Evo-Info review: Do not buy the book until…

Posted on by
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.

… the authors establish that their mathematical analysis of search applies to models of evolution.

I have all sorts of fancy stuff to say about the new book by Marks, Dembski, and Ewert. But I wonder whether I should say anything fancy at all. There is a ginormous flaw in evolutionary informatics, quite easy to see when it’s pointed out to you. The authors develop mathematical analysis of apples, and then apply it to oranges. You need not know what apples and oranges are to see that the authors have got some explaining to do. When applying the analysis to an orange, they must identify their assumptions about apples, and show that the assumptions hold also for the orange. Otherwise the results are meaningless.

The authors have proved that there is “conservation of information” in search for a solution to a problem. I have simplified, generalized, and trivialized their results. I have also explained that their measure of “information” is actually a measure of performance. But I see now that the technical points really do not matter. What matters is that the authors have never identified, let alone justified, the assumptions of the math in their studies of evolutionary models.a They have measured “information” in models, and made a big deal of it because “information” is conserved in search for a solution to a problem. What does search for a solution to a problem have to do with modeling of evolution? Search me. In the absence of a demonstration that their “conservation of information” math applies to a model of evolution, their measurement of “information” means nothing. It especially does not mean that the evolutionary process in the model is intelligently designed by the modeler.1

I was going to post an explanation of why the analysis of search does not apply to modeling of evolution. But I realized that it would give the impression that the burden is on me to show that the authors have misapplied the analysis.2 As soon as I raise objections, the “Charles Ingram of active information” will try to turn the issue into what I have said. The issue is what he and his coauthors have never bothered to say, from 2009 to the present. As I indicated above, they must start by stating the assumptions of the math. Then they must establish that the assumptions hold for a particular model that they address. Every one of you recognizes this as a correct description of how mathematical analysis works. I suspect that the authors recognize that they cannot deliver. In the book, they work hard at fostering the misconception that an evolutionary model is essentially the same as an evolutionary search. As I explained in a sidebar to the Evo-Info series, the two are definitely not the same. Most readers will swallow the false conflation, however, and consequently will be incapable of conceiving that analysis of an evolutionary model as search needs justification.

The premise of evolutionary informatics is that evolution requires information. Until the authors demonstrate that the “conservation of information” results for search apply to models of evolution, Introduction to Evolutionary Informatics will be worthless.

1 Joe Felsenstein came up with a striking demonstration that design is not required for “information.” In his GUC Bug model (presented in a post coauthored by me), genotypes are randomly associated with fitnesses. There obviously is no design in the fitness landscape, and yet we measured a substantial quantity of “information” in the model. The “Charles Ingram of active information” twice feigned a response, first ignoring our model entirely, and then silently changing both our model and his measure of active information.

2 Actually, I have already explained why the “conservation of information” math does not apply to models of evolution, including Joe’s GUC Bug. I recently wrote a much shorter and much sweeter explanation, to be posted in my own sweet time.

a ETA: Marks et al. measure the “information” of models developed by others. Basically, they claim to show that evolutionary processes succeed in solving problems only because the modelers supply the processes with information. In Chapter 1, freely available online, they write, “Our work was initially motivated by attempts of others to describe Darwinian evolution by computer simulation or mathematical models. The authors of these papers purport that their work relates to biological evolution. We show repeatedly that the proposed models all require inclusion of significant knowledge about the problem being solved. If a goal of a model is specified in advance, that’s not Darwinian evolution: it’s intelligent design. So ironically, these models of evolution purported to demonstrate Darwinian evolution necessitate an intelligent designer. The programmer’s contribution to success, dubbed active information, is measured in bits.” If you wonder Success at what? then you are on the right track.

588 thoughts on “Evo-Info review: Do not buy the book until…”

  1. Mung,

    Bullshit. it would have been better had the connection been kept from evolutionists because then it would have prevented the abuse of one domain by another that is completely unrelated.

    It is an extraordinary fact that something that does not really work – evolution – should have inspired an algorithmic approach that does.

    John Holland wasn’t an evolutionary biologist.

    No indeed. Bloody plagiarist.

  2. Allan Miller: Here’s an implementation of ‘get_fitness’, in VB so everyone can sneer at the rubbishy language I chose (!):

    We understand your skills are limited. 🙂

    In that implementation, it appears that “fitness” is undefined. Why am I wrong?

    You claimed that fitness is independent of implementation, which is questionable. Based on that particular implementation fitness appears to lack any meaning whatsoever. So maybe you are right.

