I essentially agree with you on many points in your post about information and algorithm. I am sure that you understand well these problems. But still, I think you miss some important points.
First of all, you apparently underestimate the words “on average” in Dembski and Marks’ statement. They are not saying that occasionally a search algorithm cannot work better than random search. But that is a random occurrence.
A second comment about “But the world is not mathematical chaos, but highly ordered. There’s apparently a lot of “active information” in the natural landscape.”.
I agree. But, except for a general “permissivity” to life (which would make happy theistic evolutionists), I don’t believe that such an information is in any way specific to solve the complex search problems at molecular level. The world does not know which AA sequence will determine a functional folding. The world does not know what biochemical reactions can build a complex network of interactions, to allow life. The world knows nothing of codes and transcription and translation. And so on.
So, while the world has certainly a lot of “active information” in it, it is not specifically useful and pertinent to solve the problem of biological information.
Petrushka:
“Unfortunately for ID the slope seems to be quite climbable, at least in the foothills. And if it steepens hyperbolically, that would be completely consistent with the observation that innovations thins out with time after mass extinctions.”
I suppose that you are referring here to the paper about “rugged landscape”.
I must thank you for pointing that paper to me. It has become one of my favourite.
I understand I have not yet given you any satisfaction about it: that is due mainly to reasons of time, because a complete analysis requires a very detailed discussion.
Anyway, I will give you here two important hints about my position. Just briefly, we can discuss them in detail later.
First of all, the paper:
http://www.plosone.org/article/fetchObjectAttachment.action;jsessionid=7432C19FA23F2AAFC8F72EC64AB380B5.ambra02?uri=info%3Adoi%2F10.1371%2Fjournal.pone.0000096&representation=PDF
And now my two points:
1) If you read the paper carefully, you will see that the “rugged landscape” in no way implies that the results obtained will lead to further optimization. IOWs, they are not steps to the main peak of functionality. Indeed, they are detrimental to finding it, because they represent local peaks, and therefore negative NS would prevent a walk form them to other peaks. So, in no way are they “foothills”. They are rather isolated small mountains.
From the discussion in the paper:
“More than one such mountain exists in the fitness landscape of
the function for the D2 domain in phage infectivity. The sequence
selected finally at the 20th generation has ~W=0.52 but showed no
homology to the wild-type D2 domain, which was located around
the fitness of the global peak. The two sequences would show
significant homology around 52% if they were located on the same
mountain. Therefore, they seem to have climbed up different
mountains.”
2) Please look at this part of the discussion:
“The question remains regarding how large a population is
required to reach the fitness of the wild-type phage. The relative fitness of the wild-type phage, or rather the native D2 domain, is almost equivalent to the global peak of the fitness landscape. By extrapolation, we estimated that adaptive walking requires a library size of 10^70 with 35 substitutions to reach comparable fitness.”
A library of 10^70? 232 bits? Why does that remind me of something?
dFSCI? Axe?
Toronto:
“Propose a real non-matter process that guided the organization of the matter that ended up as us, something we could try.”
We are already doing it. Protein engineering is a good example (well, not yet so succesful, but it will come). Targeted mutation and artificial selection are powerful tool.
And also a top down calculation based on a deep understanding of biochemical laws can solve many of those those problems. We are still very inefficient at that, but everybody believs it can be done (except maybe Petrushka).
As I have explained in my answer to Petrushka, I think of similar tools for the designer. The main difference would be in the interface: an interface consciousness->brain->body->matter for us; an interface consciousness->matter for the designer of biological information.
But you see, for me consciousness and design “are” non matter processes. Even in humans they are. So, I can’t see any true essential difference (except for the interface).
I sense this thread is winding down now. I join the others in thanking and admiring Gpuccio for his perseverance and good humour.
In addition to the subject matter I have a meta interest in Science communication and the scope for blogs to facilitate discussion. A couple of things I noticed:
I would love to get from feedback from any of the participants
* why they participated – 134 comments is a lot of time and effort
* what did they get out of participating (if anything) and in particular did they learn anything of value
* how did this little debate compare to other blog discussions
* what would have made the exchange more fruitful (besides total capitulation from the other side)
Thanks
Mark
The topics that start these exchanges are always secondary to the real topic just underneath.
I believe that topic is; “Our churches should be the prominent influence in our society, both personal and public.”
I believe that to be completely wrong but that is the goal of the ID side.
For instance, “Darwin, Therefore Hitler”, “There Is An Absolute Moral Code”, and “The Universe Is Fine Tuned For Life” contain the goals of the ID side. That goal is to show us there is a cosmic entity we need to deal with.
Even Dembski is not immune as he had to re-think “Noah’s Flood” as a historical global event even though scientifically, he believes it to be a local event.
I’ve heard some great scientific arguments from our side that simply get waved away, in a way no scientist actually looking for an unknown answer would do.
That’s because they “know the answer” and in time they would like that “answer” to be taught in schools to our children so that it influences everything in their material lives.
I believe it is very important to engage and understand how to reply to law makers, school boards and voters to prevent their version of “how things have to be done” from taking control of our society.
I don’t believe them to be bad people, just very frightened of their own mortality. They want to be able to escape death and if that means “behaving like they’re supposed to”, they’ll do that at any cost, even if that includes reaching into your neighbour’s life to accomplish their goals.
It was a great exchange I think, because of the people involved. I listen to some of them and I see how scientifically ignorant I am in so many areas.
I wish the other side would admit to their lack of knowledge too, but they can’t.
Msrk,
I have to disagree with you, this was not a great exchange. You are giving gpuccio far more credit for intellectual honesty and integrity than he deserves. He has not defended a single one of his claims with anything close to mathematical or scientific rigor.
If you want to improve the quality of these kinds of interactions, you need to set clear ground rules for what is on topic in the context of the thread and you need to continue to hold the intelligent design creationists’ feet to the fire by tying their statements back to the topic, demanding empirical evidence, and challenging all the handwaving that typically ensues.
As I predicted, gpuccio has not rigorously defined his terms, nor has he shown how to calculate CSI for a real biological system. In addition, he has introduced some ridiculous ideas rooted in his support of dualism, with no evidence to back them up.
ID is not a scientific endeavor and the goals of the IDCists are not to find out the truth about the natural world. I admire your patience with these people, but they really have done nothing to earn it.
Zachriel: Do you realize that your goal posts have slipped?
gpuccio: No. Please, explain.
It was right there. Here it is again.
gpuccio: And yes, my main point in this discussion is exactly “that there is an insurmountable Gap for evolution in the origin of protein domains.”
We showed you an example of a protein domain that arose from a stochastic process. So, then the goal changes to this:
gpuccio: The functional domains which can really be useful in a biological context are extremely rare.
Now, it has to “really be useful” with whatever vague notion of really useful you decide to impose.
gpuccio: Trivial, useless functions of no real complexity.
You have a very odd idea of how proteins work. Binding is not just sticking to things, but an essential process, that out of all the thousands of molecules in a cell, the protein binds only to the target molecule, in this case, a nucleotide enzyme. In other words, it ‘recognizes’ the target in a vast sea of other molecules.
gpuccio: And binding is useful only when it can determine functional outcomes in the cell context.
Again, you have moved the goalposts. It’s a protein domain. Of course, it isn’t functional in a modern cell. It just shows that new domains can be formed through stochastic processes.
gpuccio: When the final Szostak protein was added to a real biological system, the only result was to create ATP deprivation.
The “only result” demonstrates that is it performing the specified function, and only the specified function, binding to ATP.
gpuccio: This is really beyond my understanding.
You made a claim, “that there is an insurmountable Gap for evolution in the origin of protein domains.” Demanding examples of “molecular macroevolution explained by the darwinian model” when already shown that stochastic processes can create protein domains is simply a diversion.
But, in any case, your query was answered. If we look at the most distant members of protein superfamilies, they are so divergent that without the intermediaries, we would certainly consider it macroevolution, yet we can reconstruct much of that history of divergence and show that it occurred more-or-less incrementally. And there are a variety of plausible hypotheses concerning the origin of protein domains, including primoridal evolution from early peptides and exonization.
Zachriel: As we explained the “darwinian model” depends on a consilience of evidence from multiple fields of study,
gpuccio: except, it seems, molecular biology…
Consilience means the unity of knowledge. It certainly does include molecular biology. Indeed, molecular biology includes some of the most important evidence for common descent. You even cited the Povolotskaya & Kondrashov paper that reached that conclusion based on a study of protein macroevolution. What consilience means is that you can’t just exclude evidence you don’t like, such as the evolution of macroscopic structures which leave fossils, or direct observation of evolutionary mechanisms.
gpuccio: As I have already said, that’s mainly neutral variation, which does not modify folding and function. It is evidence of CD, but not of macroevolution. No new function, and above all no new complex function, emerges.
There’s no significant functional variation within protein families and superfamilies? Not sure why you would think that.
Todd, Orengo & Thornton, Evolution of Function in Protein Superfamilies, from a Structural Perspective, Journal of Molecular Biology 2001.
Zachriel: In other words, they explain why homologies are difficult to discern, and their results are consistent with a world of more primitive polypeptides that precedes the LUCA.
gpuccio: No. Their model implies an emergence of protein domain as single ancestors, and then a diversification due to neutral mutations which is kept under the strict boundaries of their functional island by negative NS.
Hmm.
Povolotskaya & Kondrashov: Our analysis shows that it is conceivable that many more proteins were present in LUCA but have since diverged beyond our ability to detect their homology, and that given enough time some of the currently identifiable orthologues among the major kingdoms of life will diverge beyond recognition. Finally, our observation of receding protein sequences provides novel evidence of the common ancestry of life.
So their results are consistent with a complex protein world preceding the LUCA.
Zachriel: We have evidence of genetics, the distinction between genotype and phenotype.
gpuccio: That the only evidence form genetics is “the distinction between genotype and phenotype”. THat’s a strange concept.
Who said “only”? The distinction between genotype and phenotype is the basis of modern genetics.
gpuccio: And in what way would that disctinction be evidence for the darwinian model (classical, neo or neo-neo)?
You’re not making much sense. We are making a list of what we know, and what needs to be explained.
Zachriel: We have evidence of common descent.
gpuccio: Which in no way is evidence of the darwinian model (classical, etc).
You’re not making much sense. We are making a list of what we know, and what needs to be explained. Any theory has to be consistent with observation. You’re not too keen on this consilience of evidence thingy.
Zachriel: We have evidence of adaptation by natural selection.
gpuccio: We have certainly some limited evidence of adaptation. We have some evidence of negative NS, and very limited evidence of positive NS in rare cases of simple mutations and extreme selective pressure (what we usually call “microevolution”).
We have significant evidence of positive natural selection. Calling it microevolution is only handwaving, because (first-order) evolutionary theory predicts that changes will be incremental. That’s rather the whole point. But we also have excellent evidence for historical transitions, such as the evolution of cetaceans, and the evolution of the mammalian middle ear. We also can show that fecundity, variation and competition for limited resources leads to changes in population, i.e. evolution by natural selection.
gpuccio: That is your sincere opinion, and I respect it. But I disagree.
Your own cited authorities disagree with you.
gpuccio: you apparently underestimate the words “on average” in Dembski and Marks’ statement. They are not saying that occasionally a search algorithm cannot work better than random search. But that is a random occurrence.
We don’t misunderstand that point. But in an evolutionary context, we are not discussing a random search or a random landscape. Again, it’s like saying a river can’t find its way to the sea any better than a random search. Terrestrial landscapes have contours, and most rivers do reach the sea.
gpuccio: I don’t believe that such an information is in any way specific to solve the complex search problems at molecular level.
We know that simple evolutionary algorithms can generate complex solutions to complex problems. And we can observe this occur in nature. What you call microevolution will result in complex structures.
gpuccio: So, while the world has certainly a lot of “active information” in it, it is not specifically useful and pertinent to solve the problem of biological information.
Szostak & Keefe didn’t know what the sequence of the functional protein would be. They didn’t aim for a particular three-dimensional structure. They just tried a bunch of random sequences and found a few that bound to the target. This already shows that domains are not that rare in sequence space.
Then they replicated, mutated and selected the protein through several rounds. At the end, they still didn’t know what the sequence or structure were. But the proteins formed complex three-dimentional structures and were functionally specific.
marktfrank: The most surprising thing perhaps is how few comments address the main theme of the post – what happened to the intermediaries?
Extinction, many probably even before the LUCA.
It’s the same ID Gap Argument made before Leakey found his hominids, or Gingerich found his leggéd whales. IDers keep getting pushed back, to the Cambrian, to the LUCA, to the dawn of life.
The thread may be overlong, but the topic will return on any evolution/ID thread.
What’s nice here is that you have the luxury of knowing that you won’t be banned for posting unwelcome information. It means you will have time and opportunity to respond to misunderstandings, and to minor points.
Mark et al.:
I have really appreciated the discussion, but probably Mark is right that it has more or less reached its evolutionary end.
Just a few comments are due:
Toronto, I have been very surprised to read your last post. I will not comment on it.
Zachriel, thank you for all your comments, but I am afraid they are now too repetitive for my taste. As I hate repetition, I will not add anything to what I have already said about the Szostak paper. I believe I have been clear enough.
Thank you anyway for pointing to the paper about superfamilies. It is very interesting, and iMO it confirms all my points.
I quote again what I wrote:
“No. They keep the same basic function. They are probably the effect of neutral variation in the island of function, and possibly in some cases of tweaking if the protein for different molecular contexts via microevolutionary events (what we could call microevolutionary adaptation, which in principle could be achieved by darwinian mechanisms).
Even when different final functions are achieved in a superfamily through minor variations in the active site and in the affinity for ligands, as could be the case in the superfamily of nuclear receptors studied in a recent paper, and in the case of nylonase, he events are microevolutionary, and not complex. A case for design could be still made in these cases, but it would require an analysis of the complexity of the whole system, and not of the single molecule. I will abstain from that, for the moment.”
Thwe paper is exactly about different final functions achieved in superfamilies. But in many superfamilies the function remains the same, even in presence of great differences of primary sequence (see for instance myoglobin). That was exactly my point.
Petrushka: you have not commented on the rugged ladsacape.
Mark:
About the hypothesis testing. I know that Bayesians are a tough clan, but still most analysis is made by hypothesis testing against a null hypothesis, for instance in medicine.
It is true that rejecyion of the null hypothesis (usually that what we observe is the product of chance) does not in itself pèrove the alternative hypothesis. But nothing in science will ever prove something in that sense.
After rejecting the null hypothesis, any convincing explanatory hypothesis can compete with the orthers. That’s what I have repeated many times here. it’s all a question of competing explanatory models.
But the models must be explanatory. If a model does not explain anything, it cannot compete.
Well maybe the discussion is not quite over.
About the hypothesis testing. I know that Bayesians are a tough clan, but still most analysis is made by hypothesis testing against a null hypothesis, for instance in medicine.
1) Just because it is common practice does not make it good practice.
2) This testing is typically Neyman-Pearson – there is a hypothesis which we have some prior reason to believe may be true and a null hypothesis and we compare the results of the two. To work the other way round – look for extremities in the data and then dream up hypotheses that explain it, is a well-known statistical faux-pas even for conventional hypothesis testing.
After rejecting the null hypothesis, any convincing explanatory hypothesis can compete with the orthers. That’s what I have repeated many times here. it’s all a question of competing explanatory models.
But the models must be explanatory. If a model does not explain anything, it cannot compete.
No! No! You can always dream up a hypothesis which will explain the facts perfectly – a designer with the appropriate powers and motivation is one such hypothesis. Its explanatory power is enormous. It can explain anything. That is why there must be an independent reason for proposing it.
gpuccio: I will not add anything to what I have already said about the Szostak paper. I believe I have been clear enough.
None of your comments change the fact that a functional domain was found in random sequences. This shows that such domains are not that rare in sequence space. It also shows, incidentally, that the domains in extant biology probably represent only a small fraction of possible domains.
gpuccio: But in many superfamilies the function remains the same, even in presence of great differences of primary sequence (see for instance myoglobin). That was exactly my point.
Your point was that they always keep the same function, therefore are of negligible consequence. But that’s false. New functions can evolve, as well as structure.
gpuccio,
Then please prove me wrong.
Stand up for Dembski and say the Bible should not interfere with the scientific beliefs held by any of us, including Dembski.
If he believes that science supports a local flood he should be allowed to hold that opinion.
Do you agree?
Zachriel:
“Your point was that they always keep the same function,”
How can you say such blatantly false things about what I have said?
Again my quote:
Even when different final functions are achieved in a superfamily through minor variations in the active site and in the affinity for ligands, as could be the case in the superfamily of nuclear receptors studied in a recent paper, and in the case of nylonase, the events are microevolutionary, and not complex.
Toronto:
“Do you agree?”
Frankly, I can’t even understand with what I should agree. Maybe your emotions have compromised your clarity.
“Stand up for Dembski and say the Bible should not interfere with the scientific beliefs held by any of us, including Dembski.”
Why should I stand up for Dembski, and for what?
Why should I say that “Bible should not interfere with the scientific beliefs held by any of us”? If any of us decides that the bible should interfere with his own scientific beliefs, he is free to decide that. I believe in free personal choices. I am not a dictator.
I am well sure that the Bible does not interfere with my own scientific beliefs, and that’s enough for me.
I am well sure that your atheistic faith does interfer with your scientific beliefs, bery much and very badly, but again that’s fine for me.
“If he believes that science supports a local flood he should be allowed to hold that opinion.”
Obviously. What would you like to do? Imprison him? Kill him?
