# Evo-Info sidebar: Conservation of performance in search

Introduction to Evolutionary Informatics, by Robert J. Marks II, the “Charles Darwin of Intelligent Design”; William A. Dembski, the “Isaac Newton of Information Theory”; and Winston Ewert, the “Charles Ingram of Active Information.” World Scientific, 332 pages.
Classification: Engineering mathematics. Engineering analysis. (TA347)
Subjects: Evolutionary computation. Information technology–Mathematics.

Denyse O’Leary, an advocacy journalist employed by one of the principals of the Center for Evolutionary Informatics, reports that I have essentially retracted the first of my papers on the “no free lunch” theorems for search (1996). What I actually have done in my online copy of the paper, marked “emended and amplified,” is to correct an expository error that Dembski and Marks elevated to “English’s Principle of Conservation of Information” in the first of their publications, “Conservation of Information in Search: Measuring the Cost of Success.” Marks, Dembski, and Ewert have responded, in their new book, by deleting me from the history of “no free lunch.” And the consequence is rather amusing. For now, when explaining conservation of information in terms of no free lunch, they refer over and over to performance.1 It doesn’t take a computer scientist, or even a rocket scientist, to see that they are describing conservation of performance, and calling it conservation of information.

The mathematical results of my paper are correct, though poorly argued. In fact, the theorem I provide is more general than the main theorem of Wolpert and Macready, which was published the following year.2 If you’re going to refer to one of the two theorems as the No Free Lunch Theorem, then it really should be mine. Where I go awry is in the exposition of my results. I mistake a lemma as indicating that conservation of performance in search is due ultimately to conservation of information in search.

The root cause of my error is a failure to recognize that the “no free lunch” theorems actually address sampling, not search. There are two main components of a search, one of which generates a sample of possible solutions to a problem, and the other of which outputs the best solution it can find in the sample. The theorems address the choice of a sampling component, assuming that the solution-seeking component is fixed. My lemma indicates that sampling processes are devoid of information. There is no conservation of something that does not exist in the first place. As everyone knows, if only by reading the news, sampling processes are distinguished by their biases. The performance (utility) of a sampling component in generating a sample of possible solutions for use by the solution-seeking component has nothing to do with information.

Evolutionary informatics is founded on the conflation of evolution and search. The main topic of the book is evolutionary search for a solution to a problem. What I hope you will remember always, after reading this post, is:

Only the sampling component of an evolutionary search is evolutionary.

The sampling component simulates an evolutionary process in which the “fitness” of a solution is its goodness. (What biologists mean by fitness is not goodness, but instead the expected number of offspring left by an organism, depending on its heritable traits.) The solution-seeking component bears no relation to biological evolution. My lemma says that the evolutionary sampling process gains no information about the fitnesses of unsampled solutions by processing the fitnesses of sampled solutions (and has no information in the first place). Technically, the sample is statistically independent of the fitnesses.

If you remember now what I hope you will remember always, then you will notice that the opening of Chapter 3, “Design Search in Evolution and the Requirement of Intelligence,” is ever so slightly misleading:

Evolution is often modeled by as a [sic] search process. Mutation, survival of the fittest and repopulation are the components of evolutionary search.

Both of the sentences are false. The first is the opposite of the truth. And the problem is not just with this passage. The authors repeatedly conflate scientific modeling of evolution with engineering of an evolutionary search for a solution to a problem. It is vital that they lead readers to misbelieve that the two are the same, because they develop engineering analysis of evolutionary search only for misapplication to models of evolution. Analysis of how well models work, under the unwarranted assumption that modelers do not model, but instead engineer evolutionary searches to solve problems, is an empty accusation of misconduct. The results of such an analysis are not evidence that the assumption holds. Marks et al. write, in the second paragraph of the preface:

Evolutionary models to date point strongly to the necessity of design. Indeed, all current models of evolution require information from an external designer in order to work. All current evolutionary models simply do not work without tapping into an external information source.

Hopefully you see now that they are referring to an evolutionary search as an evolutionary model. It is an evolutionary search for a solution to a problem that performs well (“works”). What they mean by “external information source” is a fitness function. But I established, 21 years ago, that an evolutionary sampling process does not gain information by processing fitnesses. And what sense does it make, when addressing biological evolution, to regard the probabilistic propensity of a (type of) organism to leave offspring as information coming from an external source? The fact of the matter is that Dembski et al. refer to performance as information, and to everything that causes an evolutionary search to perform well as a source of information.3 It is performance that is conserved.

1 A particularly revealing passage from the opening of Section 5.2 (emphasis added): “Here is an illustration of COI [conservation of information] using a waterbed metaphor. Because water is incompressible, if you push down on a framed waterbed at one point, it will bulge somewhere else. Consider Fig. 5.1 which is similar to a figure in Schaffer’s seminal paper [Figure 1 in “A Conservation Law for Generalization Performance“]. Each of the six images in the figure corresponds to a specific search algorithm across a space of problems. The square marked 1 is flat, illustrating an algorithm that performs exactly the same on all problems. In 2, every place the waterbed is pushed in is labeled with a ‘o’ and every place it bulges with a ‘+’. A bulge means that the algorithm performs better than average on a set of problems. An indentation indicates the algorithm does [performs] worse. For every place there is a bulge in 2, COI dictates there must be a corresponding indentation so that, on average, the algorithm here illustrated performs, on average, like the algorithm illustrated in 1. […]. Design expertise or other sources of knowledge are needed to choose a better than average [performing] search on a bulge and away from indentations. Squares 4, 5 and 6 illustrate violations of the law of information conservation. In 4, the algorithm performs better than that in 1 for a number of problems without doing worse anywhere else. A waterbed cannot bulge at a number of points without being indented somewhere else. Likewise, square 5 illustrates many indentations without any bulges. Conservation of information requires a balance between better and worse performing algorithms.”

2 Wolpert and Macready first disseminated their theorem in a 1995 technical report, “No Free Lunch Theorems for Search.” I’d already sketched a proof of the theorem in 1994, challenging Aspi Havewala in his thesis defense.