  3. Apparently Mung is demanding a function that takes a DNA sequence and returns a fitness figure.

    We’ll need to define a CoderRetardedException just in case it craps out

  4. phoodoo: We count the number of offspring to determine the best antenna?

    Made me laugh. Let’s see how Allan manages to avoid the oblivious, like Joe.

  5. Allan Miller: It is an extraordinary fact that something that does not really work – evolution – should have inspired an algorithmic approach that does.

    For some definitions of work, perhaps. You are almost certainly not claiming that GA’s always work, for any problem, are you? No, you would not say that. So what we have here is just another shovel full of bullshit. Are you a bullshit artist Allan?

  6. dazz: Apparently Mung is demanding a function that takes a DNA sequence and returns a fitness figure.

    Why can’t I just ask the DNA sequence what it’s fitness is? Didn’t Joe claim that organisms and genotypes have a fitness? If so, it should be a property of the organism or genotype, and I should be able to just ask the organism or genotype what it’s fitness is.

  7. Allan Miller: In environments where everything has the same fitness, then everything does indeed have the same fitness.

    So the “environment” determines “fitness”? Or is the environment is superfluous, since fitness is determined by number of offspring? Dancing must have some fitness value. #MoveThatGoalpost

  8. Mung: For some definitions of work, perhaps. You are almost certainly not claiming that GA’s always work, for any problem, are you? No, you would not say that. So what we have here is just another shovel full of bullshit. Are you a bullshit artist Allan?

    Is this a drinking game? Something like every time you confuse the model with what’s being modeled, you take a sip? Because you guys being tipsy would explain a lot

  9. Allan Miller: In scenarios where some genotypes have different fitnesses, then indeed, not everything has the same fitness.

    Thanks for clearing that up.

  10. Mung: [fitness] should be a property of the organism or genotype, and I should be able to just ask the organism or genotype what it’s fitness is.

    but you’re asking for a fucking function, for jeebus sake, Mung

  11. dazz: Is this a drinking game?

    Apparently it can be anything you want it to be. If you have a specific argument or contribution you can take an extra drink.

  12. dazz: but you’re asking for a fucking function, for jeebus sake, Mung

    Why didn’t you bitch at Allan for posting a function that had no arguments?

  14. Allan Miller: It does for an individual. Which, as phoodoo has made clear, is what he wants to focus on. For some reason.

    Joe F. claimed that fitness could be measured for a specific organism or genotype. I don’t think it’s fair to blame it all on phoodoo. Joe F. thinks its’ ok to assign fitnesses to specific alleles. Do alleles leave offspring Allan?

  15. phoodoo: So everyone alive who hasn’t given birth to an offspring yet has a fitness of zero?

    You make me want to die. 🙁

  16. Mung: Joe F. claimed that fitness could be measured for a specific organism or genotype. I don’t think it’s fair to blame it all on phoodoo. Joe F. thinks its’ ok to assign fitnesses to specific alleles. Do alleles leave offspring Allan?

    Tom English: Joe Felsenstein: How about fitness of genotypes? There are more than one individual of the same genotype. Consider the number of them that there are before selection, and the number after. The fraction is the estimate of the viability of that genotype.

    Phoodoo will demand to know how the fitnesses got into the program. You might explain why we can learn about evolution from simulations in which the fitnesses are made up. (Ewert, if not phoodoo, very much needs the explanation. And he’s probably reading this.)

    Many people who use the word genotype do not understand that geneticists classify organisms in a population according to (some of) their heritable traits. That is, it hasn’t registered on them that a genotype is in fact a type (class) of organism. An organism belongs to one and only one class (genotype), while a class may contain many organisms. Feel free to say this better than I have. I’m not one to believe that being a computer scientist grants me special insights into biology. I leave such self-mythologizing to people whose “faith” is severely threatened by the theory of evolution.

  17. Tom English: The process of simulating an evolutionary process is no more an evolutionary process than the process of simulating a spinning windmill is a spinning windmill.

    And now if we can just get Joe F. to agree we’ll all be happy and can sit down and have a beer together.

  20. Tom English: In evolutionary computation, we simulate an evolutionary process in order to sample the space of possible solutions to a problem.

    Far too many members here at TSZ disagree with Tom in spite of the fact that he’s probably the one person here at TSZ in a position to know.

    Of course, if some of you evolutionists out there want to adopt the position that evolution is a process which samples the space of possible solutions to a problem, don’t let me stop you.

  21. Mung: And Tom accuses me of a non-sequitur.

    Look again. “There’s your non sequitur,” immediately following a comment by phoodoo, means that you should have attributed the non sequitur to phoodoo instead of to Joe.