Zachriel:
“None of your comments change the fact that a functional domain was found in random sequences. This shows that such domains are not that rare in sequence space. It also shows, incidentally, that the domains in extant biology probably represent only a small fraction of possible domains.”
As I have said, I con’t love repetitions. It seems you do.
Stubbornly repeating the same statement will not make it true. Repeating the same statement without adding any new aspect is simply silly, or is simple brute force.
I believe we have expressed our ideas about this point very clearly. It’s all in the above posts. I am satisfied with that. And you?
gpuccio,
Here we finally see the clear problem between our sides, and it shows in almost all the exchanges on this post.
Toronto:
gpuccio:
I tell you that I support Dembski in my statement, you agree with me, and then finish by implying that I don’t.
So the question to you is, and I hope you’ll think about this even if you don’t answer, “Why can’t you engage us on the opinions we actually hold?
gpuccio: How can you say such blatantly false things about what I have said?
We just read what you write.
Zachriel: Don’t the wide variations within protein families count as “macroscopic evolution”?
gpuccio: No. They keep the same basci function.
…
Zachriel: Finally, we point to the wide variations in structure and function within protein families as examples of “molecular macroevolution.”
gpuccio: Wrong. As I have already said, that’s mainly neutral variation, which does not modify folding and function. It is evidence of CD, but not of macroevolution. No new function, and above all no new complex function, emerges.
We know that proteins organize into large phylogenetic groupings, families, which you agree imply common descent. Yet, there is wide diverence, not just in structure, but function among these phylogenetic groupings. Without the knowledge of intermediates, someone might very well not see how such divergence was possible, an insurmountable gap.
gpuccio: Stubbornly repeating the same statement will not make it true.
That’s correct.
gpuccio: Repeating the same statement without adding any new aspect is simply silly, or is simple brute force.
We provided additional cites, and also explained why binding is not a trivial function. Binding is how cellular mechanisms ‘recognize’ a single molecule out of a sea of complex interactions. Again, all you did then was repeat that it’s trivial.
–
After all, humans are ‘just’ elaborated Deuterostomes. A tube with appendages to stuff food into one end. Microevolution.
Toronto:
“I tell you that I support Dembski in my statement, you agree with me, and then finish by implying that I don’t.”
Frankly, I had not understood the sense and meaning of that post (I still don’t), as should be clear from my initial statement:
“Frankly, I can’t even understand with what I should agree. Maybe your emotions have compromised your clarity.”
I tried anyway to say what I thought about your phrases. Maybe I should just have waited for you to explain yourself better.
I apologize for not being smart enough.
Please don’t take my reply as any type of insult as I meant it when I said I’m glad you stay around to talk to us.
What I meant to point out is that Dembski is being forced to abandon his scientific beliefs. It wasn’t a choice that he made, rather it was his employer that made it for him.
The ID movement claims that evidence should be followed where it leads and here we have one of your most visible proponents being told to abandon that trail.
That is the fear I have for future generations of students. They may be told to abandon evidence for scripure. This is something your side says is not being sought and yet we’ve just seen it happen to Dembski.
What I would like to see is clarity on this issue.
If ID is introduced to schools, will students be treated like Dembski?
As a proponent of ID, I’m asking you, what is your position on this?
Evidence or scripture?
Zachriel:
This is becoming boring.
Anyway:
First, you are really unfair here. Your comment:
“Your point was that they always keep the same function”
(please, note the “always”)
is exactly in answer to my post where I wrote:
““No. They keep the same basic function. They are probably the effect of neutral variation in the island of function, and possibly in some cases of tweaking if the protein for different molecular contexts via microevolutionary events (what we could call microevolutionary adaptation, which in principle could be achieved by darwinian mechanisms).
Even when different final functions are achieved in a superfamily through minor variations in the active site and in the affinity for ligands, as could be the case in the superfamily of nuclear receptors studied in a recent paper, and in the case of nylonase, the events are microevolutionary, and not complex. A case for design could be still made in these cases, but it would require an analysis of the complexity of the whole system, and not of the single molecule. I will abstain from that, for the moment.”
The paper is exactly about different final functions achieved in superfamilies. But in many superfamilies the function remains the same, even in presence of great differences of primary sequence (see for instance myoglobin). That was exactly my point.”
Please note the:
“Even when different final functions are achieved in a superfamily”
and the
“But in many superfamilies the function remains the same”
So, apparently you did not read what I wrote.
To clarify better:
1) Superfamilies correspond to atructural calssification, based essentially on the basic folding.
2) A basic folding can experess only one function, or more than one. When it expresses more than one, that is usually achieved through variation at the level of the active site, which is something very different from the basic folding, and much less complex in terms of aminoacids involved.
3) All my discussions here have been based on the emergence of superfamilies, that is of baic foldings. That does not mean that variation at the level of the active site cannot be discussed, or that it must necessarily be explained in darwinian terms. But I stated very clearly, always in my multi-quoted passage, that such variation, being usually of moderate complexity, “in principle could be achieved by darwinian mechanisms” and that “the events are microevolutionary, and not complex.”
IOWs, basic protein superfamilies and structure almost always exhibit dFSCI, and that’s why I take them for my argumentation. Variations at the level of the active site, IMO, are not usually complex enough to exhibit dFSCI. Therefore, consistently with my procedure, I have not affirmed a design origin for them. I have anyway specified that “A case for design could be still made in these cases, but it would require an analysis of the complexity of the whole system, and not of the single molecule.”. But I have not made it.
4) Binding is of very different types and levels. There are trivial bindings and non trivial bindings. The role of the mutation selection cycles in the Szostac paper was explicitly to improve the “quality” of the binding. From the paper:
“Single representatives of each of these protein families (round 8) were chosen for further study. Only 5±15% of the mRNA-displayed protein prepared from each of these clones binds to immobilized ATP and then elutes with free ATP under selection conditions, consistent with the 6.2% binding and elution with ATP for the library as a whole. One possible explanation for this low level of ATP-binding is conformational heterogeneity, possibly reflecting inefficient folding of these primordial protein sequences.
In an effort to increase the proportion of these proteins that fold into an ATP-binding conformation, we mutagenized the library and carried out further rounds of in vitro selection and amplification.
Three consecutive rounds with mutagenic PCR amplification were
performed with an average mutagenic rate of 3.7% per amino acid
for each round. After six subsequent rounds of in vitro selection and amplification without mutagenesis, the proportion of the library of mRNA-displayed proteins that bound and eluted with ATP rose to 34% (round 18) (Fig. 2). At this point the library was entirely composed (56/56 clones sequenced) of the descendents of one of the four originally selected protein families (family B).”
So, why did they need to “improve” the binding, and why didn’t they study the characteristics of the original molecule, which would have been the only thing to do according to the objectives of the paper? You have never answered these questions. there is not convincing answer, becasue the only answer, obviously, is that the original binding was not very convincing. To loosely bind to immobilized ATP, just enough to be separated from the general population, was probably not a sufficient achievement. And I certainly agree on that.
Anyway, the fact remains that neither the initial molecule nor the final molecule could ever be selected in a biological environment, They are simply useless there. So, I really don’t see how they could represent steps of any “incremental” pathway. They can’t. They mean nothing, except for ideological propaganda. And even for that, they are not so good.
Toronto:
I have been clear on that for years, at UD:
Scriptures have no importance for ID. The attitude of starting from scriptures to make science is called “creation science”. I respect those who do that, some of them are also fine and intelligent people, but I wholly disagree: it’s simply wrong.
I don’t know the personal problems of Dembski, and I must confess that I am not specially interested in them. Dembski is both an ID proponent and a theologian, and I believe that he has always tried sincerely to keep the two things separate.
I am interested only in his work in ID, and even for that I don’t always agree with him. I respect and admire him very much, but for me he is no authority. Nobody is an authority in ID, as nobody should be an authority in science.
I have never thought that ID should be taught in schools. I believe that the mainly accepted scientific theory should be taught in schools, and at present it is neo darwinism.
But I do believe that ID should ne considered a vaible minority scientific position, and that notion can certainly be part of what is taught at school.
ID should certainly be taught and discussed in the universities, in the measure that will be appropriate as it is more widely accepted in the scientific community (which will happen).
ID is and must remain completely different and independent from any religious position. Tha’s why I would love a more active participation of lay people in the ID discussion.
It is certainly difficult to insist on that point, when most enemies of ID try daily to confound it with religion and creation science.
But I have never had any doubt: I have always embraced ID for its scientific merits: it’s a wonderful and deeply satisfyng model of reality. Religion has nothing to do with that.
Mark:
“No! No! You can always dream up a hypothesis which will explain the facts perfectly – a designer with the appropriate powers and motivation is one such hypothesis. Its explanatory power is enormous. It can explain anything. That is why there must be an independent reason for proposing it.”
But, as you should know well, there is a very strong independent reason for proposing design as an explication for biological information: it is the constant enmpirical association of design and dFSCI in all other cases. What else do you want?
I would never invoke “a designer with the appropriate powers and motivation” to explain any arbitrary thing. But I do invoke a designer for dFSCI for the reason above stated.
Sometimes I have the impression that you would admit design only if you personally could see the designer in the act of designing a protein, and commenting to you about each step of what he does.
Or maybe not even in that case…
But, as you should know well, there is a very strong independent reason for proposing design as an explication for biological information: it is the constant enmpirical association of design and dFSCI in all other cases. What else do you want?
Well I am glad you admit there has to be an independent reason. Unfortunately, as you know, I think the so called correlation of human design with dFSCI is just a sophisticated trick of definition and circular. But we have been round that many times.
Sometimes I have the impression that you would admit design only if you personally could see the designer in the act of designing a protein, and commenting to you about each step of what he does
No (although is pretty much what Behe demands of evolution). I only want to observe some part of what the designer does or some aspect of the designer or something that the designer does that is similar. Something that would allow me to assess the plausability of the hypothesis.
gpuccio,
I appreciate the clear and concise reply.
While this post seems to be winding down, I hope we can talk again in the near future.
Thanks again.
gpuccio: To clarify better:
That’s always welcome.
gpuccio: 1) Superfamilies correspond to atructural calssification, based essentially on the basic folding.
Superfamilies, like many phylogenetic categories, doesn’t have a single definition. However, there is such as thing as a fold superfamily combining aspects of sequence similarity with fold, which seems to eliminate some of the problems with phylogeny.
gpuccio: Binding is of very different types and levels. There are trivial bindings and non trivial bindings. The role of the mutation selection cycles in the Szostac paper was explicitly to improve the “quality” of the binding.
Random sequences can form functional domains. Calling it “trivial” is just handwaving. Binding is an essential protein function. Ligases have also been isolated from random libraries.
gpuccio: So, why did they need to “improve” the binding, and why didn’t they study the characteristics of the original molecule, which would have been the only thing to do according to the objectives of the paper?
Because that’s the hypothesis! That minimal function becomes optimized by evolution. Now you have it!
gpuccio: So, I really don’t see how they could represent steps of any “incremental” pathway. They can’t.
It answers the claim that “there is an insurmountable Gap for evolution in the origin of protein domains.“
Mark:
“But we have been round that many times.”
Yes, we have. And luckily, we both don’t love repetitions. That’s one of the things I appreciate in you.
“Something that would allow me to assess the plausability of the hypothesis.”
I am sure that the growing information gathered from genomes, proteomes, transcriptomes, and a deeper and more correct analysis of that information, will allow a better assessment of the plausibility of both the design hypothesis and the neo darwinian hypothesis (or of any other neo neo model which dares to compete).
That’s what science is all about.
Zachriel:
Just one comment that perphaps it is still worthwhile to make. You say:
“Because that’s the hypothesis! That minimal function becomes optimized by evolution. Now you have it! ”
No. That is not the hypothesis. The objective of the paper was to indagate the existence of function in random library, not its artificail optimizaion. So, what Szostak did was explicitly out of order for tha research model used in the paper. You seem to resuse to understand that.
And anyway, even more in general, the darwinian hypothesis is that minimal function must be selectable by NS (IOWs, must be able to confer a positive differential reproduction) to be optimized by evolution. The Szostac protein does not meet this requirement, by far it does not meet it, not even in its “refined” form, least of all in its original form.
And the refinement itself was accomplished through artificial intelligent selection in the lab. It could never have happened in a natural biological system. The ability to bind loosely ATP is barely enough to select the molecules in a very sensitive experimental device. That has nothing to do with conferring positive differential reproduction to a living being.
But I am afraid you will again deny even these elementary things.
Toronto:
I appreciate that you appreciate.
We’ll certainly meet again. Bye.
I am sure that the growing information gathered from genomes, proteomes, transcriptomes, and a deeper and more correct analysis of that information, will allow a better assessment of the plausibility of both the design hypothesis and the neo darwinian hypothesis (or of any other neo neo model which dares to compete).
I can see how this information will allow a more correct assessment of the neo darwinian hypothesis – indeed it has led to many extensions and corrections already – but I cannot see how it can throw light on the plausability of the design hypothesis unless the hypothesis specifies the designer(s) or the mechanism they use/used (by which I mean the interface between the designer and life). Given this information we can check on the plausability of there being such a designer and we can check on the plausability of that mechanism – maybe by trying using it ourselves as we have used artificial selection to assess natural selection. Without it – the most we can observe is inexplicable changes in life – which are just gaps waiting to be filled.
gpuccio: That has nothing to do with conferring positive differential reproduction to a living being. <
It answers the claim that “there is an insurmountable Gap for evolution in the origin of protein domains.“
gpuccio: And anyway, even more in general, the darwinian hypothesis is that minimal function must be selectable by NS (IOWs, must be able to confer a positive differential reproduction) to be optimized by evolution. The Szostac protein does not meet this requirement, by far it does not meet it, not even in its “refined” form, least of all in its original form.
These are standard biological functions and could certainly provide an advantage. As a simple example, binding could inhibit a toxin.
Mark:
” I cannot see how it can throw light on the plausability of the design hypothesis”
I believe it certainly can and will. No amount of evidence, however, will ever be able to change your position, given your statement that you need that: “the hypothesis specifies the designer(s) or the mechanism they use/used (by which I mean the interface between the designer and life)”. That’s no big problem. You and a few others will keep your position, and become a though minority (I envy you!).
For me, defining the designer as “a conscious intelligent agent who can interact with matter” is more than enough to build a valid hypothesis. The “interface between the designer and life” is not known even for humans (we have no idea of how our consciousness interacts with our brain).
The modalities of implementation, instead, can certainly be clarified by the coming evidence.
The “interface between the designer and life” is not known even for humans (we have no idea of how our consciousness interacts with our brain).
Quite a lot is known. We know a lot of the physical things that are necessary, if not suffient, for it to work. And we can see it in action in the sense of observing the designer manipulating things and this having a physical manifestation both in the external world and more recently in the brain. What are you ever going to observe about the design hypothesis except unexplained jumps in function in DNA?
gpuccio:
Our consciousness is a “result” of the working of our brain.
For instance, the music you hear does not “interact” with your stereo system in your living room, it is a “result” of it.
Consciousness does not come from outside of us, it comes from within us.
Let’s try to plug in a few numbers:
* s is the selection coefficient
* chance of fixation = ~2s (with large populations and low s)
* mutation rate = 10^-6 for each randomization
* chance of specified function = 10^-11
* s = 10^-6
* chance of fixation = 10^-23
On the other side of the equation:
* 1 bacteria per ml in drinking water
* 10^18 bacteria in Lake Erie filled with clean drinking water
* 10^5 generations
* generation time = few hours
* new function can be expected to fix in a few years.
These are not meant to be actual figures, of course. Mutation rates were probably very high, horizontal mechanisms were prevalent both within and between organisms, even the notion of a discrete organism may not really make sense in the primordial world. Concentrations of organisms more like 10^9 or more per ml in biotic rich areas might be more reasonable. We’re testing for a specific function, but the chance of some function arising is greater than 10^-11.
Furthermore, randomization of the entire sequence is not how new domains are posited to originate. Rather, there is a world of peptides and ribozymes. Amino acids are limited in number and already sorted by charge and other important characteristics. In addition, once the minimally functioning protein is fixed (or rather during fixation), it will be quickly optimized, so that if we were to take periodic snapshots, we would see the abrupt appearance of the optimized protein.
For me, defining the designer as “a conscious intelligent agent who can interact with matter”
__________________________________________
How is that different from defining the designer as having whatever attributes are necessary to do whatever needs to be done to explain whatever happened?
How is it different from saying that an undefined entity did undefined things at underspecified times and places?
It seems to me that we are back at the point where I got banned from UD for saying that ID proponents are abysmally ignorant of the history of science.
I say this because — just as an example — 200 years passed between the hypothesis that the earth orbited the sun, and a coherent mathematical description that relied only on observable forces. Basing the invocation of fairies to explain a phenomenon, demonstrates an ignorance of history. It’s a throwback to animism.
And even Newton relied on mysterious entities acting in unspecified was to keep the planetary orbits stable. This kind of reasoning seems to spring from some need to avoid accepting the simple fact that we don’t know everything.
Petrushka:
“How is that different from defining the designer as having whatever attributes are necessary to do whatever needs to be done to explain whatever happened?
How is it different from saying that an undefined entity did undefined things at underspecified times and places?”
Very, very different. Conscious intelligent representations are constantly observed in the process of human design, and are the origin of that process. These are facts.
“I say this because — just as an example — 200 years passed between the hypothesis that the earth orbited the sun, and a coherent mathematical description that relied only on observable forces. Basing the invocation of fairies to explain a phenomenon, demonstrates an ignorance of history. It’s a throwback to animism.”