3 From Section 5.5, “Sources of Information in Evolutionary Search” (emphasis added): “Sources of information embedded in any evolutionary search are mined for active information. […] Evolutionary search mines information rather poorly. The sources of information in the fundamental Darwinian evolutionary model include (1) a large population of agents, (2) beneficial mutation, (3) survival of the fittest and (4) initialization.”

## 102 thoughts on “Evo-Info sidebar: Conservation of performance in search”

1. As someone who advocates a minimalist approach to math, I have a theoretical question.

To me, Behe’s IC systems are somewhat like passwords. Whether or not my characterization of Behe is accurate, let’s just consider an evolutionary search for a password.

Assume the password exists in a space of 26 lower case letters and the a priori probability of any given password is the same as any other (which is not the case for practical passwords since they tend to cluster on recognizable phrases in our culture).

Under these conditions, is it fair to say, any arbitrary algorithm out of the space of possible algorithms is on average just as likely to perform like any other assuming the algorithm isn’t stupid enough to repeat passwords it has already tried?

It would seem a metric for the performance of one algorithm over another for a specific password is some sort of specified information for that particular algorithm and password, but in general I would expect there is no algorithm that will perform better on average for all passwords than any other algorithm except maybe for it’s ability to not repeat attempts already made and to keep track and index what attempts have already been made.

Thanks in advance.

PS
I’ve already gone on record as saying Bill Dembski, Robert Marks, Winston Ewert are my friends, but I tend to favor Behe’s biochemical approach to ID vs. information theory-based arguments.

I don’t teach ID with information theory stuff in general. I’m somewhat negative on it’s use for ID.

2. For the purpose of refuting evolutionary biology it would be sufficient to show that evolutionary mechanisms could not achieve adequate performance (say, adequate fitness). Recasting performance as “information” seems unnecessary.

However, I think that having it be “information” is needed if one wants to connect it to a “Logos Gospel” so that it is “the Word”. (More properly The Bit. This would be rather too easy to parody: “Hear O Israel, the Lord is God, the Lord is 1. The Lord is not 0. So take that, atheists!”)

There is also the long-term goal of providing a quantitative account of the accumulation of adaptive information, and connect this with the flow of energy and the increase of entropy. I would also like to do that, but I don’t think that Dembski, Marks and Ewert’s “Active Information” achieves that.

3. Salvador,

There are parts of your comment that interest me, as a computer scientist. But I have to start by emphasizing that I see no relation to evolutionary biology.

stcordova: let’s just consider an evolutionary search for a password.

Assume the password exists in a space of 26 lower case letters and the a priori probability of any given password is the same as any other (which is not the case for practical passwords since they tend to cluster on recognizable phrases in our culture).

Seemingly small details are actually important here. I don’t mean to criticize you personally by bringing them to the fore.

You can represent prior knowledge with a uniform distribution on the space of passwords only if you know an upper bound on the length of passwords (making the space finite, rather than countably infinite). Now, you can argue for an upper bound, exploiting what you know of the physical world. But I wouldn’t attach “a priori” to an upper bound obtained that way. To the contrary, I would say that there is, a priori, no upper bound on the length of passwords.

If you judge it physically impossible for a computer to process passwords of length greater than then you also suspect that passwords are in reality limited to lengths much less than The upshot is that a uniform distribution on the set of all passwords of length no greater than does not represent your belief about passwords.

To put it simply, you believe that it is better to test shorter passwords before testing longer passwords. (Are you familiar with Levin search, also called universal search?)

stcordova: Under these conditions, is it fair to say, any arbitrary algorithm out of the space of possible algorithms is on average just as likely to perform like any other assuming the algorithm isn’t stupid enough to repeat passwords it has already tried?

You’ve just switched from subjective probability to objective probability. I believe that you are merely using language casually. But what you’re doing is nonetheless a source of huge misunderstanding.

The “conditions” are beliefs — right?

The password is fixed, and the performance of an algorithm is a matter of fact. The algorithms in fact differ hugely in their performance.

The average you’re referring to is your subjective expectation of the performance of an algorithm. That is, taking a weighted average of performance over all passwords, the weights being the probabilities that you assign to the passwords, the average performance does not depend on which of the algorithms you choose.

The performance you expect, subjectively, does not depend on the algorithm. The performance you obtain, objectively, depends hugely on the algorithm.

To address the rest of your comment, I need to know what an algorithm does, and how you measure its performance.

4. Salvador,

What does a search for passwords have to do with any model of evolution? You are apparently describing a fitness surface on which one “genotype” has a fitness of 1, while all others have fitnesses of 0. Given such a surface, then as you say no algorithm that doesn’t repeat guesses has an advantage over any other (well, other than speed). But so what?

5. The password search thing leaves me thinking that decades of trying to explain evolution have scattered their seeds on rocks.

Which is why I no longer read VJ. He’s smart and writes well, but this looks hopeless.

6. The “password” or “needle in a haystack” fitness surface has the property that a change of one base in the fit genome (say changing A to C at position 1,769,234 on chromosome 12) is expected to be as bad for you as changing every base in the genome simultaneously.

Is that how real genomes work? Really?

7. There can’t be any such thing as an evolutionary search of a password fitness landscape. Nothing is evolving in the sense of moving incrementally closer to, or further away, from the correct sequence, guided by selection. It’s just random sampling with no feedback. You either hit the one true sequence by blind dumb luck, or nothing happens.

This has zero relation to real biology. It might be a good analogy for Michael Behe’s concept of irreducible complexity, but then there are no actual examples of IC structures in this sense, from real biology.

8. <quibble>
There can be an evolutionary “search” of a needle-in-a-jaystack (or password) landscape if there is not just one individual but a population of them doing the searching, and then having the frequencies of genotypes in the population change as a result.
</quibble>

9. Only if members of the population can survive and reproduce indefinitely having a fitness of zero.

But your point brings up neutral drift. Evolution does occur when the change in fitness is zero or close to zero. But that assumes there is no goal or target, so the word search is misleading.

