Intelligent Design

So Much For Random Searches

Spread the love

There’s an article in Discover Magazine about how gamers have been able to solve a problem in HIV research in only three weeks (!) that had remained outside of researcher’s powerful computer tools for years.

This, until now, unsolvable problem gets solved because:

They used a wide range of strategies, they could pick the best places to begin, and they were better at long-term planning. Human intuition trumped mechanical number-crunching.

Oh,my! Teleology raises its ugly head!

But, now, let’s hear it for Intelligent Design. Here’s what intelligent agents were able to do within the search space of possible solutions:

. . . until now, scientists have only been able to discern the structure of the two halves together. They have spent more than ten years trying to solve structure of a single isolated half, without any success.

The Foldit players had no such problems. They came up with several answers, one of which was almost close to perfect. In a few days, Khatib had refined their solution to deduce the protein’s final structure, and he has already spotted features that could make attractive targets for new drugs.

Random search: 10 years + and No Success
Intelligent Agents: 3 weeks and Success.

Is there a lesson to be learned here Darwinist onlookers?

76 Replies to “So Much For Random Searches

  1. 1

    Random searches for potential drugs have been going on for a while. There was a short article a couple of years ago about the significant expenses in time and machines in this area, and the nearly complete lack of success of the effort.

  2. 2

    But evolutionary processes aren’t “random search”! In fact evolutionary processes are much closer to “intelligent design”!

  3. 3
    Petrushka says:

    Oddly enough, chemistry solves the problem in milliseconds without computation.

    So the actual scorecard should read:

    Intelligent searches, ten years and no score
    Gamers score in three weeks
    Natural processes score thousands of times per minute.

  4. 4

    Actually, it’s a really interesting piece, thanks!

    But I think it makes a better pro evo argument than a pro-design argument! It shows that a distributed competitive system with proximal rewards works better than a top-down expert system with long-term goals!

  5. 5
    PaV says:

    Elizabeth:

    This is pure opinion. How in the world is what you say true. Please present some sort of an argument.

  6. 6
    PaV says:

    Are you blind to reality? Here’s the quote:

    . . . they could pick the best places to begin, and they were better at long-term planning. Human intuition trumped mechanical number-crunching.

    How can you simply turn a blind’s eye to this: “THEY COULD PICK THE BEST PLACES TO BEGIN . . .”! What in the world does this have to do with a “distributed competitive system with proximal rewards”? This is human intelligence at work using knowledge of what the final goal should look like: i.e., TELEOLOGY!!

  7. 7

    Yes, it’s opinion, but I’ve presented the argument quite often. It’s not original though! Basically, it’s the argument that the brain works by a system that is sometimes called “neural Darwinism”, with excitatory connections taking the role of positive selection and inhibitory connections the role of negative selection. Maybe we can talk about it on another thread!

    Or on my blog – I’ve been thinking of writing a piece about it. I’ll send you the link 🙂

  8. 8

    Well, read the rest of the article 🙂 It describes a distributed competitive system!

    And what happened was that the scientists, who did have some knowledge of what the final goal should look like did worse than the gamers, who were solving bits of it, for points.

  9. 9
    thud says:

    Surely we are not trying to equate randomness with evolution are we?

  10. 10

    [satire warning] Which raises an interesting possibility: Shifting some government funding from scientists to gamers to get more bang for the research buck. If they can do this for points, think what they might accomplish for real coin of the realm. Or maybe even a Nobel prize?

    Not to mention that learned capabilities are a major factor in virtually all forms of design, as opposed to natural processes that determine things such as the winding course of a river from the mountains to the sea.

  11. 11
    rhampton7 says:

    More than just opinion, it’s a hypothesis be worked into a theory – meaning it’s going to require more intellectual effort to dispute the possibility of natural evolution:

    Life as Evolving Software*
    Gregory Chaitin, September 7, 2011

    In the fi rst paper in this series we proposed modeling biological evolution by studying the evolution of randomly mutating software we call this metabiology. In particular, we proposed considering a single mutating software organism following a random walk in software space of increasing fitness … And we measured the rate of evolutionary progress using the Busy Beaver function BB(N) = the largest integer that can be named by an N-bit program.