And your words are throwback to superficial, imprecise and arrogant thinking. There is not even the start of a reasonable argument in them, only incorrect analogies.
“This kind of reasoning seems to spring from some need to avoid accepting the simple fact that we don’t know everything.”
We certainly don’t know anything, and never will. But understanding more about the causes of what we observe is the true spring of science. You seem to denigrate that motivation only when the conclusions are not what you expect and like.
And, by the way, you have not yet commented on the rugged landscape.
Others have pointed this out, but since it’s part of my argument, I’ll repeat it.
There’s no example in history of any regular phenomena being explained by reference to non-physical, capricious entities.
God of the gaps has failed miserably as a critique of the fossil record. What reason do we have to suppose it will be useful in explaining genetic fossils?
Zachriel:
You have made it! A whole post of just so stories.
Now you are a true darwinist.
gpuccio: You have made it! A whole post of just so stories.
Not at all. We’re testing your suggestion that because of the function’s weak activity, it wouldn’t be subject to selection.
Population genetics has a mathematical basis. We know that domains can form in random sequences. Even weakly selective advantages will fix in very large populations, when |s| > 1/4Ne, where Ne is the effective population size. In addition, even if a particular functional sequence is lost, we can expect that over time, it will continue to arise in the population.
Very, very different. Conscious intelligent representations are constantly observed in the process of human design, and are the origin of that process. These are facts.
______________________________
It is also a fact that no entity fitting your criteria for the designer has ever been observed, nor has any regular phenomenon ever been explained by reference to such an entity.
It is also a fact that the structure and behavior of living things is entire unlike anything engineered by humans. Except for algorithms that have been inspired by evolution.
It is also a fact that human inventions evolve incrementally.
It is also a fact that human inventions incorporate much more lateral transfer than living things. This is particularly noticeable in the genetic engineering of plants and animals. In fact if humans disappeared, and aliens examined the remaining life on earth, they would be able to distinguish the signatures of genetic engineering.
Zachriel,
Come now, Zachriel, you surely don’t expect gpuccio to apply mathematical rigor to his arguments. He hasn’t ever done so in the past.
I love watching my predictions come true.
Zachriel: Even weakly selective advantages will fix in very large populations, when |s| > 1/4Ne, where Ne is the effective population size.
Rather, selection is more important than drift with sufficiently large populations, with probability of fixation being 2s.
Zachriel:
“Even weakly selective advantages will fix in very large populations”
Which “weakly selected advantage”? In your dreams?
Petrushka:
You still are not commenting on the rugged landscape.
Zachriel:
“Not at all. We’re testing your suggestion that because of the function’s weak activity, it wouldn’t be subject to selection.”
Indeed I said that the function has no selectable activity. How are you testing that? You are assuming that the weak activity is selectable, but it is not. It is useless, therefore non selectable.
Hayashi et al., Experimental Rugged Fitness Landscape in Protein Sequence Space, PLoS One 2006.
Petrushka: Unfortunately for ID the slope seems to be quite climbable, at least in the foothills.
gpuccio: 1) If you read the paper carefully, you will see that the “rugged landscape” in no way implies that the results obtained will lead to further optimization. IOWs, they are not steps to the main peak of functionality.
Your statement isn’t very clear. A landscape may have many peaks and valleys. There may not be a “main peak”. From random sequences, the experimental evolution optimizes to 40%-55% of the wild strain. That’s very significant.
gpuccio: 2) Please look at this part of the discussion: “The question remains regarding how large a population is required to reach the fitness of the wild-type phage. The relative fitness of the wild-type phage, or rather the native D2 domain, is almost equivalent to the global peak of the fitness landscape. By extrapolation, we estimated that adaptive walking requires a library size of 10^70 with 35 substitutions to reach comparable fitness.”
You left off the rest of it.
“Such a huge search is impractical and implies that evolution of the wildtype phage must have involved not only random substitutions but also other mechanisms, such as homologous recombination.”
Indeed, evolutionary algorithms show that recombination is essential to escape local fitness peaks. The researchers didn’t test recombination because they were interested in the fitness landscape with regards to mutation only. Then they say this:
“First, the smooth surface of the mountainous structure from the foot to at least a relative fitness of 0.4 means that it is possible for most random or primordial sequences to evolve with relative ease up to the middle region of the fitness landscape by adaptive walking with only single substitutions. In fact, in addition to infectivity, we have succeeded in evolving esterase activity from ten arbitrarily chosen initial random sequences. Thus, the primordial functional evolution of proteins may have proceeded from a population with only a small degree of sequence diversity.“
You still are not commenting on the rugged landscape.
__________________________________________
Sure I have. The actual landscape is unknown. Both ID and evolution speculate on the actual history of change. Evolution assumes regular processes and ID assumes the intervention of an unspecified and never observed entity.
Zach has cited the paper I was looking for.
Your case rests on the improbability of proteins originating. We have barely begun research on this and it is already obvious that origination is not a problem. Not is evolution to mid level functionality.
Zachriel:
“Your statement isn’t very clear. A landscape may have many peaks and valleys. There may not be a “main peak”. From random sequences, the experimental evolution optimizes to 40%-55% of the wild strain. That’s very significant.”
The landscape they have studied has a very definite optimization peak, the one we find in nature. The wildtype. That peak, by admission of the authors, cannot be found by RV and NS unless one starts with 10^70 random sequences. That is very significant. But as usual, you see only the significance you like.
Moreover, the partial peaks are not steps towards the optimized peak, but rather obstacles in the search. Indeed, because of the effect of negative NS, peaks in this case can be better understood as holes. Think of it as a small ball which should arrive to the main hole, but is caught by intermediate, distant, smaller holes from which it is very difficult to come out, to regain the flat landscape.
The reason for that is that the peaks are unrelated. In the words of the authors:
“More than one such mountain exists in the fitness landscape of
the function for the D2 domain in phage infectivity. The sequence
selected finally at the 20th generation has ~W=0.52 but showed no
homology to the wild-type D2 domain”
Can you understand that? It’s simple english: “no
homology”. So, where are all your just so stories of incremental evolution?
Zachriel:
“You left off the rest of it”
Sure. The rest of it is mere specualtion, and has nothing to do with the results of the study. That’s why I left it out. I am interested in facts, not in imagination.
“Indeed, evolutionary algorithms show that recombination is essential to escape local fitness peaks.”
More wishful thinking. Why, when faced with a problem, you always evade to other arguments, and especially to wrong models of intelligently designed systems which don’t model anything pertinent?
“The researchers didn’t test recombination because they were interested in the fitness landscape with regards to mutation only.”
That’s why they cannot say anything about recombination from their work.
Where is the paper that shows that a functional optimized molecule like that can be retrieve starting from random sequences, by RV, NS “and” recombination? In a lab system? Facts please, not just stories.
“Thus, the primordial functional evolution of proteins may have proceeded from a population with only a small degree of sequence diversity.”
May. Sure. What an useful little word. What a pity that their results show exactly the opposite.
Petrushka:
“Zach has cited the paper I was looking for.”
Indeed, I have cited it, a lot of time before, asking you to comment on what I had said, which you haven’t done.
Must I remind you again that it was you who challenged me with that paper? You are supposed to know it well, and to be able to comment on it, having been you yourself who have cited it the first time in favor of gradual evolution. One should never be impulsive in life.
Zachriel has only done what you have not done: he has seriously tried to answer my comments. He has tried his best. You have not even tried. Have the dignity to keep silent now.
Can we at least keep the “natural history” of this thread correct?
Exactly what is natural or dignified bout asserting the existing of invisible ad hoc entities?
Petrushka:
“Exactly what is natural or dignified bout asserting the existing of invisible ad hoc entities?”
I admit it: you deserve a world prize for evasion. You are a master artist of it.
Toronto:
“Our consciousness is a “result” of the working of our brain.”
So, you are a fan of strong AI, and not only of darwinism. My compliments.
“For instance, the music you hear does not “interact” with your stereo system in your living room, it is a “result” of it.”
One true statement, occasionally.
“Consciousness does not come from outside of us, it comes from within us.”
Who said that it comes from “outside of us”? It is us. We are consciousness.
gpuccio,
You did when you said this, “(we have no idea of how our consciousness interacts with our brain).”
Since what we perceive as consciousness, is the result of ongoing chemical and electrical processes of our brains, there is no “interaction”. It would be like saying how does the flame “interact” with the wick of a candle. There is no “interaction” since the candle and wick are the source of the flame.
I need clarity from you again as I am not sure what you mean by “we are consciousness”.
Are you suggesting a disembodied spirit inhabits our bodies and interacts with our brain?
gpuccio,
I don’t know what you mean by “darwinism”.
I believe in evolution while you believe in ID.
It would be wrong of me to label you a “Hovindist”.
gpuccio,
Here’s a better analogy that will let let me understand what you mean by brain/consciousness interface.
Think of a client/server relationship with computers.
Are you saying our “consciousness” is like the client and the brain is like the server?
Are they peers?
Toronto:
“Since what we perceive as consciousness, is the result of ongoing chemical and electrical processes of our brains, there is no “interaction””.
In these words is implied all the errors and the mistification of strong AI (my intellectual indignation is for the theory, not for you).
Language, as usual, says more truth than philosophy.
“What we perceive” is the representations of consciousness. They come in part (but only in part) from the “ongoing chemical and electrical processes of our brains”. For example, the simplest representation, a sensation, is certainly originated by some activity in the nervous system, in turn originated by some physical outer event.
But it’s “we” who perceive. We are the transcendental self representing all representations, ad unifying them in one consciousness.
Thus, the logical order is: “we” (being conscious beings) perceive “what we perceive” (the representations of consciousness, be them sensations, inner states, or the conscious process itself) “through consciousness” (not “as consciousness”: all the representations ore objects to consiousness itself, only the perceiving self is subject). Indeed, although we observe the perceiving self in its conscious process, we just know that it exists intuitively. The intuition that we exist as subjects is the fundamental basis of reality and of our map of it.
Therefore, there is a continuous interaction in both senses: from the outer world to the perceiving self (through the nervous system, including the brain), that is sensations; and from the perceiving self to the outer world (always through the nervous system, including the brain), that is actions. That has been known for millennia.
“Are you suggesting a disembodied spirit inhabits our bodies and interacts with our brain?”
Obviously. Or at least a non physical perceiving self. That’s what has always been called “soul”. Are you new to the concept?
Anyway, this is more a philosophical concept. In a more empirical sense, I am saying that a perceiving self certainly exists (we know that), and that it obviously interacts with body and brain in both directions. The notion that it is a “product” of body and brain is only a bizarre idea of strong AI theory, unsupported by anything. The perceiving self is not explained by anything.
“I don’t know what you mean by “darwinism”.”
We have clarified that many times, even recently at UD. It’s just a quick way of reassuming more complex definitions.
In a strict sense, it means “neo-darwinism”, that is the classical modern syntesis “à la Dawkins”.
In a more general sense, the concept can include all more recent variations, which, although sometimes pretending to have new contents, can easily be reduced again to RV + NS.
The only partial exception being neo Lamarkism, which is anyway an ill defined entity, at least for now.
It’s just that it is rather tiring to write all tfhose “neo” so many times.
“Evolution” is not a good general term, because it makes no mention of the explanatory theory one refers to. I believe in evolution too, but in intelligently guided evolution. “Unguided evolution theory” is a better term, and can be used and is often used.
Finally, you can call me as you like (Hovindist is not bad), provided it is clear that I believe that biological information is the product of conscious intelligent design.
“Are you saying our “consciousness” is like the client and the brain is like the server?”
I am rather saying that our brain is like the computer (whatever computer), and our consciousness is the user. I would never liken consciousness to a machine. It is not a machine, while the brain is.
“Are they peers?”
No. Consciousness is the boss.
If you are going to get into a debate on dualism that deserves a thread of its own. I have kicked it off with a comment about dualism and ordinary language.
gpuccio: The rest of it is mere specualtion, and has nothing to do with the results of the study.
We know how recombination works, and know its important in terms of moving between fitness peaks. You cited a study about point-mutation, then you drew a faulty generalization that didn’t include recombination.
Zachriel: Indeed, evolutionary algorithms show that recombination is essential to escape local fitness peaks.
gpuccio: More wishful thinking.
Absolutely not, and easily demonstrable.
gpuccio: Why, when faced with a problem, you always evade to other arguments, and especially to wrong models of intelligently designed systems which don’t model anything pertinent?
You mean we can’t study the effects of recombination and point-mutation in a simulation? Seriously? That’s like saying we can’t study planetary mechanics or the weather by using models and equations. A simulation is just an if-then abstaction.
gpuccio: That’s why they cannot say anything about recombination from their work.
They can say they didn’t include recombination. They might also mention well-established facts about recombination.
More importantly, YOU drew the faulty generalization without accounting for recombination. Ironically, you base the overthrow of the biological sciences on something we can see in a simple model.
gpuccio: Where is the paper that shows that a functional optimized molecule like that can be retrieve starting from random sequences, by RV, NS “and” recombination?
They showed functional molecules (40%-55% of the activity of the wild versions) evolved from random molecules. And they didn’t even use recombination. This alone is sufficient to show that random molecules can have functional characteristics. They even found lyases.
Thus, the primordial functional evolution of proteins may have proceeded from a population with only a small degree of sequence diversity.
gpuccio: What an useful little word. What a pity that their results show exactly the opposite.
They used the word “may” because they don’t show a historical link, but that random origin is a workable mechanism. The real story is undoubtedly more interesting.
gpuccio: The wildtype. That peak, by admission of the authors, cannot be found by RV and NS unless one starts with 10^70 random sequences. That is very significant. But as usual, you see only the significance you like.
Your use of RV is misleading. They did not show that RV and NS requires 10^70 random sequences. What they showed was that RM (meaning point-mutation) is insufficient.
The results demonstrate that random molecules can be functional, and that evolution can dramatically increase the functional capability of weakly functioning molecules. It fills your claimed Gap concerning 0% function. Now we have a new Gap (of course!) between 40% and 100%.
gpuccio: Moreover, the partial peaks are not steps towards the optimized peak, but rather obstacles in the search.
You really don’t understand the importance of recombination, do you?
Zachriel:
“You cited a study about point-mutation, then you drew a faulty generalization that didn’t include recombination.”
I cited a study which demonstrates that random libraries and mutations and NS cannot do what you say they do. Where did your enthusiasm for ramdom libraries go?
Is now recombination the new magician? Recombination ia a random event, exactly like point mutations. And you cannot recombine what still does not exist. What are you preaching? Recombination ex nihilo?
“You mean we can’t study the effects of recombination and point-mutation in a simulation?”
No, I mean that evolutionary algorithms are wrong simulations.
“That’s like saying we can’t study planetary mechanics or the weather by using models and equations.”
I believe that here you are really gross in your epistemological metaphor.
“They can say they didn’t include recombination. They might also mention well-established facts about recombination.”
They can say what they like. But still that has nothing to do with their experimental work, which is the only reason why I read their paper.
“More importantly, YOU drew the faulty generalization without accounting for recombination. Ironically, you base the overthrow of the biological sciences on something we can see in a simple model.”
This is so silly that I will not comment on it. Zachriel, your style is deteriorating. Take care.
“They showed functional molecules (40%-55% of the activity of the wild versions) evolved from random molecules. And they didn’t even use recombination. This alone is sufficient to show that random molecules can have functional characteristics. They even found lyases.”
IOWs, you have no paper showing, as I had asked, “that a functional optimized molecule like that can be retrieve starting from random sequences, by RV, NS “and” recombination”.
Your evasions are more telling than your arguments.
I would have much more to say anout why their paper does not even show that “functional molecules (40%-55% of the activity of the wild versions) evolved from random molecules”. But now I have not the time.
“They even found lyases.”
That is another paper. I have not commented on it. Can’t you really stay on the subject when you have no more arguments?
“Your use of RV is misleading. They did not show that RV and NS requires 10^70 random sequences. What they showed was that RM (meaning point-mutation) is insufficient.”
OK, I can recognize when an important point is made. RM. Ah, yes: “meaning point-mutation”.
“It fills your claimed Gap concerning 0% function.”
OK, maybe I have to comment after all.
The reason why the paper does not demonstrate any of that is that the methodological context is extremely biased in favor of the result.
Indeed, they have chosen a complex molecule, and a function (infectivity) which refers to the whole complex. Then they change one chain of the molecule with a random sequence.
Surpisingly, that did not “zero” the function, but only reduced it very much.
That is very important, because that’s the only reason which allows NS to come in.
The random sequence alone, obviously, would never have been selected, and would never have “evolved”.
What happens here is that the whole functional molecule, which retains its basic function for infectivity even in the worst scenario (with a random sequence inserted) succeeds in partially retrieving the original level of function through some adjustement of the “disturbing” sequence. But it completely falls short of retrieving the true functional sequqence of the missing chain, the only one which would allow the complex to get back to its full, natural functionality.
Given this experimental context, your conclusions are a gratuitous hyperbole. The conclusions about the ineffectiveness of RM (“meaning point-mutation”) and NS are instead absolutely pertinent.
“You really don’t understand the importance of recombination, do you?”
Teach me. I am listening. Always ready to learn.
gpuccio: I cited a study which demonstrates that random libraries and mutations and NS cannot do what you say they do. Where did your enthusiasm for ramdom libraries go?
In the phage study, a domain is replaced with a random sequence. Some random sequences conferred more of an advantage over others.
gpuccio: Recombination ia a random event, exactly like point mutations. And you cannot recombine what still does not exist.