10. John Harshman: What does a search for passwords have to do with any model of evolution? You are apparently describing a fitness surface on which one “genotype” has a fitness of 1, while all others have fitnesses of 0. Given such a surface, then as you say no algorithm that doesn’t repeat guesses has an advantage over any other (well, other than speed). But so what?

Rumraket: There can’t be any such thing as an evolutionary search of a password fitness landscape. Nothing is evolving in the sense of moving incrementally closer to, or further away, from the correct sequence, guided by selection. It’s just random sampling with no feedback. You either hit the one true sequence by blind dumb luck, or nothing happens.

This has zero relation to real biology. It might be a good analogy for Michael Behe’s concept of irreducible complexity, but then there are no actual examples of IC structures in this sense, from real biology.

While we wait for Salvador to revivify a dead thread (hit-and-run is not his style), I’ll tell you, in terms of his comment, what Marks, Dembski, and Ewert (MDE) are saying.

MDE believe that Darwinians write programs in an attempt to prove that “Darwinian evolution” really does work. In other words, they don’t have a clue (or pretend not to have a clue) as to the objectives of modelers. [Perhaps it is that they are sure that “Darwinian evolution” does not work, and are attributing to their adversaries the negation of their belief.] When they apply their analysis, they say that the modeler first specifies that the search will output the correct password, and then defines an evolutionary process — including the fitness landscape — on the space of passwords. The modeler rigs the system to produce an evolutionary trajectory from some initial population to a population containing at least one instance of the password. So MDE say that the modeler uses prior knowledge of the password to define a fitness surface that assists the search in finding the password.

[Remarks for grownup scientists:] Obviously, if you want to analyze an agent’s informed choice (design) of a search, then you had better include the agent in your analysis. MDE do nothing of the sort. It’s a glaring error, and there’s a simple explanation of it. They set out, eleven or so years ago, to infer design in nature (avoiding, for legal reasons, a commitment even to a design-er). So they couldn’t very well posit a designer, could they? They have instead been saying that the “active information” of an evolutionary process is objectively measurable. But they actually measure performance, and call it information. (Hey, if you take the negative logarithm of the probability of obtaining the password, the result is information — right? Wrong.)

11. petrushka: Only if members of the population can survive and reproduce indefinitely having a fitness of zero.

Sounds like the human race. What’s the point of reproducing if you can survive indefinitely? Don’t answer that. You weren’t being serious, so neither am I.

12. Mung: I’m with Sal. The less math the better. 🙂

Well, you’re not with Sal, because he’s gone missing. I’m glad you’re still here at times, because I’m in the habit of thinking how you will respond to what I write.

The next post will rely heavily on pictures. It’s hard to bring myself to do that, because the obvious response is that I chose something atypical to show, as a Darwinian apologist or… whatever. In all sincerity, I’m more interested in discussing it with you than with anyone else (which is not to say that I’m uninterested in discussing it with others).

13. Tom English: Hey, if you take the negative logarithm of the probability of obtaining the password, the result is information — right? Wrong.

14. Well, you’re not with Sal, because he’s gone missing.

I’ve decided to move on for the most part from TSZ as I don’t share the views and values of the moderators and admins. I thank them for hosting my discussions while I was in exile from UD, and have gone on to other private venues. So I want to leave this forum for the most part on cordial terms…

I will show up however when I think a discussion is interesting as is the one Tom raised. As I said, despite the fact Bill Dembski, Robert Marks, and Winston Ewert are my friends and comrades, over the years I’ve grown increasingly negative on the value of all these informational entities that have been put on the table. I talk to pre-med biology and other science students who are sympathetic to ID, and I can’t recall a single one that was really into CSI. Behe’s works resonate more with them. Gunter Bechly was influenced by Behe’s works, not so much Dembski’s if at all. So, I’m not trying to pick a fight here with Tom.

What I was trying to do is try to conceptualize the search problem. Searching for passwords is exploration of a search space that doesn’t give feedback like say a travelling salesman problem or other things amenable to Genetic algorithms.

So, I felt the supposed special information in an algorithm that somehow made it seem somehow optimized to find a particular password faster than another algorithm might be possibly quantified, but really all that is a measure of the algorithm’s dumb luck relative to the password, not actual knowledge of the search space.

By way of extrapolation from password searches to something like travelling salesman, it would seem information metrics are either a specialized knowledge of the search space (i.e. we know in advance this is a travelling salesman problem) or we have an algorithm with some dumb luck over a small sample size of searches.

I wasn’t really trying to make a point or argue. I was trying learn and understand. I read Tom’s papers, and the math was difficult and I don’t think I understood 10% of what was written. It’s not in my specialty (and I’ve studied Shannon’s theorems, general relativity, quantum mechanics). So if I have trouble understanding Dembski’s math and the counter arguments in Tom’s paper, I don’t know how very many people would care enough to devote the time to learn the fine points of what is being argued.

Imho, most of this is moot even within the ID community. Bill Dembski has moved on, and so has most of the ID community from CSI. I certainly don’t use it when I argue ID or creation.

I participated in this discussion because it seemed to me, there might be a way to frame the issue is simpler mathematical terms, or there might be special cases that can be framed simply like password searches. Not that I think anything except relatively trivial theorems might result, but it was still an intellectual curiosity.

Thanks to Tom for responding. I wasn’t attempting to make any argument. I was just brainstorming and also trying to understand.

15. What does a search for passwords have to do with any model of evolution? You are apparently describing a fitness surface on which one “genotype” has a fitness of 1, while all others have fitnesses of 0. Given such a surface, then as you say no algorithm that doesn’t repeat guesses has an advantage over any other (well, other than speed). But so what?

As far as your question about passwords, Dawkins Weasel is like a search for a password where “METHINK IT IS LIKE A WEASEL” is like a password except his system gives positive feedback to help the search along, whereas if we’re dealing with trying to crack a password in the cybersecurity world, the situation would be different. So is natural selection more like Dawkins weasel where there is positive feedback or more like searches for passwords in the man-made world of cybersecurity? I and my ID comrades think real evolution performs even worse than the search for passwords (not formally, but that is our intuition anyway).