    I’m “accusing” you and phoodoo of not understanding the difference between an evolutionary search (for a solution to a given problem) and an evolutionary model (that specifies an evolutionary process along with an event that tends to occur in the process).

    (I hope you won’t repeat the bit about software development as modeling. I built lots of airplane models as a kid, therefore…. By the way, I was working with broad-spectrum specification languages in 1987, as a student of Joe Urban in software engineering.)

  22. Mung: Far too many members here at TSZ disagree with Tom in spite of the fact that he’s probably the one person here at TSZ in a position to know.

    Of course, if some of you evolutionists out there want to adopt the position that evolution is a process which samples the space of possible solutions to a problem, don’t let me stop you.

    chu-chu-chu-chu-chu-chu…pito!

  23. Rumraket: Like Dawkins WEASEL. And other simulations of evolution that do, like BoxCar2D.

    You think Dawkins’ WEASEL program is a simulation of evolution in spite of the fact that he explicitly wrote that “Life is not like that”? Why do you think Dawkins is wrong?

  24. Tom English: Look again. “There’s your non sequitur,” immediately following a comment by phoodoo, means that you should have attributed the non sequitur to phoodoo instead of to Joe.

    ok. I thought that “your non sequitur” had me as the subject. 🙂

  25. Mung: You think Dawkins’ WEASEL program is a simulation of evolution in spite of the fact that he explicitly wrote that “Life is not like that”? Why do you think Dawkins is wrong?

    Wait, how close to the real thing must be the model before we drink?

  26. Tom English: I’m “accusing” you and phoodoo of not understanding the difference between an evolutionary search (for a solution to a given problem) and an evolutionary model (that specifies an evolutionary process along with an event that tends to occur in the process).

    You make a distinction that almost certainly demands a wider audience than phoodoo and I.

    There is something we can rightly call an evolutionary search. An evolutionary search searches for a solution to a problem. That is what I have been saying all along.

    But then there is also something else, which we must carefully distinguish from evolutionary search, which are models of evolution. A model of evolution specifies an evolutionary process along with an event that tends to occur in the process.

    I will honestly admit that I do not see any distinction that makes a difference. But to hear that the Dawkins’ WEASEL program is not an evolutionary model makes me smile.

    But I think you’re wrong. It is a model. A misguided model. But nevertheless, a model.

  28. Mung: An evolutionary search searches for a solution to a problem

    God’s got so many problems… poor guy

  29. Alan Fox:
    Phoodoo

    Please resist the urge with the “have-you-stopped-beating-your-wife” routine. As you are a minority member, I give you some leeway and I request you not to abuse it..

    What the fuck are you talking about Alan? Are you up to your same old bullshit again, trying to jump in and save one of your team, whenever they are doing bad.

    What line did I write that was against the rules? What leeway are you giving me??

    Go away Alan. You apparently have nothing to contribute to this thread and are just spamming.

  30. Rumraket: That’s the same thing. lol

    Right Rumraket. As in there is ONLY one genotype for ONE individual. Every new strand of DNA is a brand new genotype, so no genotype survives from one generation to the next now does it? If I make any changes to the DNA I have a new genotype, thus we can’t count its fitness.

    It is a completed entity. 1 or 0. It exists or it doesn’t.

  31. Mung: LoL! Until you take the actual measurement, how do you know how many offspring they have?

    Hahaha. If you measure them and they have no offspring, then they can’t be measured!

    Fitness is a matter of timing. Its sort of like quantum mechanics.

  32. Allan Miller: The fitness of a genotype in any EA is the number of offspring produced.

    [working through old comments]

    FYI: We rarely go with the biological conception of fitness in evolutionary computation. All we mean by fitness is goodness. That is a big part of why people like me get wrong ideas about biological evolution. (I’ve put a lot of work into un-knowing a lot of what I thought I knew.)

  33. Mung:

    Where’s keiths?

    Busy, but dropping by every now and then to watch you and phoodoo make phools of yourselves.

    God never answers my prayer for worthy opponents. I’m beginning to think he doesn’t exist.

  34. To all, prompted by phoodoo:

    phoodoo: We count the number of offspring to determine the best antenna?

    [working through old comments]

    An evolutionary computation (EA, GA, GP, etc.) uses a simulation of evolution for the purpose of generating a sample of possible solutions to a problem, treating possible solutions as though they were organisms, and the goodnesses of possible solutions as though they were fitnesses. However, we usually proceed under the misconception of Darwinian evolution as “survival of the fittest,” thinking that the “organisms” that survive are the ones that are the best (by some criterion or another).