Evolution that includes recombination can explore regions of the fitness landscape that mutation alone cannot. And yes, recombination creates novel combinations.
gpuccio: OK, I can recognize when an important point is made. RM. Ah, yes: “meaning point-mutation”.
Thank you. Even if you don’t recognize the importance of recombination to moving off fitness peaks, it is an observed mechanism of variation, so any broad claim based on this experiment concerning random variation being limited is an overgeneralization.
gpuccio: No, I mean that evolutionary algorithms are wrong simulations.
Tell us why, and tell us how to make a suitable simulation.
Zachriel:
“Tell us why, and tell us how to make a suitable simulation.”
Why? Because they are engineered, introducing hidden information in many different ways. Again you can refere to the Dembski and Marks papers on this important point, but I will try to sum up some concepts here, very simply.
Hidden information is the difference between a truly unguided system and one where, in one form or another, design has an important role.
First of all, it must be clear that any simulation of a darwinian mechanism must not include any artificial intelligent selecttion.
This is the main point which is wrong in many so called “evolutionary algorithms”.
Usually, that comes in the form of some intelligently chosen “fitness function”.
I have discussed many times at UD the difference between artificial intelligent selection and natural selection.
In intelligent selection, something is recognized by the system, and the system actively promotes the replicators according to that recognition.
Nothing like that can model NS.
In NS, the environment (or the system which simulates it) must be completely “unaware” of the replicators, and must not recognize any particular thing. It’s the replicators themselves that must reach “selection” through some positive differential replication in them which takes advantage of the existing environment.
Moreover, the mechanisms of mutation or variation can be variously implemented, but it must again be totally unrelated to any specific anticipated effect.
So, “how to make a suitable simulation”? That too I have discussed in some detail at UD. Petrushka should remember.
Replicators can be programmed for some independent system (I have suggested Windows, but any environment, more or less complex, is fine, provided that one important condition is satisfied: the environment must be pre-existent to the experiment, or at least it must not have been in any way designed for the experiment. Only that condition can guarantee that no hidden information is introduced, vluntarily or not, into the system. That is a perfect simulation, because after all the natural environment is not supposed to have been designed for life in the model (unless it is a TE model).
The replicators, instead, are designed to use the system to replicate, and they can in some degree take advantage of the system resources. That’s fine, because already existing functional replicators are part of the darwinian model. That’s also something like a “blind” experiment, preventing many forms of bias.
The replicators must be programmed so that they incorporate a variation system. The variation system can be variously regulated or implemented, the only important requirement is that it must be truly random and no hidden information must be inputted into it.
And then? Then we wait, and check periodically the results. Aim of the study is to demonstrate that some new complex function can evolve in the code of the replicators in that system, and be selected, either incrementally or in other ways.
The environment needs not be static. Random changes could be intoducted in it, but always “in blind” (for instance, by programmers who are not aware of the purpose of the research, and who know nothing of the replicators).
That’s my idea of a correct simulation.
gpuccio:
Correct me if I am wrong, but I sense the same error here that UD contributors such as GilDodgen have made.
You seem to imply, that the artificial landscape where this simulation will take place is Windows itself and not a contained application simply hosted by Windows.
Here is why this is wrong.
Imagine a testing lab that is trying to determine the flashpoint of material used in pillow cases.
The technician puts a thermometer on his workbench, pours gasoline in the sink, drapes the material under test over the faucet, ignites the gasoline and then runs out of the building as fast as he can before he dies of smoke inhalation.
What do you think the chances are of the firemen finding the thermometer intact so he can record the results of his test?
This is what Gil and others on UD have proposed for a “Darwinian” simulation on a computer, an unresricted change in bits a running program would randomly make system-wide.
That is the same as burning down your lab as part of your experiment.
It makes absolutely no difference whether a computer simulation is done on Windows, a home-made operating system or distributed among multiple servers miles apart.
It is the algorithm you are supposed to be testing, not your lab.
gpuccio: In NS, the environment (or the system which simulates it) must be completely “unaware” of the replicators, and must not recognize any particular thing. It’s the replicators themselves that must reach “selection” through some positive differential replication in them which takes advantage of the existing environment.
Yes, the simulated environment or fitness landscape is separate from the replicators. They interact only with regards to competition for limited resources and fit to that environment. How did you think evolutionary algorthms worked?
gpuccio: (I have suggested Windows, …
That doesn’t make any sense. Windows is not a fitness landscape, or can we imagine how you think to represent it as an environment for replicators.
A typical evolutionary algorithm consists of a fitness landscape and replicators. The landscape represents the limited resources required for replication. Those higher on the landscape have more resources and leave more offspring. It’s an abstraction, but it is sufficient to show how evolution can hill-climb, and how recombination can allow replicators to explore between peaks. Indeed, the Hayashi paper is based on the principle of a fitness landscape.
gpuccio: any environment, more or less complex, is fine, provided that one important condition is satisfied: the environment must be pre-existent to the experiment, or at least it must not have been in any way designed for the experiment.)
There’s all sorts of preexisting landscapes that can be explored by evolutionary algorithms. It could be a complex mathematical problem, a signal reception problem. It could even be the dictionary. If you want, you could represent the fitness landscape so it more closely resembles the natural world, with food resources and replicators with a mechanism of movement that evolves in terms of coordination. But the essential results are the same.
That may be a bit off-topic, though. On the original point, the origin of protein domains has a simple, natural mechanism.
Toronto: It is the algorithm you are supposed to be testing, not your lab.
That’s an excellent point. A simulation is an implementation of an algorithm, an elaboration of an if-then scenario. The algorithm must be independent of any programming language, operating system, or computer.
If bodies are attracted proportional to the product of their masses and inversely proportional to the square of the distance between the bodies. If they have given initial velocities and distance. Where will they be tomorrow?
If we have replicators. If we have a given fitness landscape. If the replicators randomly mutate. If replicators are more successful the higher up the fitness landscape they are found. Then what?
Zachriel and Toronto:
I find your remarks very off the point.
I never said it must be Windows. It can be any software environment, but it must exist of itself, with its rules, and not have been programmed for the experiment.
Why do you insist so much that the environment must be simulated as a fitness landscape? Any software environment or operationg system has limited resources. Why shouldn’t replicators evolve to make a better use of them?
You are only avoiding the problem. You must simulate the replicators, and put them in an environment where they can replicate, better or worse. The environment must not “measure” anything in the replicators. It’s the replicators which must be able, through random variation, to use better the environment.
Why should that be possible in nature but not in a software environment?
gpuccio: Any software environment or operationg system has limited resources. Why shouldn’t replicators evolve to make a better use of them?
Try to reread Toronto’s comments. The algorithm is independent of the implementation. You should be able to describe the algorithm without reference to a computer in terms of a number of rules.
Population
Genome/Phenome
Mutation
Mating
Fitness
Instead of a fitness landscape, you could have a rule concerning the acquisition of resources, but you have to be explicit if you want to actually implement the algorithm.
gpuccio: You must simulate the replicators, and put them in an environment where they can replicate, better or worse. The environment must not “measure” anything in the replicators. It’s the replicators which must be able, through random variation, to use better the environment. Why should that be possible in nature but not in a software environment?
The fitness landscape abstractly represents the relationship between the environment and fitness. Consider how it is used in
Consider how it is used in the Hayashi paper. The fitness landscape represents the sum-total relationship between the sequence and the reproductive fitness of the organism. We don’t have to account for how the molecule makes the organism more fit, or its shape or stereochemical properties. Just its relative fitness.
In any case, you could simulate a watery world with food resources, and organisms with primitive swimming motility and coordination that evolves. When they find food, they replicate. If, after a time, they don’t find food, they die. There is still a fitness landscape (or rather, we could create one from the data à la Hayashi et al.).
There are countless such experiments that can and have been run. There is no controversy about this anywhere outside the closed ID community.
gpuccio:
I shall try to be clearer why this can’t be done.
In an operating system like Windows or Linux, paging or another form of memory protection/partitioning scheme is used.
All these schemes take a page or block of “logical” memory and assign attributes to it. These attributes are typically things like, “Read Only”, Readable-Writable”, or “Executable”, etc.
An application almost never in any of these systems is allowed to set the attributes of memory pages assigned to it by the OS when the app is launched.
That means that at most, we can have “Readable-Writable” pages but never can we change one into “Executable”. We could also never change an OS’es page to anything other than what the OS set it to.
You would never be able to have any code randomly develop that could in any way change even a single bit in any code or data page/segment in the OS.
The point is that if we agreed with you and tried to do it, the OS itself would stop us.
It won’t let us accidently burn down the lab, so to speak.
You must set up a controlled test/simulation however you like with the restriction that the OS proper, by design, will not allow you to modify it’s code in any way.
That is why Gil is so wrong in what he suggested. OS’es are the lab for the software that we run as applications.
You can develop your own interpreter and randomly change it to your heart’s desire, but even this interpreter is an app as far as the OS is concerned and is restricted in the same manner as any other app.
So, come up with an algorithm that you can monitor for our simulation, since that is what we are testing, the algorithm.
Toronto:
I don’t understand your point, or you don’t understand mine.
“You would never be able to have any code randomly develop that could in any way change even a single bit in any code or data page/segment in the OS.”
Why should that happen?
All the replicators should do is improve the way they exploit the system’s resource. They can copy themselves to the hard disk, replicate, use the RAM, compete one with the other. Think of them as simple viruses which should evolve to more elaborate spyware, or just some well structured program. What is your problem?
Zachriel:
Whatever you say, an implemented fitness landscape is nothing sinilar to an unaware environment which just “hosts” replicators. Your insisting that the environment must be implemented is motivated only by one thing: you well know that, unless the whole system is programmed to allow some specific result, the result will not come, or will be trivial.
Again, my point is simple, and you continue to elude it. If it is true that random variation and NS can evolve complex functions in replicators in an environment, than my test must work.
We have an environment, with necessity rulesm resources and complex landscapes (the computer system). The computer system is not the lab. it is part of the experiment. it is the environment.
We program simple replicators which can naturally replicate and survive in the environment. Simple computer viruses. And which can undergo random variation (tweaked as you like, provided it is random).
You observe the whole system, to see if in time new complex functions arise and are selscted in the replicators which can allow them to use better the environment’s resources, and to compete successfully with the original replicators.
That’s all. It’s very simple. What are your problems with that? What has that to do with labs and fire, or with programmed fitness functions?
The fitness function is already in the laws of the computer system. Nobody creates it. Nobody should interfere with it.
It’s the replicators which should become fitter through RV and NS. Can you understand these words you use so often? “Natural” selection.
gpuccio,
The OS assigns your App “Page1000” as a RAM page and “Page1001” as your CODE page.
The OS marks “Page1000” as Readable/Writeable but not executable.
“Page1001” is marked Readable/Executable but not Writeable.
“Page1000”, the space the OS allows you to modify, cannot execute any instructions, while “Page1001” which is allowed to execute, cannot be modified.
Do you see the problem now?
Toronto:
No, I can’t see the problem.
If the replicators are replicatirs in the system, that means that they can copy themselves. Many viruses do that.
Why shouldn’t an evolved virus be better than the original?
gpuccio: Simple computer viruses.
A typical computer virus doesn’t replicate in your computer. It parks on your computer. It then *infects* across networks. Nor is it subject to random mutation.
What you’re supposed to do is carefully construct an algorithm that exhibits the features you are attempting to simulate. Then you implement the algorithm in a software environment. The same algorithm should be able to run in different software environments, giving the same results. You should even be able to do it with pen and paper, albeit very slowly in the case of an evolutionary algorithm.
Think about a weather simulation again. Scientists don’t put the computer in a wind tunnel or the refrigerator. They make a pretend world with all the features and interactions that are relevant to understanding the phenomena.
The basics of population genetics has a mathematical basis. It starts with Hardy-Weinberg who showed that, under certain circumstances, genotypes in a population will stay in equilibrium. But if there is mutation, selection, limited population, or non-random mating, then the population will change, that is, evolve. Weinberg discovered this through statistical studies in epidemiology, while Hardy derived it from pure mathematics.
We can simulate and study this fundamental process by creating a pretend world, then varying the parameters. It’s not hard to do. It’s been done, and it can be compared to biological data, such as Weinberg’s.
gpuccio,
Zachriel:
Exactly!
You should be able to show us what you expect to accomplish on a piece of paper and even run a small version of your experiment.
If it’s do-able, then find a way to run that algorithm legally on a PC.
I don’t think anyone will accept that if Microsoft stops your virus from running that you have in some way proved the mechanism of evolution doesn’t work.
The fact that you have to run your replicators as a virus is proof that you have a bad experiment.
You can’t run your experiment in a normal fashion since the operating system won’t let you and therefore you have to come up with a virus scenario.
You have set up an experiment which is guaranteed to fail.
If the OS doesn’t prevent your virus from running, the virus may prevent the OS from running!
When you first mentioned this I thought of GilDodgen’s very bad idea, now that you mention replicators, I think of kairosfocus.
The RAM you modify cannot execute without a re-boot or help from the OS to locate and load it as executable code.
You can change any bit in DATA, but the OS will vector to an exception handler if you try to execute it as an instruction.
That’s why this has to be done as a virus.
Why don’t you try to come up an experiment that will prove evolution true?
By doing that, you will have given the experiment every chance to succeed and if it still fails you will have a very good talking point in future arguments.
I would do exactly that for ID. I would try to prove that it is workable. If no one ever attempts it, why should we accept ID as something we would even consider?
Zachriel:
No. Here the problem is to test if random variation in replicators can generate new complex naturally selectable functions or not. That is all the issue. The rest is only useless discussions.
I say that in no context random variation can generate that kind of result. You say it can do it incrementally. The only purpose of a computer test must be to answer this question.
You can implement what you want, on whatever system you want, but respect these requirements:
The environment must be programmed by someone who is not aware of the purposes of the test. In blind. It must have its rules, but those rules must be set for other purposes, and not to test replicators.
The replicators can be programmed as you like, and random variation can be introduced as you like.
The system must, after time, spontaneously generate new complex functions in the replicators. The code of these new functions and their purpose must be completely new, and complex.
The system nust not in any way “measure” any property of the replicators. It’s the replicators that, thorugh RV and NS, will evolve new ways to use the properties of the system to better replicate and survice. But there must be no fitness function intentionally input in the system by the programmers.
That woudl be a test of the concept of RV + NS. Nothing else. All the rest is a form of intelligent engineering.
Toronto:
No. You can program your viruses so that they replicate, but don’t cause serious damage to the system, at least in the starting condition.
If, by RV and NS they evolve such powers that they can really threaten the system with new original functionsa, well that would be evidence of the model.
Even in nature evolved beings can threaten the system: we are a good example.
Moreover, I have no intention to do any test if that kind: you are maybe forgetting that I am the one who does not believe that teh model of RV + NS works.
Darwinists, on the contrary, who are the believers, should try to put to test what they believe in.
You and Zachriel asked why evolutionary algorithms are useless exercises which give no evidence for the darwinain model. I have answered: because the system is intelligently engineered to obtain the expected result, directly or indirectly. IOWs, they are invariably flawed by a serious cognitive bias, and by huge errors in methodology.
Then you asked how should a satisfying simulation be. I have answered. If you think it cannot work, I agree with you. It cannot work, because the model cannot work.
That’s all.
gpuccio:
Then your side should should do the same for intelligent design.
Show us that ID is more powerful than evolution.
Let’s you and I come up with respective algorithms for each of our positions.
We’ll grade them on their ability to generate novelty and modify themselves for unknown environmental changes in the future.
We’ll then run the algorithms on a computer and see what happens.
I think that’s a fair test of which theory is more valid.
Toronto:
“Show us that ID is more powerful than evolution.”
Obviously ID is more powerful than darwinian evolution. ID can create complex functional information. I am doing that in this blog. protein engineers are doing that. Computer programmers are doing that. What do you want more?
“Let’s you and I come up with respective algorithms for each of our positions.”
Any designed computer program is an algorithm for my èposition. Designed systems like so called “evolutionary algorithms” are an algorithm for my position. The truly random, naturally selected, algorithm I have described would be an algorithm for your position. And would fail.
“We’ll grade them on their ability to generate novelty and modify themselves for unknown environmental changes in the future.”
You just don’t get the point, do you?
You must match a designer, or a designed algorithm where information has been intelligently used, with a system where no information about how novelty could be generated and no “pseudo-fitness based” intelligent control of the generation algorithm has been introduced. Which will have the greatest “ability to generate novelty and modify itself”?
Difficult question indeed.
“We’ll then run the algorithms on a computer and see what happens.”
You still don’t get it.
To understand the difference between natural selection and artificial intelligent selection, consider what Szostac did and compare it to what Lenski is trying to do, or to what was done in the rugged landscape paper I cited.
Szostak has selected his molecules. How? Passing them through a lab system where he could isolate those which stuck to fixed ATP.
This is a common lab technique to separate molecules with some specific property. It is not natural selection. It is intentional, intelligent, designed selection. The molecules are not selected because they confer a positive differential replication to replicators. They are selected by the researcher (the designer). (After having been also modified by him, just to remind).
Lenski is more correct in his methodology (and, therefore, mich more unsuccesful in his results). He just fixes and environment (although potentially selective for something, which already is not true NS), but at least he lets his bacteria replicate, and he does not measure which of them has such and such property to artificially amplify it or “help” it in any way. So, except for the initial choice of a generically selective environment, he is really testing NS. His model makes some sense.