The S-coefficient of a present trait in a present population, imho, has little or nothing to do with the S-coefficient of precursors to the emergence of that trait before the trait emerged. For example, if a eukaryote fails to make a critical protein for its spliceosome, it probably will die. So there is strong selection pressure for preserving that trait (the critical spliceosome protein) in the present. That says nothing of the selection pressure to evolve that protein before it existed. It would seem to me, it’s like searching for a password — hit or miss — to create that protein in a context that it is actually useful. In fact, as I said, I think real world natural selection performs worse than a password search because it simply gives up trying. That was sort of what might be extrapolated from Behe’s paper on Natural Selection for loss of function. You know, “Behe’s first rule of adaptive evolution.”

I had a conversation with Bob Marks the one time I met him where I talked about the situation where selection selects against complexity, like selecting against a possible password. Half-formed traits may be dysfunctional. To paraphrase Gould, “what good is half a wing.” So I actually think natural selection prevents the evolution of complexity by trapping things in local fitness peaks permanently. But then Bob and I weren’t able to finish our conversation, we ran out of time together…

16. “Only the sampling component of an evolutionary search is evolutionary.”

What does “evolutionary” mean in this context? That it (viz. the sampling) evolves? And what do you mean by sampling? The sampling process or the sample obtained? The sample obtained after each round can easily be said to evolve compared to the sample of an earlier round, but the sampling process (which is really a scientific procedure) cannot evolve.

The way I conceive of evolution, it’s the thing called “natural selection” that does the equivalent of sampling (process) when evolution is understood as an analogy to statistics. Natural selection doesn’t evolve, does it? Or are you saying that with each next ever-fitter generation also natural selection evolves, becomes ever fitter?

It’s the key thesis of your post, so let’s make it very sure we understand it right.

17. stcordova: As far as your question about passwords, Dawkins Weasel is like a search for a password where “METHINK IT IS LIKE A WEASEL” is like a password except his system gives positive feedback to help the search along, whereas if we’re dealing with trying to crack a password in the cybersecurity world, the situation would be different. So is natural selection more like Dawkins weasel where there is positive feedback or more like searches for passwords in the man-made world of cybersecurity? I and my ID comrades think real evolution performs even worse than the search for passwords (not formally, but that is our intuition anyway).

I’m not sure why anyone would take Dawkins as the person who is accurately representing evolution. Dawkins would like people to think of himself as arch-Darwinist, but in reality he is a straightforward Paleyist. Dawkins doesn’t deny the Designer and his/her creative activity, just that the Designer is blind. Dawkins is very much on the ID-ist side of the debate.

18. stcordova:
As far as your question about passwords, Dawkins Weasel is like a search for a password where “METHINK IT IS LIKE A WEASEL” is like a password except his system gives positive feedback to help the search along, whereas if we’re dealing with trying to crack a password in the cybersecurity world, the situation would be different. So is natural selection more like Dawkins weasel where there is positive feedback or more like searches for passwords in the man-made world of cybersecurity?

Since the environment clearly provides feedback in the form of both fecundity and death, the answer is clear unless you have an irrational bias that causes you to ignore the evidence. You’ve been involved in these discussions long enough to know this.

I and my ID comrades think real evolution performs even worse than the search for passwords (not formally, but that is our intuition anyway).

When your intuition is refuted by objective, empirical evidence, which one do you go with? No need to respond, we know the answer.

19. stcordova: So is natural selection more like Dawkins weasel where there is positive feedback or more like searches for passwords in the man-made world of cybersecurity? I and my ID comrades think real evolution performs even worse than the search for passwords (not formally, but that is our intuition anyway).

One consequence of a password-like fitness surface, as I have said earlier in this thread, is that once we have got the password, all mutations of a single letter are just as bad, fitness-wise, as mutating all the letters simultaneously. I asked, in that comment, whether this is what we see in real organisms. Sal seems to have missed that comment.

If I have a single mutation, do I go all the way back to the primordial ooze? Is it no worse than having all sites in my genome mutated?

I think that the answer is obvious — a single mutation is a lot less serious. That points to a fitness surface that does provide feedback. And the reason for that, as I have said here and at Panda’s Thumb many times, is that the weakness of long-range interactions in physics lead to mutations at different sites of the genome not interacting infinitely strongly.

So no, our genome is not like a password.

20. Sal is adopting an odd view of evolution, in which (to use his example) a spliceosome sits around, useless, for millions of years before the last necessary protein appears (poof) in its final form to take its proper place. Of course, this scenario is one that only creationists ever mention. It’s a classic strawman. No wonder he’s confused.

21. Dr. Felsenstein,

Thank you for your comment. With regards to genomics and individual base mutations, I agree, and thank you for pointing out my mistake in my choice of words.

The password analogy should be applied to higher level discrete conceptual entities like protein domains in a coherent system that uses those domains rather than individual nucleobases.

For example there are a bare minimum number of parts needed to create something as complex as the histone readers, writers, erasers and chromatin remodellers required to implement eukaryotic chromatin.

As another example, transitioning from either a eukaryotic from prokaryotic transcription system (or some common ancestor of both) requires re-arranging complex parts much like taking different alphabetic letters from one password and using them to find another password (except in the case of eukaryotes, supposedly more letters need to be added).

The example I’ll be teaching my creationist students is this one regarding the initiation complexes as can be found in Lehninger’s biochem chapters on DNA transcription. Depicted below is the eukaryotic initiation complex. [but here is a link if any one wants to zoom in
http://theskepticalzone.com/wp/wp-content/uploads/2017/05/eukaryote_protein_synthesis_initiation111.png%5D

Which can be contrasted to the prokaryotic initiation complex.
http://theskepticalzone.com/wp/wp-content/uploads/2017/05/prokayote_protein_synthesis_initiation1.png

The deep conservation of both architectures in their respective life forms suggests the transition from one to the other is prevented, rather than facilitated, by natural selection.