    I prefer to say that the algorithms of EC are biologically inspired, rather than that they use simulations of evolution. What I’m doing, when I say that they use simulations of evolution, is to concede as much as possible to Marks, Dembski, and Ewert when they conflate evolutionary search and evolutionary model. The very furthest I can go is to say that an evolutionary search uses an evolutionary (simulation) model to generate a sample of possible solutions to a problem.

    We do sometimes use a technique called fitness-proportionate selection (or roulette wheel selection) that does jibe with the biological conception of fitness. Parents are selected randomly (with replacement) from the population, with probability proportional to their fitnesses. The expected number of offspring for an individual in the parental population is proportional to its fitness. The response of a biologist to this is “well, of course.” But it generally doesn’t work very well in practice. In fact, our sundry impressionistic renditions of “evolution” in evolutionary computation indicate that we collectively do not regard evolution as some sort of magical problem solver.

    Back to phoodoo’s question. What we measure, in evolutionary computation, is the goodness of a possible solution to a problem. The measurement of goodness has nothing to do with the measurement of fitness that Joe Felsenstein described above. In evolutionary computation, we treat the possible solution and its goodness (somewhat) analogously to an organism and its fitness. Marks, Dembski, and Ewert have got the analogy backwards.

  35. Mung: Joe F. thinks its’ ok to assign fitnesses to specific alleles. Do alleles leave offspring Allan?

    Care to provide a link to where I said that? I deny saying it.

    In some models, with some patterns of fitness, we can do that, but generally, not.

  36. Allan Miller: OK though, let’s try wind turbines. We have a population of genotypes, and we have some means to convert genotypes into actual wind turbines. Perhaps it’s a series of coordinates in xyz space.

    Obviously, we are looking for something that distinguishes better turbines from worse. We’ve coded the EA for a purpose, not so we can just sit and look at it. We set them up and spin ’em, and measure the efficiency or whatever. The genotypes that don’t do so well according to those criteria, we eliminate. Then we do some reproducin’. Then we repeat. Have we coded ‘better’ and ‘worse’ into the program? No. Yet the genotypes have differing numbers of offspring – fitness. The fitter ones produce more, but we have coded no ‘def_fitness’ routine.

    [responding to old comments]

    The EA uses a simulation of evolution to sample the space of descriptions of turbines, treating descriptions analogously to genotypes of organisms, turbines analogously to phenotypes of organisms, and efficiencies (whatever) of turbines analogously to fitnesses of genotypes mapped to phenotypes.

    A sampling process (evolutionary or otherwise) is statistically independent of the data (not information) associated with the sample. Thus the intuition that the association of genotypes with fitnesses (data) informs the evolutionary process is dead wrong. There is no gain of information in data processing. That is, there is no way to justify a prediction of the efficiencies of as-yet unsampled descriptions on the basis of the efficiencies of the already-sampled descriptions. Philosophers know this as the problem of induction.

    It makes no difference whether the EA gets its efficiency data from an industrial lab or from a simulation module (not that you said it did). The data are not information.

    I have to mention also that biological fitnesses are not data, let alone information. Some types of organism tend to leave more offspring than do others. How is that “informing evolution what to do”? Reproduction is what organisms do. Evolution occurs as a consequence.

  37. Mung: Bullshit. it would have been better had the connection been kept from evolutionists because then it would have prevented the abuse of one domain by another that is completely unrelated.

    John Holland wasn’t an evolutionary biologist.

    Oh, so he was the first to do an “evolutionary algorithm”? He published in 1975 and called his algorithms “genetic algorithms”.

    I recommend for your light reading the volume of reprinted papers by David Fogel, Evolutionary Computation: The Fossil Record”. David put together that book because he was annoyed that people attributed the whole history to Holland. He knew that was not the whole story because his father, Leslie Fogel, had done genetic simulations of evolutionary processes in 1963.

    As made clear in Fogel’s excellent compilation, the first genetic simulations of evolutionary processes were done by Nils Aall Barricelli in 1954. In 1957 an Australian quantitative geneticist, Alex Fraser, simulated multiple-locus traits under artificial selection. By 1973 there were already two books published about how to do genetic simulations of evolutionary processes and of artificial selection.

    I helped Fogel a little when he did his book, by providing him with references from my Bibliography of Theoretical Population Genetics which referenced computers, and by helping him get in touch with Alex Fraser (who Fogel arranged to get a pioneer award from the IEEE interest group on Evolutionary Computation).

    Or Mung could just read the History section of the Wikipedia page on Genetic Algoirthms.

    No Mung, when evolutionary biologists do computer simulations of evolutionary processes that is not an irrelevant of misleading thing to do.