The model in the rugged landscape paper is even better. The function tested, infectivity, is a natural function. And NS is truly tested. And the results are telling.
As I pointed out before, the only limit of the model is that it is testing the retrieval of an existing, partially maintained function, and not the emergence of a new one. So, the system is correct to tell us what RV (oh, I apologize!: RM ““meaning point-mutation”) and NS can do to retrieve an existing, non completely lost function.
In these last two experimental models, and especially in the last one, the selection part is truly “NS”, because the replicators do achieve some useful function which confers them positive differential reproduction, whithout being artificially selected by added information in the system (well, in the Lenski case partially so: but still, the replicators are not directly “measured” and helped).
That is NS. All the rest is artificial selection.
So, in brief, if you implement a fitness function, you are doing artificial intelligent selection. If you implement a variation system for replicators in an environment that truly has not been engineered for the test, and let the whole system run for variation and NS, then you are doing right. But you will have no results.
Because the model does not work. Very simply, no new complex function can come out by RV + NS. And if you test the model correctly, you get no results. It’s as simple as that.
gpuccio,
The point is the phrase, “unknown environmental changes in the future.
Your ID mechanism will fail when designing for an unknown environment.
If you design a life form for a tropical climate that exists now but the environment in the future becomes very cold, you life form will fail to survive.
An evolutionary mechanism on the other hand, is designed to cope with environmental changes and respond with an adaption that might survive.
So, ID has a static goal, evolution has no goal at all.
ID is like a well designed rod and reel for fishing. It’s very functional and very specific.
Evolution is like a hand grenade. Nobody ever designed it for fishing, but I will be eating a buffet of fish before your boat even gets to your fishing spot.
Then take me up on my offer.
If you say you don’t have an algorithm yet for ID, I’ll wait until you work one out.
If you say there is no way to come up with an ID algorithm that’s supposed to test it’s functionality, then what you consider ID is useless as a theory.
Instead of just talking, let’s do some work.
I’ll accept your challenge for evolution, you accept mine for ID.
Get kairosfocus or GilDodgen to help you with the software and let’s put together some results that people here and at Uncommon Descent can evaluate.
Let’s prove our theories.
gpuccio,
To be even more fair, on each pass, we’ll let the ID side change the environment for the evolution model and we’ll let the evolution side change the environment for the ID model.
That way, no one can claim an oracle or a static function.
Toronto:
“If you design a life form for a tropical climate that exists now but the environment in the future becomes very cold, you life form will fail to survive.”
Which is more or less what happens. Ediacara phyla went extinct, like many other species.
But obviously, the designer can also incorporate adaptational algorithms which, in some measure, can deal with part of the environmental variation.
And finally, it is obvious that the designer can design new beings from the old ones, more adapted to new conditions, or simply, which IMO is the main reason, to express new, more complex functions.
Maybe you should remember that new species have been appearing throughout natural history, endowed with new protein families, new characteristics, and so on.
“An evolutionary mechanism on the other hand, is designed to cope with environmental changes and respond with an adaption that might survive.”
Ehm, you used the word “designed”…
I have already commented on evolutionary algorithms, and the limited part they can have in modeling species.
“So, ID has a static goal”
Not true. The designer is very dynamic.
“evolution has no goal at all”
Darwinian evolution certainly has no goal at all, and definitely achieves no goal at all.
“ID is like a well designed rod and reel for fishing. It’s very functional and very specific.”
No. ID is a causal explanatory model. Your idea of ID is certainly many things that ID is not.
“Evolution is like a hand grenade. Nobody ever designed it for fishing, but I will be eating a buffet of fish before your boat even gets to your fishing spot.”
Picturesque metaphor, I owe you that. Darwinists have no lack of imagination. So, I suppose evolution should be made illegal?
“Then take me up on my offer.
If you say you don’t have an algorithm yet for ID, I’ll wait until you work one out.”
An algorithm for ID? What do you mean? I and my keyboard are an algorithm for ID. I don’t understand your point. An algorithm for ID must include a designer, I supposed that was clear.
“If you say there is no way to come up with an ID algorithm that’s supposed to test it’s functionality, then what you consider ID is useless as a theory.”
I never said such a silly thing.
“Instead of just talking, let’s do some work.
I’ll accept your challenge for evolution, you accept mine for ID.”
OK, it’s simple. I have just generated a functional output of more than 500 bits (this post). I have done my part.
Now, you build a model with any environment which does not contain hidden information (according to the requirements I have outlined) and please call me when you have generated more than 500 bits of functional information (a text in english which has sense and expresses original ideas will do, not even necessary that the ideas be particularly good).
Do that by any kind of replicators (which obviously must not contain the information to be found), and any kind of random variation (I will be generous, including recombination!).
I am waiting. But, as usual, I will not hold my breath.
And I have not even used kairosfocus or Gil Dodgen!
Toronto:
“gpuccio,
To be even more fair, on each pass, we’ll let the ID side change the environment for the evolution model and we’ll let the evolution side change the environment for the ID model.
That way, no one can claim an oracle or a static function.”
I am not sure what you are thinking of, but it’s fine for me. You change my environment (please, leave me a place where to sleep the night), and then I will adjust my design.
As for you, I will wait that you have some minimal result, before.
gpuccio,
Ok, mine’s simpler. You, the designer of this post, are a product of evolution. There, I have done my part.
Are you saying the designer is still active?
gpuccio</b.: Here the problem is to test if random variation in replicators can generate new complex naturally selectable functions or not.
Apparently so, as your cite to Hayashi et al. shows. Having shown that, we’ve moved on to trying to correct your misunderstandings of simulations. Of note, you simply ignore our comments and repeat those misunderstandings.
A simulation is just an if-then algorithm. If we want to simulate a planetary system, we don’t bounce ping pong balls inside a computer housing. We devise an algorithm. Then we implement it either with paper and pencil or nowadays with a computer. But to be a valid simulation, the algorithm *must* be independent of the simulation’s substrate.
So, we create a world with bodies, bodies with masses and momentum, rules of attraction. Then we run the simulation. It doesn’t matter if we do it on an Apple computer, or a Windows computer, or use pen and paper. It’s independent of the substrate. You’ll find, by the way, that the general n-body system is chaotic and unpredictable over sufficiently long time scales.
A simple simulation in population genetics might be Hardy-Weinberg. We create a large population of diploid genomes, and see if it supports the expectation. We find that a properly constructed model will be supported by the mathematics (Hardy 1903) and by biological data (Weinberg 1903). We can then add mutation to the simulation, or use a limited population and test that against the biological data and the relevant mathematics. In this way, we can build a more complete model that allows us to study evolutionary processes.
gpuccio: I have no intention to do any test if that kind
Of course not. That was a given. But without understanding simulations, you reject the results of computer scientists and bioinformaticians who have constructed such models.
Preview or edit would be nice.
gpuccio: Now, you build a model with any environment which does not contain hidden information …
The environment does contain information. What made you think overwise?
Toronto:
” You, the designer of this post, are a product of evolution. There, I have done my part.”
That’s a rather circular way of proving evolution, isn’t it?
“Are you saying the designer is still active?”
Why not?
Zachriel:
I don’t agree with your restrictive idea of simulation.
The test I have proposed may not be a simulation as you define it. Let’s say that it is a lab test for the RV + NS model, where the only thing which is “simulated” are the replicators and the variation model, in the sense that they are inspired to the replicators and the variation model of the biological model.
So, we have a software model where the same principles of RV, NS and function are acting. If it is true that complex functions are incrementally evolvable through RV and NS, they should evolve. They will not be biological functions, they will be software functions. But the principle is the same.
With your concept of simulation, the truth is simple: you can implement all the simulations you like, but they are not a model of NS. As soon a as you predefine a fitness function which in any way “measures” artificially something in the replicator, you are out.
In NS, a spontaneous functionality in the replicator, unexpected, unprogrammed, unhelped by any intelligent designer, must be the cause of the differential survival. Any differential survival which is in any way “helped” bu the system according to a programmed plan is not, and never will be, a model of NS, or a simulation of it.
Zachriel:
“gpuccio: Here the problem is to test if random variation in replicators can generate new complex naturally selectable functions or not.
Apparently so, as your cite to Hayashi et al. shows.”
The Hayashi paper shows nothing of that kind. The function generated is neither new nor complex. The function here is pre-existing and present, even with a completely random molecule substituted. And there is no calculation of the complexity needed to retrieve partial function. Indeed, if we calculate it indirectly form how easily it is achieved, it is no example of dFSCI. On the contrary, optimization to the wildtype, although not a “new” function, seems to be a clear example of dFSCI, as it requires a starting library of at least 10^70 molecules, which is already well beyond my threshold of 150 bits.
“The environment does contain information. What made you think overwise?”
I don’t think otherwise. But it is not information about protein domains, or the way protein sequences fold. An ancient rule says that what you don’t know, you cannot say.
gpuccio,
If the designer is still active, we should be able to catch him in the act and maybe even devise a test to force him to act.
This is now something testable by the ID side.
Put life forms under stress and see how they react. If they react as evolutionary theory predicts, it’s a positive for evolution, if they react like ID theory predicts, it will be a positive for ID.
Hey gpuccio! Happy Thanksgiving!
This topic is interesting to me and we never really finished our last iteration of discussing it. I did point out at that time, though, that Tom Ray’s Tierra simulator does exactly what you suggest and did get some interesting results, including parasitism.
If you expect to see even more complex functionality, you’ll need a correspondingly richer environment. We can’t even begin to fully model the real world in software, but as our simulated environments get more complex, we do see that evolutionary mechanisms continue to generate new functionality.
Toronto:
“If the designer is still active, we should be able to catch him in the act and maybe even devise a test to force him to act.”
Catch him in the act is certainly possible, at least in principle. The idea of forcing him shows that you must have a better understanding of him than I have.
“This is now something testable by the ID side.”
We can expect to see something corresponding to design in action, but if design input is “punctuated” in time as I believe, the chances are not so big.
“Put life forms under stress and see how they react. If they react as evolutionary theory predicts, it’s a positive for evolution, if they react like ID theory predicts, it will be a positive for ID.”
I am not sure that stress is the best “argument”, both for darwinian evolution and for ID. Darwinists are probably more violent people than I am.
gpuccio,
That is a very unfair thing to say.
It’s the sort of thing you say when you want to show solidarity with your group, a group the target of your statement doesn’t belong to.
It is much easier to dismiss the ideas of people you disrespect than it is to do the same to someone you consider your equal.
That’s why evolutionists are termed “Darwinists”. This way the failings of an individual can be cast on a whole group.
We could easily start calling you “Hovindists” and then point out that “Hovindism” leads to tax evasion.
It would do nothing to dispute any good scientific arguments you might make for ID, but it would work well for political purposes and public opinion.
You of course, would immediately see through that and point it out to us, just like I’m pointing it out to you now.
MathGrrl:
Hi, fine to see you here. I was just wondering if you had followed the debate.
I well remember the discussion about Tierra.
Well, I will keep my judgement suspended about that system, because probably I don’t know it well enough, but again my impressionis that it corresponds more to Zachriel’s concept of a “simulation” than to mine of a real test. IOWs, I am not at all sure that it is not really full of hidden information. I remember that someone at UD who had more expertize than myself, some time ago, commented very negatively on it.
An interesting point you raise is the relationship between complexity of the environment and complexity of the functions.
I believe darwinists make many gross generalizations about this point.
First of all, one thing is the “complexity” of a function, another thing is the complexity of its implementation.
For instance, a protein can be very complex, but its biochemical function, for instance as an enzyme, can sometimes be described rather simply: for instance, accelerate this reaction. But to achieve that biochemical, apparently simple, task, a lot of 3d and chemical complexity is needed. You have to do that with 20 aminoacids, maybe you need 300 of them, and you have to create the correct secondary and tertiary structures, and maybe stabilize them with post-transcriptional interventions, and so on. Without considering all the regulatory and higher level integration stuff, obviously.
Now, where in the world the environment, complex as it may be, could help in that implementation?
Sometimes, the environment is simple, but the function itself can be very complex. Take flight, for instance, which has evolved apparently many times independently. I think there can be no doubts that it is a very complex function. But the reason to be able to fly is just the gross structure of our planet: a solid ball with an atmosphere all around, and the possibility to move through that atmosphere. How would that structure give information about how it is possible to fly? About how wings can work, and the muscular coordination which must give them power, and the requirements of the bone system, and the terrific flight abilities of insects, still rather mysterious for us, and so on?
The environment, complex or simple as it may be, has only a passive role. It is as it is, and it changes as it changes. It knows nothing of replicators or of life. It will never know.In a word, it is truly “random” in respect to life. (Well, there maybe the TE argument, which can be partially shared even by ID and even by me, that in part the environment could have been designed to be “supportive” of life, but I am not considering it because I assume that people here don’t believe that way).
Complexity in itself means nothing. It’s specific functional complexity that counts. That kind of complexity which can generate function. And not any possible function, anoyther myth of darwinists (up to its most recent incarnation, Toronto’s “hand grenade” giving us a “buffet of fish”). Just the function which is necessary in the context we are in. Everything else will be useless, or more probably harmful.
The problem with simulation is that the complexity of the environment is created exactly with the purpose of making the system work, of creating an evolutionary algorithm which can be published and used as a weapon against those irritating IDists. The environment is not random at all, it is indeed strongly motivated and very smartly designed. Even sometimes, maybe, with a conscious intention to hide the added information as well as possible. Or, other times, unintentionally.
That’s why I insist so much on a “blind” environment. A truly blind environment, created by truly “blind” programmers, is the only guarantee that we are testing NS, and not intelligent selection.
And the other important point, I repeat it because it is extremely important, is: the function which is naturally selected should really be naturally selected. It must not be measured or recognized in any way. It should “stand on its own legs”. There must be no active “rewards” from the system: we are not training a dog, after all.
The system is what it is. If the replicator must really evolve by random variation a new function, that function must be such that it can, of itself, allow the replicator to replicate better in the environment as it is. That’s the point.
Toronto:
I was just joking! I am not sure if I can use smiles here. Now I try 🙂 .
Well, it works.
I absolutely consider darwinists as my equals (well, maybe a little dumber 🙂 …joking again, Toronto!).
I am not very sure the other way round, anyway…
And another point: I am not a group man. As I have said many times, I am a minority guy. That is my vice. Even in my group, I am minority (you can find many examples at UD).
And I like it.
gpuccio,
Yes, the smiley worked. 🙂
You seem to know the designer better than I as you seem to recognize his designs better than I do.
That is the key thing here, that if you claim you can recognize his design as an artifact, then you should be able to recognize his new designs while they are in progress.
Come up with a way of determining this and you will be proving ID to be a valid mechanism that may have been used to put us here.
If you don’t, again, you are left with nothing but a belief.
gpuccio: So, we have a software model where the same principles of RV, NS and function are acting. If it is true that complex functions are incrementally evolvable through RV and NS, they should evolve. They will not be biological functions, they will be software functions.
But there is no variation. There’s not even replication unless you are referring to infection across a network. Indeed, a properly designed operating system will immediately disable a virus. What you are proposing is broken.
You ignored the examples given. They show the relationship between the model, the mathematics and the data.
gpuccio: As soon a as you predefine a fitness function which in any way “measures” artificially something in the replicator, you are out.
Well, no. A fitness function is an abstraction that simulates the relationship of the organisms with the environment. However, if this disturbs you, then you can simulate an environment without a fitness function.
You might try to understand simple examples before proposing a broken model.
Merry Christmas? What right have you to be merry? What reason have you to be merry? You’re poor enough.
Come, then. What right have you to be dismal? What reason have you to be morose? You’re rich enough.
–
gpuccio: OK, it’s simple. I have just generated a functional output of more than 500 bits this (post). I have done my part.
Toronto: You, the designer of this post, are a product of evolution. There, I have done my part.
gpuccio: That’s a rather circular way of proving evolution, isn’t it?
It’s a parallelism which shows the vacuity of your previous argument.
gpuccio: The function generated is neither new nor complex.
The functional domain was replaced by a random sequence. But some of the random sequences were functional enough to be selectable. Throwing words at it doesn’t change this fundamental fact.
Zachriel: The environment does contain information. What made you think overwise?
gpuccio: I don’t think otherwise. But it is not information about protein domains, or the way protein sequences fold. An ancient rule says that what you don’t know, you cannot say.
Now you got it! And that’s exactly how evolution works. The environment, whether a fitness function, a simulated environment, or a real environment, doesn’t direct the sequencing of a genome. Rather, if a genome conveys an advantage, then it is selected. And this causes the genome to evolve into complex configurations.
That’s what happens in Hayashi or Szostak’s work. They don’t tell the genome what to do. They only determine the structure after the fact. Yet, the genomes evolve from barely functional molecules into highly adapted structures, and do so in response to random variation and selection.
Zachriel:
If you want to deliberately ignore the difference between natural selection and intelligent selection (like in the Szoastk case), be my guest.
If you want to deliberately ignore the difference between evolving a new function or optimizing an existing one which has been hampered, again be my guest.
But it’s not me who am “throwing words”. It’s you who are avoiding facts.
Zachriel:
“They don’t tell the genome what to do.”
Obviously. Svostak just mutates and then selects what he decides to select. And Hayashi just sets up a system where a function has been compromised, but not completely.
For Szostak, that is a full criticism: he had no right to do what he did.
For Hayashi, it’s not a criticism, just a limit of an experimental set which is however very interesting.
And you are right, they don’t tell the genome what to do. Have I ever said differently?