I call that the Nelson Paraodox after Paul Nelson. I don’t know the original author of the Nelson’s paradox (but I don’t think it’s Nelson), but it points out that deep conservation of an architecture also is evidence the transition to that architecture is not common place (compared say to evolving a convergent base in antibiotic resistance).

22. Erik:
[Quoting OP:] “Only the sampling component of an evolutionary search is evolutionary.”

[…]

It’s the key thesis of your post, so let’s make it very sure we understand it right.

Thesis: While Marks, Dembski, and Ewert would have you believe, and I once would have had you believe, that information is conserved in search, it is actually performance that is conserved.

The sentence you quote is a prompt to recall more than is stated in the prompt. I should not have highlighted it. I believe that the questions you ask about it are answered clearly enough in the preceding paragraph and the following sentences. It appears that you do not see them because you think that I’m writing about biology. Rereading the post, I see that I should have made it more clear that I was addressing technology. Marks, Dembski, and Ewert address “computer search” through the end of Chapter 5, “Conservation of Information in Computer Search,” and then apply their analysis of search in Chapter 6, “Analysis of Some Biologically Motivated Models.” I’m trying to give a simple, easy-to-remember explanation of why an evolutionary search is not an “evolutionary model” (MDE’s term, promoting conflation of technology and models).

The following is clipped from something else I’ve written. Perhaps it will help.

_________________

Search is a method for solving a problem. A search generates a sample of possible solutions to the problem, and then outputs the best solution it can find in the sample. A search is referred to as evolutionary when the sampling is done by simulation of an evolutionary process. Thus an evolutionary search is not itself a model of evolution. Design of an evolutionary search is engineering, not science.

23. Tom English: Search is a method for solving a problem. A search generates a sample of possible solutions to the problem, and then outputs the best solution it can find in the sample. A search is referred to as evolutionary when the sampling is done by simulation of an evolutionary process. Thus an evolutionary search is not itself a model of evolution. Design of an evolutionary search is engineering, not science.

Damn Tom,. Are you looking to start another weasel war?

😉

24. stcordova: The deep conservation of both architectures in their respective life forms suggests the transition from one to the other is prevented, rather than facilitated, by natural selection.

Let’s call that the Axe fallacy, supposing that we must explain a transition from one modern state to another modern state, rather than that from some common ancestral state to two modern states.

25. Let’s call that the Axe fallacy, supposing that we must explain a transition from one modern state to another modern state, rather than that from some common ancestral state to two modern states.

Fair enough, but the problem is then one is invoking conjectural entities not rooted in direct observation. The conjectural entities would be at least believable if they could even be constructed in principle, and that’s not been proven that it can be done. One can invoke conjectural entities, but then at what point does this cease to be empirical science? If one is going to invoke transitions that may well be indistinguishable from miracles, one may as well invoke miracles.

One doesn’t need 322 pages of evolutionary informatics to point all this out.

But I should point out, if the peptide synthesis initiation takes place in some ancestral form, is there any reason to think it’s part aren’t comparable well matched and coordinated as in extant eukaryotic and prokaryotic forms? Given that the timing, location and precision of events is involved, it doesn’t seem plausible to suggest the system is very pliable to change — much like a password!

There may be multiple login/passwords to gain access to a system, but they are generally not very pliable to change in one part without corresponding changes in other parts if not the whole system.

I just don’t find an evolutionary scenario probable. One can still assert common descent, but that doesn’t mean the event was within ordinary mathematical, chemical and physical expectation, far from it, imho. Such a transition would seem anything but natural. It is evolution by un-natural selection.

26. stcordova: Fair enough, but the problem is then one is invoking conjectural entities not rooted in direct observation. The conjectural entities would be at least believable if they could even be constructed in principle, and that’s not been proven that it can be done.

Look up the method called Ancestral Sequence Reconstruction. You’ll be suprised by how many ancient DNA and protein sequences have been reconstructed, synthesized and actually tested in the lab.

It’s not conjecture, history happened. In fact, it raises key questions for creationism, particularly of the young earth type.

If the inferred evolutionary transitions didn’t actually happen, why can ancestral states be resurrected in the lab and tested functional, often with broader or different ancient functions? Someone like you would have to, well what the hell can your answer even be here? It seems to me it is YOU will be forced to invent that it is due to some happenstance statistical miracle that it just so happens that with some large protein dataset, inferred to be ancestral states are functional.

Seriously, this, probably more than anything else, should cause you to seriously doubt your young earth creationism.

27. The Thornton lab did ASR on the ATP synthase and found how at least parts of it became more complex over time:
https://www.ncbi.nlm.nih.gov/pubmed/22230956
Evolution of increased complexity in a molecular machine.
Finnigan GC1, Hanson-Smith V, Stevens TH, Thornton JW.

Abstract
Many cellular processes are carried out by molecular ‘machines’-assemblies of multiple differentiated proteins that physically interact to execute biological functions. Despite much speculation, strong evidence of the mechanisms by which these assemblies evolved is lacking. Here we use ancestral gene resurrection and manipulative genetic experiments to determine how the complexity of an essential molecular machine–the hexameric transmembrane ring of the eukaryotic V-ATPase proton pump–increased hundreds of millions of years ago. We show that the ring of Fungi, which is composed of three paralogous proteins, evolved from a more ancient two-paralogue complex because of a gene duplication that was followed by loss in each daughter copy of specific interfaces by which it interacts with other ring proteins. These losses were complementary, so both copies became obligate components with restricted spatial roles in the complex. Reintroducing a single historical mutation from each paralogue lineage into the resurrected ancestral proteins is sufficient to recapitulate their asymmetric degeneration and trigger the requirement for the more elaborate three-component ring. Our experiments show that increased complexity in an essential molecular machine evolved because of simple, high-probability evolutionary processes, without the apparent evolution of novel functions. They point to a plausible mechanism for the evolution of complexity in other multi-paralogue protein complexes.

That’s at least parts of a multi-component molecular machine’s evolution, resurrected and tested in the laboratory.