    Our two results employing the framework are

    * with random mutations, random point mutations, we will get to fitness BB(N) in time exponential in N (evolution by exhaustive search),

    * whereas by choosing the mutations by hand and applying them in the right order, we will get to tness BB(N) in time linear in N (evolution by intelligent design).

    We were unable to show that cumulative evolution will occur at random; exhaustive search starts from scratch each time.

    This paper advances beyond the previous work on metabiology by proposing a better concept of mutation … Using this new notion of mutation, these much more powerful mutations, enables us to accomplish the following:

    * We are now able to show that random evolution will become cumulative and will reach fitness BB(N) in time that grows roughly as N2, so that random evolution behaves much more like intelligent design than it does like exhaustive search.

    * We also have a version of our model in which we can show that hierarchical structure will evolve, a conspicuous feature of biological organisms that previously was beyond our reach.

  12. 12
    PaV says:

    Elizabeth, you’re missing the forest for the trees.

    Here’s what the guy who made this success possible said:

    “These results indi­cate the potential for integrating video games into the real-world scientific process: the ingenuity of game players is a formidable force that, if properly directed, can be used to solve a wide range of scientific problems.”

    The success didn’t depend on “a distributed competitive system with proximal rewards”. You’re being almost willfully blind here. I don’t know why. (But I have my opinion, 😉 )

  13. 13
    PaV says:

    No, we’re trying to equate intelligent design with evolution.

  14. 14
    Eugene S says:

    It is claimed that randomness is an essential part of evolution. Coupled with natural selection it is claimed to have produced all variety of life we observe. The problem with this explanation is that it is highly implausible on the grand scale. While it certainly explains “the small world” variability (roughly within existing species) which is observable, it cannot realistically explain giant leaps of structural complexity between taxa. These giant leaps correspond to substantial increases of information that cannot be reliably accounted for other than by intelligent agency.

  15. 15
    dmullenix says:

    PaV, read the article: “But until now, scientists have only been able to discern the structure of the two halves together. They have spent more than ten years trying to solve structure of a single isolated half, without any success.”

    Do you really believe those trained molecular biologists were randomly choosing structures and then seeing if they worked? They’ve been using all of their knowledge and expertise (in other words, “…knowledge of what the final goal should look like: i.e., TELEOLOGY!!”) for the last ten years without success. Then the problem was recast so amateurs with good general intelligence and little or no molecular biology knowledge could work on the problem as a game and swarms of those amateurs solved most of the problem in three weeks. At that point an expert stepped in and finalized the solution.

    Or, in short, distributed intelligence beat isolated intelligences.

    F/N The owner of the house we’re renting this week just came in wearing a Baylor sweatshirt. I brought up Dembski and ID. NOT a fan of either! In spades!

  16. 16
    PaV says:

    Elizabeth, this is how the solving took place:

    The controls are intuitive; tutorial levels introduce the game’s mechanics; colourful visuals provide hints; and the interface is explained in simple language. While protein scientists concern themselves with “rotating alpha-helices” and “fixing degrees of freedom”, Foldit players simply ‘tweak’, ‘freeze’, ‘wiggle’ and ‘shake’ their on-screen shapes.

    Human intelligence pervades the whole process. This wasn’t some computer program number-crunching itself through some kind of search alogrithm. Human INTUITION was what solved the problem. You know, “intelligence”. And, of course, intuition is that part of our intelligence that leaps immediately towards some final end. So, it’s teleology.

  17. 17
    dmullenix says:

    Eugene, contra ID, evolution only changes a very small part of the genome at a time – typically only one or two bases. This means that 99.99999+ percent of the new genome is identical to the parental genome.

    This means that instead of doing a random search, evolution searches a very small area very close to a known workable part of the search space.

    To do a Dembski/ID style random search, an organism would have to change every base in its genome randomly all at once every time it reproduced. And everybody agrees with ID here – that would never work.

  18. 18
    PaV says:

    Did you read this part:

    Foldit takes a different approach, using the collective efforts of causal gamers to do the hard work. And its best players can outperform software designed to do the same job.