Zachriel:
“Well, no. A fitness function is an abstraction that simulates the relationship of the organisms with the environment. ”
It is a designed abstraction, made with purpose.
“However, if this disturbs you, then you can simulate an environment without a fitness function.”
I am not disturbed at all. I just give no credit to those results of those designed systems. If darwinists want to be more convincing, it’s them who have to “simulate an environment without a fitness function”. In blind.
And then simulate the replicators. And then just look for natural selection of replicators for their intrinsic function, without any measurement or active reward from the system.
gpuccio:
Evolutionary theory says that your environment rewards you with the ability to replicate.
Evolution doesn’t care “explicitly” about any specific function at all, only that you somehow survive to reproduce.
How you do that depends only “implicitly” on your set of functions, and pure luck that the environment does not become lethal to you suddenly.
That IS your reward, and your ONLY reward, that you can reproduce/replicate.
If you take that particular reward out of your test, then you are not testing evolution at all.
MathGrrl:
there is one more aspect in the “environment and complexity” problem which I would like to stress.
A great part of biological complexity is not directly finalized to the environment, but rather to create the context of life itself.
I will take the simplest, and strongest, example: the DNA – transcription – translation system, including the genetic code and the aminoacyl-trna synthtetases.
That whole system has one big function: keep the biological information stores in prtoen coding genes, and translate it into functional proteins.
That function is necessary for life, whatever the environment. It is intrinsic in the biological system, it is not a “response” to some characteristics of the environment.
That complex fucntional system is found in all known life, and it is remarkably constant. According to current theories, it was already formed very early in the history of life (at least in LUCA). According to my personal opinion, it started with life itself.
Was it helped by information in the environment? What part of the environment could lend the information that it is necessary to have information to get a complex systems, that it is necessary to store it and to be able to retrieve it when necessary? That different molecules (DNA, RNA, proteins), none of which exists in the environment, are perfectly suited to perform different tasks in such a system?
And so on.
gpuccio: If you want to deliberately ignore the difference between natural selection and intelligent selection (like in the Szoastk case), be my guest.
We’re not ignoring it. It’s called evidence (Darwin 1859).
gpuccio: If you want to deliberately ignore the difference between evolving a new function or optimizing an existing one which has been hampered, again be my guest.
The domain was replaced with a random sequence. Hence, the domain had no functional capability except insofar as one of the random sequences might have function.
gpuccio: For Szostak, that is a full criticism: he had no right to do what he did.
No right? That’s funny. It was a hypothesis straight from abiogenetic theory. Lucky guess, huh?
gpuccio: If darwinists want to be more convincing, it’s them who have to “simulate an environment without a fitness function”.
Let’s say we create a space in which food grows. Those organisms that reach the food, replicate. Otherwise, they eventually die from starvation. The organisms have uncoordinated motility, and we see if they can evolve coordinated motility.
But if you want to propose a clearly defined algorithm, then we can talk. But if you can’t even describe the process in algorithmic terms, then it means you probably don’t understand how evolution is posited to work.
gpuccio: According to current theories, it was already formed very early in the history of life (at least in LUCA). According to my personal opinion, it started with life itself.
Can you propose a test?
The evidence thus far suggests primordial biology before the LUCA.
gpuccio: Was it helped by information in the environment? What part of the environment could lend the information that it is necessary to have information to get a complex systems, that it is necessary to store it and to be able to retrieve it when necessary?
As you don’t understand how evolution works in less ancient times where we have very strong evidence, it will be difficult to understand theories of abiogenesis that are still largely tentative and speculative.
Toronto:
“Evolutionary theory says that your environment rewards you with the ability to replicate.”
No. I used the words:
“without any measurement or active reward from the system”
exactly to avoid misunderstandings. But it seems that english language is not enough to avoid misunderstandings with darwinists.
a) First: “measurements”. The system must not have any algorithm which measures any specific function.
b) Second: “active” reward. The system must not “actively” reward (that is, though a specific programmed algorithm which reward what has been measured) the replicators.
Those are characteristics of intelligent engineering, not of NS. You cannot simulate NS by intelligent engineering. It is a contradiction in terms.
So, evolutionary theory does not say that your environment rewards you with the ability to replicate.
Evolutionary theory says that your environment allows you to replicate if you have the ability, and if your ability is better, in that environment, than the ability of the others, the result is positive differential reproduction (natural selection).
The environment does not reward anything. The environment is what it is. Please, stop using ID metaphors against ID.
Zachriel:
“The evidence thus far suggests primordial biology before the LUCA.”
What evidence?
“As you don’t understand how evolution works in less ancient times where we have very strong evidence, it will be difficult to understand theories of abiogenesis that are still largely tentative and speculative.”
Just to paraphrase:
As nobody understands how evolution works in less ancient times because we have no evidence, it will be even more desperate to understand theories of abiogenesis that are, if possible, vastly more tentative and speculative.
Zachriel:
And yes, Zvostak had no right to do what he did.
gpuccio,
Here’s the next line of my comment explaining the metaphoric sentence that you commented on.
You then proceed to “agree” with me, in a way where even I, (according to you, a Darwinist with English language issues), can finally understand what it is that I actually told you.
Here again you don’t seem to understand what a computer simulation means.
Intelligent engineering is how we build the platform, controls, and measurement techniques we will use when we run the simulation of the model under test.
This is how simulations are done for earthquake tolerant buildings. Intelligent engineering “rewards” the “resistance” of a “steel beam” to “warp under stress”.
All quoted items don’t actually exist. No reward, no steel beam and there is no warping of any part of the PC’s housing or even the OS that ran the simulation.
As Zachriel has mentioned before, numbers are crunched and algorithms are tested.
If scientists have achieved success by modeling the real world in this manner, why can’t it be done for evolution or ID?
gpuccio,
You say NS cannot be tested with intelligent engineering, but then on the flip side of that coin, ID should be!
Let’s come up with testable models. Let’s pit ID against evolution and see what happens.
Trying to dismiss an attempt at science is not doing your side any good at all.
If you want to help ID be accepted by the scientific community, come up with a model.
ID is the only field I have ever seen that has not been excited about being able to use computers to further their own knowledge.
The only use your side seems to have for computers is to try and prove our side wrong instead of trying to prove yourselves right.
gpuccio: Without the expansion, the probabilities multiply, the probabilistic resources for the second event are 10^12 times lower than those for the first event, and therefore the total time needed is 10^12 times greater than the time needed for a single event.
With the expansion, the time needed to have both events in an individual is more or less double than the time neede for one event, plus the time needed for the expansion.
So, Spetner is wrong.
And gpuccio is right.
This is something we can easily simulate. If you remember, a simulation is just an elaborated if-then, which you have just outlined. From above, we can create a simulation that approximates Hardy-Weinberg Equilibrium. Now, we add selection. According to your scenario, we don’t have to show all the details of selection. Some trait is conferring a reproductive advantage, which we can represent with a selection coefficient. From there, we can watch as the simulation evolves, including the introduction of mutations; neutral, selective, epistatic.
If 1/4Ne << s << 1 then the probability of fixation is 2s. We can show this mathematically and we can verify it, and get a better grasp on the process, with such a simulation. Given reasonable time, meaning enough trials, it is inevitable that the mutation will become fixed in the population. Then it's just a matter of time before the second, epistatic mutation occurs and fixes.
Toronto:
“Here again you don’t seem to understand what a computer simulation means.”
You and Zachriel are confused. It’s not that I don’t understand. It’s that the kind of simulation you offer does not simulate NS at all.
“Intelligent engineering is how we build the platform, controls, and measurement techniques we will use when we run the simulation of the model under test.”
Indeed. And the point is, you need not measure anything, except the results. Differential reproduction ,ust be a result of a true new reproduction function, not of other functions “measured and rewarded”. Other wise, you are not simulating NS, but intelligent engineering. You can certainly simulate IE by IE, but that will not be a simulation of NS at all.
“This is how simulations are done for earthquake tolerant buildings. Intelligent engineering “rewards” the “resistance” of a “steel beam” to “warp under stress”.”
That is a correct simulation. Builidngs are intelligently engineered, they don’t arise because of NS.
“If scientists have achieved success by modeling the real world in this manner, why can’t it be done for evolution or ID?”
It depends on what you are trying to model. You can easily model necessity mechanisms. And you can model random events. But you must do those things correctly.
When you model random events, you must be sure that you are modeling them randomly.
If you are modeling a necessity mechanisms, you must be sure that you are modeling that mechanism, and not another one.
The distinguishing characteristics of NS is that the function selected arises spontaneously and has itself the capacity of generating positive differential reproduction without being explicitly measured and without being actively rewarded. The function must be the cause of its own reward, so to speak.
If you don’t respect this property, you are not modeling NS.
And if you want to model ID, you must insert a conscious designer in the model.
Toronto:
“Trying to dismiss an attempt at science is not doing your side any good at all.”
You are confused again. I am well ready to do the test you syggest.
So, you build a computer model which really corresponds to RV + NS. And I give you my model: a conscious designer at a computer.
I could volunteer, if you can pay my expenses 🙂 .
I am waiting.
Zachriel:
I am happy we agree on that point.
As you can see, I try to defend what I objectively believe to be right, either it is in accord with what others in ID say, or not.
I have always made clear that probability calculations must be reserved to the parts where NS does not intervene.
On the other hand, the intervention of NS must be credibly detailed and realistically modeled, and not assumed by blind faith.
gpuccio: As you can see, I try to defend what I objectively believe to be right, either it is in accord with what others in ID say, or not.
Of that we had no doubt. No one has yet responded to your comment.
In any case, your claim is that such models can’t be built, yet you yourself just created a qualitative model, a model which can be improved, verified against the mathematics, and then checked against the data.
gpuccio: But, if the first mutation in itself gives the single individual a definite reproductive advantage in the population, IOWs if it is selectable by NS, after a finite time the whole population will be made of descendants of the mutated individual, each with the first mutation. That is the darwinist scenario …
Fixation depends on a number of factors detailed above.
gpuccio: (which I don’t believe to be true, because complex functions are not made of individually selectable steps; but here we are assuming it for our reasoning).
The “don’t believe to be true” part only refers to the evolution of complex function. At this point, our model only concerns the evolution of two separately selectable traits.
gpuccio: I have always made clear that probability calculations must be reserved to the parts where NS does not intervene.
Your very own model assumes that there exists a trait which confers a reproductive advantage. It’s a valid model of how selection can propagate in a population.
gpuddio,
You have a great suggestion there.
Just as a chess program competes with a real player, we could have an evolution program compete with and intelligent designer sitting in front of a computer.
Both can analyze each generation’s performance and prepare for the next unknown environment they will have to survive in.
The environment will be changed each generation by an unknown third party.
The evolutionary program and the designer will not know how well each others lifeforms are performing and neither will know what the next environment will look like.
So Dembski’s oracle is not in the picture for the evolutionary program which should negate any improbability arguments.
What do you think?
Sorry, that’s gpuccio, not gpuddio! 🙂
Zachriel:
“The “don’t believe to be true” part only refers to the evolution of complex function. At this point, our model only concerns the evolution of two separately selectable traits.”
Well, that’s true, but my idea was that the two separately selectable traits should be steps to one new function (even if only of two mutations). IOWs, something like Behe’s cloroquine resistance, if the two mutations were separately selectable (which does not seem to be the case).
There is obviously no difficulty in having many single mutation new functions, independently achieved and selected.
The problem is when a more complex new function emerges.
So, just to be clear, if form A a new protein B emerges, with a new function which absolutely requires a 4 AAs change (4 coordinated mutations), there are two possibilities:
a) None of the 4 single mutations is individually selectable (and none of their combinations, except for the whole 4 AAs change). In that case, probabilities multiply, and the necessary time (all other parameters being the same) increases exponentially.
b) The single mutations are individually selectable for some definite reproductive advantage (more or less unrelated to the new function of B) they confer. In that case, each of them, as soon as it happens, will expand in a time t (let’s say for simplicity to the whole population, even if that needs not be the case). Therefore, probabilities do not multiply, and the final time is a relatively reasonable multiple of the time necessary for each mutation.
Intermediate situations are obviously possible.
So, such a model can be applied only to complex functions which are deconstructable into individually selectable steps.
Toronto:
“What do you think?”
It’s fine for me, provided the “evolutionary” algorithm really models RV + NS. I would not like to compete with some smart algorithm, like a chess program (or worse, Gil’s checkers program). I am not so good.
The necessary premise is that the evolutionary algorithm may really bear some functional fruit. Just stay there waiting would be boring.
Mark:
I would recommend that you consider carefully the above discussion about the importance of expansion, and its implications for the original subject of this thread, the missing expanded intermediaries.
I believe you have stated somewhere that expansion is not really necessary for the darwinian mechanism to work. I can’t see how that can be.
Without expansion, functional mutations behave just as neutral ones (except for the small contribution of fixation by negative NS). The whole model looses then its “necessity” part, and becomes almost completely random.
gpuccio: Well, that’s true, but my idea was that the two separately selectable traits should be steps to one new function (even if only of two mutations). IOWs, something like Behe’s cloroquine resistance, if the two mutations were separately selectable (which does not seem to be the case).
We merely represented the simple model that you yourself qualitatively outlined.
gpuccio: There is obviously no difficulty in having many single mutation new functions, independently achieved and selected.
Or, if they exist, evolution through selectable pathways, as you stated. We’ll get to that, but you still seem confused on simulations.
So, to review, while this simple simulation may not answer your questions about complex adaptations, it is a valid model. We can compare our simulation to natural situations, such as bacteria evolving in response to antibiotics, and show that the model reasonably represents how selection works in such populations. Once having done this, we can then improve our simulation to account for other effects, such as epistasis.
Or recombination.
Zachriel:
Please remember that in that “simple model” we have just assumed that a function is selectable, but in no way we have modeled natural selection. There is nothing in the model which says which mutations will be naturally selectable and which will not.
That is the point you have to cover first.
gpuccio: Please remember that in that “simple model” we have just assumed that a function is selectable, …
Yes, we assume in the model that a mutation can be selectable. We can check this assumption against biological observation, and that is exactly what we find.
gpuccio: but in no way we have modeled natural selection.
We have not provided a detailed model of the source of variation, but we have, indeed, provided a testable model of selection.
Zachriel:
“that is exactly what we find.”
In what sense?
“we have, indeed, provided a testable model of selection”
Of selection, not of natural selection.
gpuccio,
Let’s more clearly define exactly what we are doing.
The only requirement for our test is that you input intelligence and that we don’t.
A computer can mutate 100,000 members of a population in less than a second.
How fast can you type?
We are not talking of a project for entertainment here. I am serious about finding something out from this simulation.
The goal is not not make you a good player at something, it’s to prove an evolutionary algorithm.
We are trying to prove that no intelligence is required to guide the algorithm.
You don’t believe that evolutionary algorithms work but you seem to be suggesting how we must incorporate our algorithm under test.
To you, this is a simple black box test. The black box doesn’t know how the environment is going to change and neither do you.
Can the black box react to unforeseen changes and still generate survivable offspring with each generation?
If the black box doesn’t require your guiding intelligence, it doesn’t need any outside intelligence at all.
Zachriel: Yes, we assume in the model that a mutation can be selectable. We can check this assumption against biological observation, and that is exactly what we find.
gpuccio: In what sense?
The statement should be quite clear. We can observe mutational variations that lead to reproductive advantage.
Zachriel: we have, indeed, provided a testable model of selection
gpuccio: Of selection, not of natural selection.
That is incorrect. If we have a beneficial mutation, the model shows how that mutation will spread through the population even if the benefit is due to competition in nature.
Now, the model does assume *the existence* of beneficial mutations, that is, mutations that provide a reproductive advantage. But, as we can observe beneficial mutations, the model can be used to predict the spread of the mutation. And these predictions fit the observations.
Gpuccio seems to be an intelligent guy, but it´s almost painful to watch how he struggles with the concept of a “model”. This seems to be a common problem of ID people, but I really wonder why? What´s so difficult about this?
Toronto:
“You don’t believe that evolutionary algorithms work ”
You are equivocating. I don’t believe that evolutionary algorithms model natural selection. Evolutionary algorithms modeling intelligent selection do work, as should be clear from all my discussions above.
I don’t want to measure myself against an intelligent algorithms finding an intelligent solution to some specific problem. But I will gladly measure myself against a RV model whose only specific property is to select reproductive advantage in the replicators.
I believe I can be much better at increasing reproductive ability in the replicators, even at my slow typing rate (and possibly assisted by some smart virus programmer!).
Toronto:
“Can the black box react to unforeseen changes and still generate survivable offspring with each generation?”
Let’s try and see.
However, I believe the original idea was that it should produce new complex functions. After all, mere survival has been achieved very successfully by the first known living beings, bacteria and archea.
Zachriel:
“We can observe mutational variations that lead to reproductive advantage.”
All of them microevolutionary, that is of a complexity of one or two aminoacids.
All of them minor variations of an existing structure, usually working as a loss of function which accidentally confers protection from an environmental threat (as well documented by Behe in TEOE). Or, possibly, (but I am not sure I am aware of definite models of that), as an optimization of an existing function.
And yes, that kind of variation can be selected under strong selective pressure.
I can’t see how that can be generalized to the concept that such variations can be naturally selectable steps of more complex functions. There is no evidence of that, neither in nature nor in your models.
Zachriel:
“Now, the model does assume *the existence* of beneficial mutations, that is, mutations that provide a reproductive advantage. But, as we can observe beneficial mutations, the model can be used to predict the spread of the mutation. And these predictions fit the observations.”