The limitation here isn’t that the entire machine couldn’t have evolved. The limitation to this kind of work is lack of data. Either the sequences diverge so far over time, that the ancestral state can’t be estimated (effectively, the evidence is erased), or gene-loss in different lineages means it just isn’t possible to obtain the data sets that would allow phylogeneticists to infer the ancestor sequences. Or perhaps we just haven’t sampled enough of the diversity out there.

What is remarkable is that, where the data does exist, ancestral states can be reliably inferred, biosynthetically resurrected, and tested for function. And invariably the results conform to evolutionary processes.

There are other such articles out there. Here’s another one: https://www.ncbi.nlm.nih.gov/pubmed/26971075
How to Build a Complex, Functional Propeller Protein, From Parts.
Clark PL1.

Abstract
By combining ancestral sequence reconstruction and in vitro evolution, Smock et al. identified single motifs that assemble into a functional five-bladed β-propeller, and a likely route for conversion into the more complex, extant single chain fusion. Interestingly, although sequence diversification destabilized five-motif fusions, it also destabilized aggregation-prone intermediates, increasing the level of functional protein in vivo.

Or how about this one, resurrecting a billion year old ancestral state and finding it not only functional, but functional in the way expected from the knowledge geologists have assembled from the precambrian global environment?

Biochemical characterization of predicted Precambrian RuBisCO
Patrick M. Shih, Alessandro Occhialini, Jeffrey C. Cameron, P John Andralojc, Martin A. J. Parry, and Cheryl A. Kerfeld

Abstract
The antiquity and global abundance of the enzyme, RuBisCO, attests to the crucial and longstanding role it has played in the biogeochemical cycles of Earth over billions of years. The counterproductive oxygenase activity of RuBisCO has persisted over billions of years of evolution, despite its competition with the carboxylase activity necessary for carbon fixation, yet hypotheses regarding the selective pressures governing RuBisCO evolution have been limited to speculation. Here we report the resurrection and biochemical characterization of ancestral RuBisCOs, dating back to over one billion years ago (Gyr ago). Our findings provide an ancient point of reference revealing divergent evolutionary paths taken by eukaryotic homologues towards improved specificity for CO2, versus the evolutionary emphasis on increased rates of carboxylation observed in bacterial homologues. Consistent with these distinctions, in vivo analysis reveals the propensity of ancestral RuBisCO to be encapsulated into modern-day carboxysomes, bacterial organelles central to the cyanobacterial CO2 concentrating mechanism.

How do you explain these experimental findings from a YEC perspective? How many absurd, ad-hoc hypotheses do you need to erect, for every such study out there, to save your creationism from what is unassailably evidence against it?

28. The Thornton lab did ASR on the ATP synthase and found how at least parts of it became more complex over time:
https://www.ncbi.nlm.nih.gov/pubmed/22230956/
Evolution of increased complexity in a molecular machine.
Finnigan GC, Hanson-Smith V, Stevens TH, Thornton JW.

Abstract
Many cellular processes are carried out by molecular ‘machines’-assemblies of multiple differentiated proteins that physically interact to execute biological functions. Despite much speculation, strong evidence of the mechanisms by which these assemblies evolved is lacking. Here we use ancestral gene resurrection and manipulative genetic experiments to determine how the complexity of an essential molecular machine–the hexameric transmembrane ring of the eukaryotic V-ATPase proton pump–increased hundreds of millions of years ago. We show that the ring of Fungi, which is composed of three paralogous proteins, evolved from a more ancient two-paralogue complex because of a gene duplication that was followed by loss in each daughter copy of specific interfaces by which it interacts with other ring proteins. These losses were complementary, so both copies became obligate components with restricted spatial roles in the complex. Reintroducing a single historical mutation from each paralogue lineage into the resurrected ancestral proteins is sufficient to recapitulate their asymmetric degeneration and trigger the requirement for the more elaborate three-component ring. Our experiments show that increased complexity in an essential molecular machine evolved because of simple, high-probability evolutionary processes, without the apparent evolution of novel functions. They point to a plausible mechanism for the evolution of complexity in other multi-paralogue protein complexes.

That’s at least parts of a multi-component molecular machine’s evolution, resurrected and tested in the laboratory.

The limitation here isn’t that the entire machine couldn’t have evolved. The limitation to this kind of work is lack of data. Either the sequences diverge so far over time, that the ancestral state can’t be estimated (effectively, the evidence is erased), or gene-loss in different lineages means it just isn’t possible to obtain the data sets that would allow phylogeneticists to infer the ancestor sequences. Or perhaps we just haven’t sampled enough of the diversity out there.

What is remarkable is that, where the data does exist, ancestral states can be reliably inferred, biosynthetically resurrected, and tested for function. And invariably the results conform to evolutionary processes.

There are other such articles out there. Here’s another one: http://www.ncbi.nlm.nih.gov/pubmed/26971075/
How to Build a Complex, Functional Propeller Protein, From Parts.
Clark PL.

Abstract
By combining ancestral sequence reconstruction and in vitro evolution, Smock et al. identified single motifs that assemble into a functional five-bladed β-propeller, and a likely route for conversion into the more complex, extant single chain fusion. Interestingly, although sequence diversification destabilized five-motif fusions, it also destabilized aggregation-prone intermediates, increasing the level of functional protein in vivo.