  38. Tom English,

    I give Tom credit for pointing out what I assumed everyone should have already been aware of. Fitness in an EA refers to quality, or at least to a particular quality, that we then search for.

    That differs entirely from the concept of fitness in nature. The concept of fitness in nature works as the exact opposite, we don’t look for that quality, we look at existence, and then consider the quality. In a computer we don’t look for what exists, and then say it is good, we look for what is good, and then say it should exist.

    If anything, a computer EA is the closest we can come to modeling what many IDist believe, usefulness proceeds existence, not existence proceeds usefulness.

    (I didn’t write that perfectly, but I still acknowledge Tom for providing some measure of sanity to the issue).

  41. phoodoo: That differs entirely from the concept of fitness in nature. The concept of fitness in nature works as the exact opposite, we don’t look for that quality, we look at existence, and then consider the quality. In a computer we don’t look for what exists, and then say it is good, we look for what is good, and then say it should exist.

    We can watch genotypes for one or a few generations, and making careful counts, estimate fitnesses. Equivalently, we can try assigning numerical values to the fitnesses, run a model of the evolution of those genotypes, and predict the changes in the genotypic composition of the population. Then we take as the best estimates those that best predict the genotype frequencies that we see.

    So the before/after distinction between EAs/GAs and nature is not quite as large as phoodoo says.

  42. All of the talk about computers in this thread has bugged me. The computer is a tool for automating processes. It’s the process, not the automation, that matters. From the Wikipedia biography of Hans-Paul Schwefel:

    While attending the Hermann Föttinger-Institute for Hydrodynamics (HFI) at TUB, [Schwefel] and Rechenberg began performing experiments upon wings, kinked plates, and other objects related to fluid dynamics. The main objective of those experiments concerned optimizing the shape and/or parameters through mostly small modifications on the real objects, a “technique” they called experimental optimization, in order to reduce the drag, increase the thrust, and so on. Applying classical optimization methods (such as Gauss–Seidel and gradient-based techniques) on such experiments showed that those methods are not well suited to be adopted in experimental optimization, mainly due to noisy measurements and/or multimodality. They realized modifying all the variables at same time via a random manner (e.g., small modifications are more frequent than larger ones). This was the seminal idea to bring to light the first, two membered, evolution strategy, which was initially used on a discrete problem (optimization of a kinked plate in a wind tunnel) and was handled without computers.

  43. Mung,

    Let’s see how Allan manages to avoid the oblivious, like Joe.

    Haha! Was that a typo? Good joke if not. I have plenty of means of avoiding the oblivious.

  44. Mung,

    In that implementation, it appears that “fitness” is undefined. Why am I wrong?

    No. In that implementation, fitness differential is unspecified – but it’s still there, just happening to be zero between any two genotypes. Which is itself a specification, one could say. All genotypes have the same chance. There is no fitness differential, but there is fitness.

  45. Mung: All you’ve done is replace one term with a different term. Define the function measure_reproductive_success. So far we’ve found out that get_fitness is just an alias. It’s like pulling teeth around here.

    Which makes it all the more perplexing that you people are having a problem with it. Just compare the number of offspring of different organisms and then you have their relative reproductive success, aka their fitness.

    This is what is so amazing about this discussion. Why the hell is this such a problem? It’s so simple.

  46. Mung,

    Do alleles leave offspring Allan?

    They leave copies. Pretty interchangeable concepts, really. As an expert in GAs, you should be able to see that the fitness of a digital organism is the number of copies made from that instance. I don’t know why you might think real genes could not be treated the same way.

  47. Mung: LoL! Until you take the actual measurement, how do you know how many offspring they have?

    You obviously don’t. Why is this even a question you ask? Should I call an ambulance?

  48. Mung: How? Define the function get_fitness.

    Pick an organims (A) that had offspring and count how many it had. If it had eight, say, call that a fitness of 1.
    Pick another (B) one and count (say another one had four), now compare the numbers. B had half as many as A, so it had a fitness of 0.5.

    What is the problem?

  49. Mung: Who said it’s an issue?

    Who said it’s circular?

    Lol. I guess we’re just saying words here for no reason.

    Let me ask you, why do you think we’re having this discussion?

  50. Mung: Bullshit. it would have been better had the connection been kept from evolutionists because then it would have prevented the abuse of one domain by another that is completely unrelated.

    Bullshit. The use of computers to simulate evolutionary processes is neither a form of abuse, nor is it unrelated to actual biological evolution.

    See, anyone can just sit and assert shit like that.

Leave a Reply