You are saying that, if we assume beneficial mutations which can be naturally selected, that is which can of themselves give a reproductive advantage in the system, without any intelligent intervention to select them, then those mutations will be naturally selected and will spread.
But I have always agreed on that. So, why all the discussion?
Zachriel:
If all your model is saying is that microevolutionary variation of the kind we already observe in lab situations can spread under due selective pressure, again it should be clear that I have always agreed with that. That is no big issue.
Indiumas:
Thank you for the compliment.
I am sorry if I cause you pain at all, but indeed I am not struggling with the concept of a model.
I am only saying that the models suggested here are not good to model natural selection.
The reason should be clear now, but I will repeat it again just the same.
The distinguishing mark of NS is that it is a process which can recognize and select only one thing: a reproductive advantage.
Indeed, nothing is selected and recognized (these are just ID metaphors which have unfortunately become common, but that does not make them true). We can say That NS is a process where a function selects itself, because the fucntion is a reproductive advantage.
No other type of function can select itself. All other functions must be recognized and rewarded if they have to be selected, and not in metaphor, but in reality. That is called intelligent selection.
So, we have a simple difference: all functions can be intelligently selected, at any level. But only a minuscule set of functions cab be “naturally selected”.
That is a big, big difference.
If you build a model on the intelligent selection of a function, you are building a model of intelligent selection. That is no model of NS.
A model of NS should recognize only the functions which are naturally selectable. The only way to do that is to measure the spontaneous reproductive rate of replicators in some environment, independently from any measurement of reward. And even if you measure the emergence of that kind of function, still there should be no need to reward it, because the function is already “rewarding” itself.
I really hope you will not struggle with these concepts.
Gpuccio,
if you run a simulation which simulates certain aspects of evolutionary processes you can include a fitness function and base the reproduction rates of your organisms on this fitness vale. In this way you can model the effect of different reproductive success in nature, and you can then see how selectable or neutral mutations and variants can be fixed in populations. For many questions this is a viable route (see the Red Lynx simulator on pandasthumb.org for example).
Of course in principle you can also build a model without an extrinsic fitness function. This has been done in Tierra. Maybe instead of just relying on something you vaguely remember being said on UD you take a look for yourself and tell us what is wrong with it?
gpuccio,
In order for generation (X) to survive in environment (X + 1), is to modify itself in some way that will allow it to survive and reproduce.
A mutation may not help, but might not hurt your survivability, and so one population member might survive with different functionality than another.
That is the blind test to be passed. to survive, to reproduce and to pass on your new “information”.
We are incrementally, generation by generation, modifying “information”, and thus, functionality.
Can the black box do this without a guiding intelligence?
Agreed. The black box will “intelligently select” nothing.
There will be no intelligent input.
Can the black box, with no guiding intelligence, and no feedback, mutate “information”, and thus the functionality of that “information”?
According to ID, not only should a process guided by intelligence outperform the black box, the black box should not be able to produce new “information”, and thus functionality.
gpuccio: All of them microevolutionary, that is of a complexity of one or two aminoacids.
That’s irrelevant to the point. You said we couldn’t simulate natural selection, when you yourself proposed such a model. We can’t even begin to discuss complex adaptations when you have such troubles with simple models.
gpuccio: You are saying that, if we assume beneficial mutations which can be naturally selected, that is which can of themselves give a reproductive advantage in the system, without any intelligent intervention to select them, then those mutations will be naturally selected and will spread.
Of course the model assumes the existence of beneficial mutations. As we observe beneficial mutations, this is a reasonable assumption.
gpuccio: But I have always agreed on that.
Because you claimed we can’t have a testable model of natural selection. But beneficial variations are observed, and we can use to model to accurately predict the trajectory of such variations in populations.
gpuccio: If all your model is saying is that microevolutionary variation of the kind we already observe in lab situations can spread under due selective pressure, again it should be clear that I have always agreed with that. That is no big issue.
Which you follow with:
gpuccio: I am only saying that the models suggested here are not good to model natural selection.
Sigh.
gpuccio: If you build a model on the intelligent selection of a function, you are building a model of intelligent selection. That is no model of NS.
This is where you are becoming confused. If there is selectable natural variation, then we have quite valid models of natural selection. Your problem is with modeling variation.
gpuccio: A model of NS should recognize only the functions which are naturally selectable. The only way to do that is to measure the spontaneous reproductive rate of replicators in some environment, independently from any measurement of reward.
You are attempting to draw a picture of *evolution* by natural selection. Ideally, we would want to model replicators in an environment with limited resources.
Indiumas:
“you can then see how selectable or neutral mutations and variants can be fixed in populations.”
That’s fine. But that’s not the problem. The problem is how naturally selectable variation arises. And, more specifically, if any complex naturally selectable variation can ever arise. That is what we have been debating here. Don’t forget that the only reason we are debating evolutionary algorithms is because Zachriel cited them as support to the darwinian model for the emergence of complex functional information.
The work of Dembski and Marks is aimed exactly at demonstrating that, when evolutionary algorithms seem to attain some interesting results, that’s because intelligent information about the search has been introduced in the system. There are many ways that can be done, and the fitness function is not the only one.
That’s why I have suggested two criteria to ensure that no intelligent information has been added to an experimental informational system testing RV and NS:
1) The system must have been programmed independently from the replicators, by some other programmer, and in blind. That is to avoid any cognitive bias, and any possible introduction of active information about the replicators or about possible functions to be selected into the system.
2) The reproductive advantage of the replicators must be intrinsically functional and select itself without any outer measurement and reward system.
If in such a system dFSCI can arise, my position is falsified.
Regarding Tierra, before I spend a lot of time enquiring about it, can you answer two simple questions?
a) Do you believe it satisfies the tow above requirements?
b) Has any complex function come out of it? And if possible, how complex?
Thank you for your cooperation.
Toronto:
“Can the black box do this without a guiding intelligence?”
Maybe, but not producing new complex information. If simple, non cumulative variation can do that, it will probably do that. At least in some cases. But no new complex information will come out of that.
Toronto:
“According to ID, not only should a process guided by intelligence outperform the black box, the black box should not be able to produce new “information”, and thus functionality.”
That’s correct. At least, no new complex information. Simple functional changes are in principle possible (although IMO rare).
Zachriel:
You can try to maniulate my words as much as you like, but it’s only you who generate confusion.
If in your model you assume NS, then your model is in no way testing if NS will happen and what it can do. The only thing your model assumes is that, if something is assumed as selectable, it will be selected and expand. Very interesting indeed.
So, let’s go step by step.
We have an environment. We have replicators. We have random variation.
The first question is: what type of new functional information can random variation produce of itself?
You say that we observe beneficial naturally selectable variations in nature. That’s true, but they are always simple, and they are selected only in special conditions of strong selective pressure. we do know that, in those conditions, those simple variations will expand. We have examples of that, for instance, in antibiotic resistance. I have always admitted that, for the simple reason that it is true.
Again, if your model wants only to model mathematically this scenario, you are welcome, but what relationship has that with our discussion? We already know that it works, I have never denied it, so what’s the purpose of it?
You sigh at my phrase:
“I am only saying that the models suggested here are not good to model natural selection.”
And yet, the meaning should be very clear, in the context of all the discussion. But you like to take phrases our of context.
What I have said many many times is that, if you want to demonstrate by a simulation that NS can produce complex information, you need a model where NS really happens, and not a model based on intelligent selection. IS can certainly produce complex information.
None of your models offered any guarantee that NS was being observed, and that possible results had been naturally selected.
So again, if you are saying that assuming NS, what is naturally selected will expand, I agree, but it is only a tautology.
If you are saying that your model shows how NS can lead to new complex information, you are simply wrong.
gpuccio,
If our black box can change one bit per successfully replicating generation, after 150 generations, 150 bits of information have changed.
I believe that 150 bits meets your requirements for “complex information”.
Is that true?
Toronto:
“Is that true?”
No, it isn’t. After all this time, you still make this simple error.
150 bits of dFSCI means a function for which 150 bits of variation are needed for the function to be available.
Bits scattered in different functions have no meaning. dFSCI is the functional information needed for one function.
Toronto:
To be more clear:
If you can vary 150 bits in one protein, in the sequence and time you like, to attain a new function for which those 150 bits are required, than that is dFSCI.
gpuccio,
Let’s try to clarify what we are really talking about.
We are trying to see if an algorithm can generate the required functionality to survive and reproduce in environment (X).
In order to survive environment (X + 1), the algorithm may or may not require any functionality change at all.
There may be some environments that do require a change or additional new function.
In our simulation, we will change the environment every single generation just to stress the offspring.
Do you agree, that if the offspring generated by the black box with no knowledge of the future, have whatever new functionality is required to survive to let’s say, generation ( X + 1,000,000,000 ), and no intelligence guides the development of the offspring through the generational changes, that the evolutionary algorithm has passed your test?
Zachriel: If there is selectable natural variation, then we have quite valid models of natural selection. Your problem is with modeling variation.
gpuccio: You can try to maniulate my words as much as you like, but it’s only you who generate confusion.
gpuccio: The problem is how naturally selectable variation arises.
gpuccio: If in your model you assume NS, then your model is in no way testing if NS will happen and what it can do.
It doesn’t test the origin of beneficial mutations, because it’s not a model of the origin of beneficial mutations. It’s a model of selection.
gpuccio: The only thing your model assumes is that, if something is assumed as selectable, it will be selected and expand.
That’s not necessarily correct. The trajectory of a beneficial mutation is not necessarily positive, and may even result in extinction. That’s why we construct testable models.
Zachriel:
I really don’t understand what you want to demonstrate.
Again, the only object of this discussion has been if the darwinian model can generate complex functional information or not.
You have raised the issue of evolutionary algorithms to support the explanatory power of the darwinisn model.
Again, if the only purpose of your model is to “trace the trajectory” of a beneficial mutation, that’s fine for me, but I have no special interest in that. I have declared that, if single naturally selectable mutations can build a new complex function, then such a function is in the range of the darwinian model. I am conceding that the naturally selectable trait will expand, even if I am aware that it may not always be the case. And I have stated clearly, both at UD and here, that I have no problem with that.
The problem, as I have stated many times, is that:
1) Naturally selectable variation is rare and simple.
2) In no way it can build complex functions.
So, if you have any model which can falsify these statements, bring it on, otherwise it’s useless that we go on talking of models to prove something we already agree upon.
Toronto:
Ahain, to you as to Zachriel:
the problem is how complex functions arise. The problem is not how a living organisms can survive changes in the environment. When have I ever said such a thing?
If a replicator can survive changes through simple mutations, it will probably do exactly that. In principle, that can happen. In practice, sometimes it happens.
A sum of simple mutations is not a complex mutation, unless those simple mutations contribute to a new complex function.
Is that clear?
Toronto:
“Do you agree, that if the offspring generated by the black box with no knowledge of the future, have whatever new functionality is required to survive to let’s say, generation ( X + 1,000,000,000 ), and no intelligence guides the development of the offspring through the generational changes, that the evolutionary algorithm has passed your test?”
The evolutionary algorithm will pass my test if, and only if, it generates at least one new functional structure through a transition of at least 150 functional bits, through a mechanism of mere random variation (of any type) and true natural selection (satisfying the requirements I have detailed).
I hope that’s clear, after having repeated it tens of times in a completely unequivocal way.
Zachriel:
“Your problem is with modeling variation.
gpuccio: You can try to manipulate my words as much as you like, but it’s only you who generate confusion.
gpuccio: The problem is how naturally selectable variation arises.”
You are wrong. “how naturally selectable variation arises” can be decided only through a correct modeling of NS. If I had said “how variation arises” you would be right. But I have said “how naturally selectable variation arises”: you cannot model that, unless you clearly model natural selection.
So, my problem is definitely not only with modeling variation, but also with clearly defining and modeling what naturally selectable variation is.
gpuccio: So, let’s go step by step.
That’s the best way.
gpuccio: We have an environment. We have replicators. We have random variation.
The environment should represent limited resources required by the replicators in order to replicate. That means the presupposition of existing replicators means some primitive ability to acquire those resources.
gpuccio: The first question is: what type of new functional information can random variation produce of itself?
gpuccio: You say that we observe beneficial naturally selectable variations in nature. That’s true, but they are always simple, …
They are generally incremental changes, but then, that’s the whole point!
gpuccio: and they are selected only in special conditions of strong selective pressure.
That is incorrect. If 1/4Ne << s << 1 then the probability of fixation is 2s. Again, that's why we make models, in this case, a mathematical model.
gpuccio: So again, if you are saying that assuming NS, what is naturally selected will expand, I agree, but it is only a tautology.
It’s not only not a tautology, but it’s not even necessarily true.
You sometimes seem to be arguing that {evolution by} natural selection can’t explain complex adaptations, but then you always cap it off with this:
gpuccio: None of your models offered any guarantee that NS was being observed, and that possible results had been naturally selected.
Natural selection is something that can be observed in a variety of circumstances. It’s a consequence of fecundity, variation and limited resources. If you want to discuss whether evolution by natural selection is a sufficient to explain complex adaptations, we can have that discussion, but you need to be clear in your argument.
Yes, and that is what evolution does!
gpuccio: Again, the only object of this discussion has been if the darwinian model can generate complex functional information or not.
A small percentage of random sequences can be functional, even providing a reproductive advantage, such as in phage.
gpuccio: You have raised the issue of evolutionary algorithms to support the explanatory power of the darwinisn model.
We introduced evolutionary algorithms to show that recombination, an observed mechanism of variation, is essential to escape local fitness peaks.
gpuccio: I have declared that, if single naturally selectable mutations can build a new complex function, then such a function is in the range of the darwinian model.
That’s where we thought we were long ago. In that case, we might consider well-established instances, such as the evolution of the mammalian middle ear, which exhibits cooption and irreducible complexity.
gppucio: “how naturally selectable variation arises” can be decided only through a correct modeling of NS.
It would have been thought you would have dropped this line of argument. A model of natural selection presupposes the existence of selectable variation, just like a model of evolution presupposes the existence of replicators, and a model of gravity presupposes bodies.
However, as we observe selectable variation, it’s not something that is in much doubt. But you’re not consistent.
gpuccio: I have declared that, if single naturally selectable mutations can build a new complex function, then such a function is in the range of the darwinian model.
You {sometimes} seem to be accepting of natural selection of traits, but that there are no pathways leading from these simple, selectable variations and the building of complex structures.
We certainly can’t show the evolutionary path for every complex structure. Most don’t leave fossils. Most molecular structures evolved eons ago. But we can show that some complex structures evolved, and can predict homologies for large classes of others.
Zachriel:
“But you’re not consistent.”
And you’re not clear.
“You {sometimes} seem to be accepting of natural selection of traits,”
I have always accepted that some traits can be selected.
“We certainly can’t show the evolutionary path for every complex structure. ”
Let’s say for none.
“Most don’t leave fossils.”
My discussion has nothing to do with fossils, only with molecular biology.
“Most molecular structures evolved eons ago.”
Not all of them. And eons ago the laws of reality were not different.
“But we can show that some complex structures evolved”
How? Through RV + NS? Which?
“and can predict homologies for large classes of others”
And so? Proteins in the same superfamily usually show some homology. And so? Homology is at most evidence of CD, not of causal mechanism. You just repeat the same non arguments.
Please, answer some simple questions:
a) Do you believe that complex molecular structures evolved through RV + NS?
b) Have you any direct evidence, either in biological systems or in informational models, that such a thing happens?
Zachriel:
“We introduced evolutionary algorithms to show that recombination, an observed mechanism of variation, is essential to escape local fitness peaks.”
I must have missed that part. Could you explain better?
gpuccio: I have always accepted that some traits can be selected.
And we can certainly model changes to a population when a trait is under selection. Good.
Zachriel: We certainly can’t show the evolutionary path for every complex structure.
gpuccio: Let’s say for none.
Okay.
gpuccio: My discussion has nothing to do with fossils, only with molecular biology.
Changed your mind already? Is there a reason why you exclude macroscopic evolution where our evidence is more complete?
gpuccio: Homology is at most evidence of CD, not of causal mechanism.
Common Descent provides the historical framework. But you’ve already excluded whole classes of evidence.
gpuccio: a) Do you believe that complex molecular structures evolved through RV + NS?
The evidence indicates that evolution by natural selection can lead to complex structural adaptations.
gpuccio: b) Have you any direct evidence, either in biological systems or in informational models, that such a thing happens?
Yes. We can observe evolution solve complex problems, e.g. by balancing multiple constraints.
gpuccio: a) Do you believe that complex molecular structures evolved through RV + NS?
For clarity; yes, though other mechanisms may be involved.
Zachriel:
“gpuccio: My discussion has nothing to do with fossils, only with molecular biology.
Changed your mind already?”
Why do you say that? I have changed nothing. I quote form one post of mine in the opening part of this thread:
“I am saying that darwinian evolution does not work at all as an explanatory model. Macroscopic forms are only a consequence of genomic information, and the relationship between the two things is too poorly understood to build explanatory models on macroscopic considerations.”
Your position is rather clear now, and so should be mine. Why insist in attributing to me positions I have never had?
“I am saying that darwinian evolution does not work at all as an explanatory model. Macroscopic forms are only a consequence of genomic information, and the relationship between the two things is too poorly understood to build explanatory models on macroscopic considerations.”
___________________________________________
Evolution doesn’t require prior understanding of the effects of genomic change in order to work. It isn’t necessary to understand the effects of genotype on phenotype, and it isn’t necessary to understand the effects of phenotype on reproductive fitness.