Or how about this one, resurrecting a billion year old ancestral state and finding it not only functional, but functional in the way expected from the knowledge geologists have assembled from the precambrian global environment?
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4735906/
Biochemical characterization of predicted Precambrian RuBisCO
Patrick M. Shih, Alessandro Occhialini, Jeffrey C. Cameron, P John Andralojc, Martin A. J. Parry, and Cheryl A. Kerfeld

Abstract
The antiquity and global abundance of the enzyme, RuBisCO, attests to the crucial and longstanding role it has played in the biogeochemical cycles of Earth over billions of years. The counterproductive oxygenase activity of RuBisCO has persisted over billions of years of evolution, despite its competition with the carboxylase activity necessary for carbon fixation, yet hypotheses regarding the selective pressures governing RuBisCO evolution have been limited to speculation. Here we report the resurrection and biochemical characterization of ancestral RuBisCOs, dating back to over one billion years ago (Gyr ago). Our findings provide an ancient point of reference revealing divergent evolutionary paths taken by eukaryotic homologues towards improved specificity for CO2, versus the evolutionary emphasis on increased rates of carboxylation observed in bacterial homologues. Consistent with these distinctions, in vivo analysis reveals the propensity of ancestral RuBisCO to be encapsulated into modern-day carboxysomes, bacterial organelles central to the cyanobacterial CO2 concentrating mechanism.

How do you explain these experimental findings from a YEC perspective? How many absurd, ad-hoc hypotheses do you need to erect, for every such study out there, to save your creationism from what is unassailably evidence against it?

29. Rumraket: Seriously, this, probably more than anything else, should cause you to seriously doubt your young earth creationism.

An honest YEC responds, “God’s Word, above all else, should cause you to seriously doubt your Darwinism.”

30. Mung: Damn Tom,. Are you looking to start another weasel war?

A war to end all wars, of course. I’ll be going nuclear with a Christian apologist’s extension of the arch atheist’s model. In the meantime, a “sword of Damocles, hanging by the slenderest of threads,” etc., etc.

31. stcordova: Imho, most of this is moot even within the ID community. Bill Dembski has moved on, and so has most of the ID community from CSI. I certainly don’t use it when I argue ID or creation.

I have to note that I’ve been addressing active information, not complex specified information. The latest form of CSI, algorithmic specified complexity, is inversely related to active information.

32. Now of course ancestral sequence reconstruction (or ancestral anything reconstruction) is difficult if there are only two taxa with two states between them, as Sal is alleging for the protein initiation complex. But is he sure? Are there nothing more than prokaryotes and eukaryotes, each with only one condition? No variation within eukaryotes or eubacteria? Most especially, no variation in the various flavors of Archaea? I don’t know whether there is, but it would certainly aid ancestor reconstruction if so.

33. stcordova: Imho, most of this is moot even within the ID community. Bill Dembski has moved on, and so has most of the ID community from CSI. I certainly don’t use it when I argue ID or creation.

Someone should tell the posters and commenters at Uncommon Descent. Both Barry Arrington and Denyse O’Leary seem to think that the observation of specified complexity is an unanswerable argument against natural evolutionary processes having produced the adaptation. Jonathan McLatchie, in his “1-minute apologist” video, spends 2 minutes on it, and looks very pleased with himself. And Arrington and O’Leary like to invoke that video.

All are using SC as if it were something one can conclude is present, even without yet knowing whether natural evolutionary processes could produce it. They are in effect assuming that there is some proof that those processes couldn’t produce it. That was William Dembski’s original stance in No Free Lunch. He changed his tune in 2005 in his paper Specification: The Pattern that Signifies Intelligence”. There he gets it to signify intelligence by defining it so that a pattern is only Specified Complexity if it can be shown to be extremely improbable that it arise by natural evolutionary forces. I kid you not — that’s how we know it signifies intelligence, because we have defined it as such.

Someone should straighten out Arrington, O’Leary, and McLatchie. They are still not among “most of the ID community”.

34. Tom English: An honest YEC responds, “God’s Word, above all else, should cause you to seriously doubt your Darwinism.”

I think that was the 14th commandment, but I can never be sure, because it’s part of “God’s Word” that isn’t actually put into writing.

😉

35. Tom English: A war to end all wars, of course.

There has to be at least an ICBM full of regulars here at TSZ who disagree with you but are just too cowardly to say so.

36. Specified Complexity is like a swiss army knife! Someone should have warned Crick about the can of worms he was opening.

37. Mung: Specified Complexity is like a swiss army knife! Someone should have warned Crick about the can of worms he was opening.

I think you mean Orgel. But Dembski actually formalized what Dawkins said about complicated objects that seem designed to us: “statistically improbable in a direction specified not with hindsight alone” (quoting from faulty memory — see Chapter 1 of The Blind Watchmaker). About five years ago, he started giving Dawkins credit, but without being too, um, specific.

38. Joe Felsenstein: There he gets it to signify intelligence by defining it so that a pattern is only Specified Complexity if it can be shown to be extremely improbable that it arise by natural evolutionary forces. I kid you not — that’s how we know it signifies intelligence, because we have defined it as such.

Sounds more like the version in No Free Lunch to me. There’s both probabilistic complexity and descriptive complexity, IIRC, in the 2005 paper. It’s increasingly hard to keep everything straight, as many times as the story has changed (without explanation).

39. Yes, it is confusing. I was thinking only of the definition of CSI that uses scales such as “viability”. The whole algorithmic specified complexity thing never made much sense to me, because

(1) You have high ASC if you can be described *simply*?
(2) So a party balloon has higher ASC than a hummingbird? Huh?

Furthermore they have sometimes said flatly that ASC is the only way to talk about CSI. And sometimes they have said that it is one of the ways Specification can be “cashed out”. It makes the head spin.

.

40. John Harshman: Now of course ancestral sequence reconstruction (or ancestral anything reconstruction) is difficult if there are only two taxa with two states between them, as Sal is alleging for the protein initiation complex.

Ahh yes, I forget how the inference is made in practice, even though I had you explain it to me at some point. It’s too easy to forget stuff if you don’t use it.

41. Mung,

Specified Complexity is like a swiss army knife! Someone should have warned Crick about the can of worms he was opening.

Real scientists can never control what messes amateurs turn their work into. Just look at carbon dating for instance.

42. Tom English: I’m trying to give a simple, easy-to-remember explanation of why an evolutionary search is not an “evolutionary model”…

Well, thanks, I guess. This leaves us still in the dark about what evolution is and what it can be legitimately likened to.

So you are saying that all those people are operating with a faulty notion of evolution. This makes this place pretty much like UD where whenever one challenges ID theory a standard tack is to reply that one does not understand ID theory. I can grant that ID theorists weasel about evolution, I have seen them do it myself, but the weasel program, supposed to simulate evolution, in reality non-different from a password cracking tool, is by Dawkins. Did he get it wrong also? And the idea that certain errors of perception and visual illusions suggest that consciousness itself is an illusion and this proves that it evolved just so is by Dennett. Yeah, maybe I am stretching the last one, but nobody has shown what conclusion Dennett is really getting at when he makes all these connections.

And did Darwin get it wrong too? Darwin raised the analogy of biology and linguistics. He stated that common descent is “the only possible” genealogical arrangement of both species of life forms and of linguistic variety (Origin of Species, Chapter XIV). In reality no linguist, as long as evidence is concerned, posits a common origin to all languages. The only way to posit a common origin to languages is to believe in (or actually to disregard the true import of) the story of the tower of Babel. So Darwin got it wrong with this analogy and thus evolution seems to be something nobody can ever grasp (which should include yourself) because it’s peerless, incomparable, inscrutable, while at the same time being omnipresent in all of us (sounds like a theology rather than science, right? but this is exactly the impression one gets from Darwin himself).

If Dawkins’, Dennett’s and Darwin’s accounts of evolution hold no water, and Dembski & Co.’s obviously don’t either, then whose account does? Who has the model or simulation or analogy that is close enough and correct enough to be worthy of consideration, not of dismissal?

My point is that unless we have properly identified what evolution is and what “evolutionary” means and entails, your attempts to fend off ID theorists’ attacks on evolution are as futile as ID theorists’ attempts to undermine it. We don’t have a settled topic so we are bound to be talking past each other.

Tom English: A search is referred to as evolutionary when the sampling is done by simulation of an evolutionary process. Thus an evolutionary search is not itself a model of evolution. Design of an evolutionary search is engineering, not science.

If you wanted to be really clear, then you would, in addition to saying what evolutionary search is not, say what it is. What aspects of it are “evolutionary” so that it can be termed “evolutionary search”? Does sampling have the role of natural selection as I assume? Which aspect of the evolutionary search represents genetic drift?

Or are you really saying that evolutionary search is engineering, not science, and therefore evolution stands unscathed? I would come to the opposite conclusion – if there is nothing in principle engineerable about evolution, then it has no leg to stand on. “Naturalistic” theories should be engineerable, mechanistic, otherwise they are not naturalistic.

43. Erik: I can grant that ID theorists weasel about evolution, I have seen them do it myself, but the weasel program, supposed to simulate evolution, in reality non-different from a password cracking tool, is by Dawkins.

The Weasel program is not a password-cracking tool. It is not a semi-accurate simulation of evolution. It is a teaching example intended to show, and showing well, that cumulative selection can increase fitness far, far faster than a purely random-sampling process. It was intended to counter the blatant falsehood spread by creationist debaters, when they repeatedly say that evolution explains adaptation by “random” change. They rely on their audience to be ill-informed about evolution, and thus to reject evolutionary biology because everyone knows that purely random change will lead nowhere. The Weasel simulation is a very dramatic demonstration of the effectiveness of cumulative selection.

Erik: And did Darwin get it wrong too? Darwin raised the analogy of biology and linguistics.

Erik: Darwin got it wrong with this analogy and thus evolution seems to be something nobody can ever grasp (which should include yourself) because it’s peerless, incomparable, inscrutable, while at the same time being omnipresent in all of us (sounds like a theology rather than science, right? but this is exactly the impression one gets from Darwin himself)

Sheesh! The analogy of trees of languages to trees of species is not so bad. If there is no single ancestral language, that does not invalidate the conclusion that Indo-European languages are related. Or does Erik fauk to see that analogy?

Erik: If Dawkins’, Dennett’s and Darwin’s accounts of evolution hold no water, and Dembski & Co.’s obviously don’t either, then whose account does? Who has the model or simulation or analogy that is close enough and correct enough to be worthy of consideration, not of dismissal?

Erik really doesn’t understand the role of models in evolutionary biology, does he? Dawkins and Dennett’s models are simplifications for teaching purposes. Dembski, Marks, and Ewerts’s models are even simpler. The problem is not their models but the dramatic conclusions they draw from them, and the invalid comparisons they make. Evolutionary biologists make models too. None of these models is a complete and accurate description of nature.

So Erik holds all models to an impossibly high standard — but somehow is quite impressed by the model of the Tower of Babel!

44. Erik: And did Darwin get it wrong too? Darwin raised the analogy of biology and linguistics. He stated that common descent is “the only possible” genealogical arrangement of both species of life forms and of linguistic variety (Origin of Species, Chapter XIV). In reality no linguist, as long as evidence is concerned, posits a common origin to all languages

Darwin overplayed an analogy with something outside of his area of expertise? My God, he must have been ignorant scum!

Here’s where he used the analogy better:

“I wish I had time to write you an account of the very absurd lengths to which [Francis?] Bowen & [Louis] Agassiz—each in their own way—are going. The first denying all heredity (all transmission except specific) whatever. The second coming near to deny that we are genetically descended from our great-great-grandfather; & insisting that evidently affiliated languages e.g. Latin Greek Sanscrit owe none of their similarities to a community of origin,—are all autochtonal [i.e., born in the region where they are found]. Agassiz (foolish man) admits that the derivation of languages & that of Species or forms stand on the same foundation, & that he must allow the latter if he allows the former,—which I tell him is perfectly logical.” [To C. Lyell; Feb 2, 1861]

You have amused me much by your account of Agassiz’s denying the community of descent of allied languages, [To Asa Gray; Feb 17, 1862]

Source

They note there some crucial differences, but the analogy regarding the evidence is actually quite good for most languages (English not as good as most), since the evidence of derivation relies primarily upon what is conserved in both biologic evolution and in language evolution.

Darwin made some mistakes that matter a great deal more to science than any mistakes in an analogy. What matters is not the persons, however, but the evidence.

This site uses Akismet to reduce spam. Learn how your comment data is processed.