Design, on the other hand, would require all of this, plus an ability to forecast changes in geology, cosmology and the ecosystem. Are you exempt from explaining where this information comes from and how it is acquired?
Evolution acquires this information by learning. By trial and success. You may not like this process and may think it is inadequate to the task, but it is the only process on the table.
Zachriel: We certainly can’t show the evolutionary path for every complex structure.
gpuccio: Let’s say for none.
gpuccio: My discussion has nothing to do with fossils, only with molecular biology.
The first statement indicates we can’t show an evolutionary path for any complex structure. The second excludes macroscopic structures. This is your reason.
gpuccio: Macroscopic forms are only a consequence of genomic information, and the relationship between the two things is too poorly understood to build explanatory models on macroscopic considerations.
Yet there is ample evidence of incremental and selectable evolution in macroscopic structures, e.g. the mammalian middle ear.
gpuccio,
Actually, it’s a lot more complex than that. The real world has all the laws of physics and chemistry that are not typically modeled in a simulation. Not only has this resulted in flight evolving multiple times, but also in specialization. Compare the wings and flight style of a condor and a seagull, for example. The complexity of the real world means there are more niches to exploit than in a simulation.
That certainly doesn’t mean that we can’t learn from simulations, of course. The most complex aspect of the Tierra environment is actually the other simulated organisms. It’s very interesting that evolutionary mechanisms such as mutation and crossover resulted in virtual parasites that took advantage of those components of the environment.
My core point is that the more complex the simulated environment, the more ways there are to exploit it. That’s why we would expect to see more complex functionality evolve in a more complex simulation.
The environment, whether real or simulated, doesn’t explicitly give any information. It is the stage on which members of populations act out their lives. The genetic make up of the population will change over time as evolutionary mechanisms operate and individuals leave more or fewer offspring relative to each other based on how well they take advantage of the environment. Natural selection is a result of this process, not a process itself.
Do you disagree with any of this?
gpuccio,
I’ve read the Dembski and Marks papers about active information and, while that may be their goal, that is not what they have demonstrated. Dembski and Marks misstate the No Free Lunch theorems in their abstract here when they say “any search algorithm performs, on average, as well as random search without replacement unless it takes advantage of problem-specific information about the search target or the search-space structure.” It is more accurate to say that the NFL theorems say that any algorithm performs no better than random search when averaged over all search spaces.
For a particular search space, some algorithms will perform worse than a random search and some will peform better — quite possibly much better.
For now I’ll leave aside my discomfort with modeling evolutionary mechanisms as a search algorithm, except to note that we must keep in mind that it is only a model. With that caveat, it is clear that the real world can be modeled as one search space. It is not surprising that some algorithms are better at traversing that search space than others, and are much better than a random search. It is further unsurprising that those are the algorithms, or mechanisms, we observe driving evolution in the real world. If they didn’t work, we would either observe other mechanisms or we wouldn’t be here to observe anything.
There is no need to posit a guiding intelligence, and the NFL theorems upon which Dembski’s and Marks’ work is based do not suggest otherwise.
Now that does leave open the question of, if we stick with the search space model, why do some algorithms work better than others in the model based on the real world? The answer is that the mechanisms that work do so by modifying populations of organisms in such a way as to allow them to take better advantage of the environment. Again, if those mechanisms didn’t work, we’d observe different ones.
You might be able to model this process in information theoretic terms, which is what Dembski and Marks seem to be heading towards, but all that would show is that known evolutinary mechanisms are able to create information about the environment in populations of organisms. We know that by observation.
I’ll touch on your issues with Tierra in a subsequent reply.
“but all that would show is that known evolutionary mechanisms are able to create information about the environment in populations of organisms”
___________________________________
Evolution doesn’t create information. It learns. As a crude metaphor you might say the genome stores information about what works. It does so by trying everything and preserving some things.
The math is fairly inexorable. Large populations are better able to adapt to change than small populations, because parallelism enables them to try more variations per generation.
How would a designer enable this phenomenon except by inventing evolution? I’m still waiting for the designer’s source of information.
gpuccio,
I think is a very important point.
How does the designer know what his ultimate design should look like for an unknown future environment?
Where is the designer’s spec in this case?
Does he have the power to see the future?
Petrushka,
I agree. I was presenting my point in the context of modeling evolutionary mechanisms using information theory. It’s not a good approach, in my opinion, but Dembski and Marks seem to be trying to pursue it.
How does the designer know what his ultimate design should look like for an unknown future environment?
________________________________
How does a designer know how a change in the genotype will affect the phenotype, without trying it, or without having a perfect simulator, or without having an infinite lookup table?
If there are rules for predicting emergent properties that are simpler than the infinite lookup table, wouldn’t finding such rules, or at lest a subset of those rules, be something ID should produce before asserting that design is even possible?
That would be a prerequisite for design, even before you set out to make something that would survive in a complex ecosystem.
gpuccio,
Your goal is valid, but your suggested approach is only one of many ways to achieve it. If you look at the Tierra code or the code for any publicly available GA engine, you will see that it is factored into several discrete components:
a) A simulation of the environment
b) An implementation of evolutionary mechanisms such as mutation and crossover
c) A means of translating a bit string or equivalent to some behavior within the environment
d) A loop to run the simulation
There are also typically tools for monitoring and reporting. In some cases there may be a way to specify explicit fitness functions.
As long as the components are separate with clearly defined interfaces it is straightforward to inspect the code and demonstrate that the system does not use anything other than the evolutionary mechanisms being simulated.
Some GAs use explicit fitness functions. Others, like Tierra, do not. I’m not sure what you mean by “select itself”, but in Tierra the only measurement of success is successfully using limited resources of memory and clock cycles.
Another GA system that might interest you is 3DVCE, based on Karl Sims’ work. It demonstrates the evolution of digital organisms that are capable of moving in a simulated environment that includes Newtonian physics.
Tierra definitely keeps the environment simulator decoupled from the replication mechanism as specified in your goals. It also does not use an explicit fitness function.
What metric are you using to measure complexity?
If you can’t apply your measurement to the output of Tierra for some reason, I suggest looking at Tom Schneider’s ev. Like Tierra, ev does not have an explicit fitness function. In addition, ev is even more closely aligned with real biological systems since it was written to test Schneider’s PhD thesis. Most importantly for our purposes here, Schneider measures the Shannon information generated by the evolutionary mechanisms he is simulating. I suggest reading both his thesis on the real biological system and the ev paper describing the simulation.
Petrushka:
“Evolution doesn’t require prior understanding of the effects of genomic change in order to work.”
Not so. Any explanatory theory of evolution must explain the genomic changes, because it’s the genome that evolves. The genome is largely the cause of the phenotype. If you don’t know or can’t explain the cause of genomic variation, you haven’t an explanatory theory of evolution.
Petrushka:
“Evolution acquires this information by learning. By trial and success.”
That’s just your unsupported hypothesis. There is just no evidence that any relevant information is “learned” at the genomic level by the darwinian mechanism.
” You may not like this process”
It’s not a question of what i “like”
“and may think it is inadequate to the task”
That’s certainly true.
“but it is the only process on the table”
Not so. Design is on the table, even if you don’t “like” it at all.
Zachriel:
“The first statement indicates we can’t show an evolutionary path for any complex structure. The second excludes macroscopic structures. This is your reason.”
For the nth time: I am not excluding macroscopic structures from science. I am only stating the obvious: that they don’t help us in our explanatory causal models, because the relationship between macroscopic structure and genomic information is not well understood.
“Yet there is ample evidence of incremental and selectable evolution in macroscopic structures, e.g. the mammalian middle ear.”
But we have no idea of what genomic variation is the basis for that. Therefore, we cannot compare model to explain macroscopic variation, we cannot say if a darwinian mechanism or design explain it best, because we simply don’t know the genomic basis. I don’t think this is so difficult to understand, or to accept.
Mathgrrl:
Welcome back!
“Actually, it’s a lot more complex than that. The real world has all the laws of physics and chemistry that are not typically modeled in a simulation. Not only has this resulted in flight evolving multiple times, but also in specialization. Compare the wings and flight style of a condor and a seagull, for example. The complexity of the real world means there are more niches to exploit than in a simulation.”
OK, but the main problem remains: can RV in a replicator create complex information through improvement of the replicator’s replicating fitness?
If that is not true (and we have really not one piece of evidence in that sense), then however many “niches” you may have, they will never be filled by a darwinian mechanism, unless they can be filled exclusively by very simple variations.
I would say flight is certainly not a good example of simple variation.
“That certainly doesn’t mean that we can’t learn from simulations, of course. The most complex aspect of the Tierra environment is actually the other simulated organisms. It’s very interesting that evolutionary mechanisms such as mutation and crossover resulted in virtual parasites that took advantage of those components of the environment.”
Again, about Tierra, I ask: how complex were the variations which conferred that advantage? (I suspend for the moment my other question about Tierra, because probably it is more difficult to answer it).
“My core point is that the more complex the simulated environment, the more ways there are to exploit it. That’s why we would expect to see more complex functionality evolve in a more complex simulation.”
I don’t agree with that. A complex environment contains more constraints. And the ways to exploit it become more complex.
Another point often overlooked is that , the more complex the replicator, the more the constraints. In a complex replicator you have to take into account all the complex structures which already exists, and any new function will have to be strictly integrated to the logic of the already existing functions.
“The environment, whether real or simulated, doesn’t explicitly give any information. It is the stage on which members of populations act out their lives. The genetic make up of the population will change over time as evolutionary mechanisms operate and individuals leave more or fewer offspring relative to each other based on how well they take advantage of the environment. Natural selection is a result of this process, not a process itself.”
That’s certainly true for the real environment, much less for the simulated ones. The only way to be sure that a simulated environment does not give explicit or implicit information about functions to be evolved is that the programmer of the environment must be blind to the replicator experiment which will be performed. I have said that many times, but nobody seems to accept this very obvious methodological principle.
For the rest, I agree with your description of NS. I would just emphasize that the only “evolutionary mechanism” which can operate, before NS takes place, is RV. There is no other (except, obviously, design).
gpuccio,
Evolution is an implicit process. It doesn’t care at all about what any affect of a change is.
Out of a population of 1 billion, some mutations may help survival while some may hinder.
If I was the actual “Process Of Evolution”, I wouldn’t care what the changes were, I would just congratulate the winners and start the next round.
Evolution is like American Idol where audience reaction is the primary issue and talent is a distant second.
Week to week, we don’t know how the audience is going to react, but always it seems, the best advice from the judges is to give the audience want they want.
In evolution, the environment is the audience and we and other life-forms, are the contestants.
But design is NOT on the table because the “process” is not defined well enough to be modeled.
Evolutionary algorithms have been defined well enough to be modeled and we can see that because those simulations exist.
Where is a design algorithm. Show us one that knows the future well enough to “explicitly” specify functions for that future environment.
Petrushka:
“It is more accurate to say that the NFL theorems say that any algorithm performs no better than random search when averaged over all search spaces.
For a particular search space, some algorithms will perform worse than a random search and some will peform better — quite possibly much better.”
OK. I am fine with that. Now you only have to show that our world is a space where the darwinian algorithm of RV and NS works much better than chance.
” It is further unsurprising that those are the algorithms, or mechanisms, we observe driving evolution in the real world.”
Ehm, that’s not a demonstration at all. First of all, it smells suspiciously of an ontologic argument, or of an nth version of the anthropic principle: both bad philosophy, and certainly not science.
And second, I would humbly mention that we have never observed those algorithms “driving evolution in the real world”. That’s exactly the problem.
“There is no need to posit a guiding intelligence, and the NFL theorems upon which Dembski’s and Marks’ work is based do not suggest otherwise.”
Yes, they do, if you consider the extreme complexity of what we observe in real life, and the complete absence of any evidence that our world is special in regard to your algorithm. It is not.
As far as we know, your algorithm works exactly as expected by Dembski and Marks, and even if by sheer luck it worked a little better, that would not explain the extraordinary intelligent results we observe in the living world.
Artificial evolutionary algorithms work a little better than chance, sometimes a lot better, and that is explained, not by mere luck, but by the active information they contain.
“We know that by observation.”
No.
Ehm, I apologize. My last post was an answer to Mathgrrl, not Petrushka.
Petrushka:
“How would a designer enable this phenomenon except by inventing evolution? I’m still waiting for the designer’s source of information.”
As far as I can say, like us, he already knows many things. And, like us, he can certainly learn others.
Toronto:
“How does the designer know what his ultimate design should look like for an unknown future environment?”
I suppose he is intelligent and intuitive. Like us, he makes plans and anticipates things.
“Does he have the power to see the future?”
At least like us. Maybe more.
Petrushka:
“How does a designer know how a change in the genotype will affect the phenotype, without trying it, or without having a perfect simulator, or without having an infinite lookup table?”
Maybe by understanding the laws of biochemistry better than us?
And anyway, he can certainly try. Like us.
“If there are rules for predicting emergent properties that are simpler than the infinite lookup table, wouldn’t finding such rules, or at lest a subset of those rules, be something ID should produce before asserting that design is even possible?”
Protein engineers are working exactly at that. I suppose they know what they are doing.
“That would be a prerequisite for design, even before you set out to make something that would survive in a complex ecosystem.”
As I have said many times, especially to you who keep repeating stubbornly the same questions, a designer can use both top down and bottom up strategies. Like us.
Mathgrrl:
“Your goal is valid”
Then, why has nobody really tried that? A blind trial is a very common tool of research in modern science.
“As long as the components are separate with clearly defined interfaces it is straightforward to inspect the code and demonstrate that the system does not use anything other than the evolutionary mechanisms being simulated.”
That’s true, but I have neither the time nor the expertise to do that. I will wait for some analysis from Dembski and Marks. I cannot, for obvious resons, trust the mainstream evolutionary enthusiasts.
“Tierra definitely keeps the environment simulator decoupled from the replication mechanism as specified in your goals. It also does not use an explicit fitness function.”
It could use an implicit one. One question. Are the replicators in Tierra true replicators? IOWs, are they like viruses in a computer environment? IOWs, is Tierra a computer environment where small programs can replicate and copy themselves, and the environment is completely passive to that process?
And are the functions acquired by the replicators true replicating function? IOWs, do they actrively favour an independent replication of the replicators, without any reward, explicit or implicit, from the environment? (Other than the environment being what it is?)
“What metric are you using to measure complexity?”
dFSCI, obviously.
Regarding ev, well, another topic in my list of priorities.
Toronto:
“In evolution, the environment is the audience and we and other life-forms, are the contestants.”
I have already praised your talent for metaphors. Maybe you could take part in some contest for that: even if talent is not always correctly recognized, you could win just the same. I have seen stranger things happen 🙂
Toronto:
“Evolutionary algorithms have been defined well enough to be modeled and we can see that because those simulations exist.”
Well, that is not even a metaphor. I suppose you are less brilliant at mere wishful thin king.
“Where is a design algorithm. Show us one that knows the future well enough to “explicitly” specify functions for that future environment.”
Protein engineering is a design algorithm. Antibody maturation is a design algorithm.
gpuccio,
Could you please explain the repeated use of the phrase, “like us”?
gpuccio,
Before addressing your other points, I would like to clarify what you mean by this. Are you asserting that the mechanisms of modern evolutionary theory, including mutation, crossover, and variable reproductive success have not been observed to change the allele frequency in populations?
If that isn’t what you mean, could you please clarify?
About 100 comments ago I wrote that I thought this thread was winding down. Clearly I was wrong. I think the performance is getting unacceptably slow because of the number of comments so I have created a new post simply to allow the discussion to continue.
Genuinely definitely very good website article which has got me considering. I never looked at this from your point of view.
I apologise that WP took your comment as SPAM and I only just noticed. Thanks for your comments.
gpuccio,
Could you please explain the repeated use of the phrase, “like us”? 1 brb
To date all evolutionary algorithms have been targeted searches, which makes them ID algorithms as there is’t any target according to the theory of evolution.
Joe G:
“..there isn’t any target according to the theory of evolution.”
That’s true. Please inform kairosfocus, bornagain77 and Bill Dembski that calculating the improbability of a “specific” result does not refute evolution since “..there isn’t any target..”.
Joe, you should probably answer on one of the latest continuations of this thread. I think there are four in total.
Mathgrrl?
MarkF?
Bon jour!
Joe G “To date all evolutionary algorithms have been targeted searches, which makes them ID algorithms as there is’t any target according to the theory of evolution.”
I believe evolution targets survival, Joe.
Hello! I simply want to give you a huge thumbs up for the excellent info
you’ve got here on this post. I’ll be returning to your
site for more soon.
Long blonde haired Amanda Blow masturbates
with a magic wand at Anilos Naughty blonde 48 yr previous
Sherri Donovan in fishnet lingerie posing Tanned blonde MILF
Anna Pleasure from AllOver30 gets bare by the pool Voluptuous Anilos blonde in black stockings has a great time fingering
her juice field Hirsute girl gets toyed by perverted blonde mom Pussy-loving woman laps up a sexy blonde mature and gets eaten up in return Pretty MILF
blonde Malula blowjobs and fucks her neighbor man in her dwelling Fats mature blonde Zara
with shaved pussy makes use of big pink dildo to masturbate pushing it deep into
her cunt Fairly blonde mature cougar peels her sexy lingerie and thigh highs Blonde nurses in latex lingerie and gloves fucking Blonde
lina in a pink costume and blowing a geek F Large Cock Inside
new Sizzling Mothers