    And, did you read what you just finished writing?

    . . . distributed intelligence beat isolated intelligences.

    I don’t see any mention of random processes or natural selection. The selecting that took place was that done by intelligent agents. This is, of course, an ID sine qua non.

  19. 19
    Eugene S says:

    I think it is not disputed that we are comparing different patterns of intelligent intervention. The experiment itself was designed as well, no doubt. In either scenario, there is what we call a decision maker (something or somebody that compares different outcomes in view of the final goal, to varying degrees of success) and a non-trivial starting point. It is so to say a higher level of intelligence that drove and directed lower level intelligent input. In any case, we have a known optimisation function. With blind search, nothing like this really holds. So I believe it is an example of careful design.

  20. 20
    Eugene S says:

    Hello,

    I have no problem with this explanation. This is indeed part of what ID asserts. ID claims that to get to those isolated islands of biological meaning it is impossible without deliberate fine-tuning (intelligent intervention). After the system is put in that favourable state in its phase space, provided the system has enough capability to adapt to environmental change, microevolution is possible.

  21. 21

    Elizabeth: Here are a couple of key quotes:

    “Last year, Cooper showed that Foldit’s gamers were better than the Rosetta programme at solving many protein structures. They used a wide range of strategies, they could pick the best places to begin, and they were better at long-term planning. Human intuition trumped mechanical number-crunching.”

    and

    “[Those who didn’t do as well] focused too heavily on tweaking already imperfect solutions that other teams achieved better results by making large-scale changes.”

    The objective lessons from this are clearly: (i) intelligent agents are better than number-crunching brute force approaches (i.e., throwing lots of probabilistic resources at it doesn’t really help much when there is a huge search space), and (ii) “tweaking” imperfect solutions (wow, that is practically a definition of how evolution is supposed to work, progressing from one point to the next) is not as good of an approach as looking at the problem more broadly and making large-scale changes.

    Kudos, though, if you can somehow twist this into “evidence” for the efficacy of evolution.

  22. 22
    PaV says:

    For the quasi-willfully blind:

    “We wanted to see if human intuition could succeed where automated methods had failed,” Firas Khatib of the university’s biochemistry lab said in a press release.

    “The ingenuity of game players is a formidable force that, if properly directed, can be used to solve a wide range of scientific problems.”

    One of Foldit’s creators, Seth Cooper, explained why gamers had succeeded where computers had failed.

    “People have spatial reasoning skills, something computers are not yet good at,” he said.

    Link

  23. 23
    rhampton7 says:

    Which is the reason for my previous post (2.1.2). Gregory Chaitin is beginning to demonstrate that a “fine-tuned” search space can be a completely natural phenomena. If his ideas are credible, then Intelligent Design theory will have to adapt.

  24. 24
    Eugene S says:

    RHampton7,

    Let him try and demonstrate this. I have enough reason to doubt it can be demonstrated, unfortunately. The no free lunch principle is ubiquitous: in mechanics, in physics, in biology: to get something beneficial you have to pay, and pay significantly more than the amount of gain you want to get. That;s the way this fallen world operates, unfortunately…

  25. 25
    ciphertext says:

    Based upon the comments I’ve read so far, the experiment’s “results” appear to be more like a Rorsach test than anything else. Sure, the 3-D protein formations are real enough, but whether the experiment demonstrates concepts of ID or TOE appears to be as much a product of the individual’s preconceptions as anything else.

  26. 26

    If “microevolution” is possible, and of course it is, and has been observed, then a great many ID arguments simply fail, including the “No Free Lunch” argument.

    Microevolution happens precisely because evolutionary processes work better than random search when the search is a population adapting to its environment.

  27. 27

    I’m still not getting the message you are apparently drawing from this.

  28. 28

    Well, that’s good, because evolution is a fairly intelligent system.

  29. 29

    So what was not distributed, competitive or proximally rewarding about the Foldit operation?

  30. 30

    There was certainly selection – the stable configurations were rewarded.

    But certainly the variance wasn’t random. However, neither was it far-sighted, unlike the scientists, who, apparently, failed at the same task.

  31. 31

    Yup.

    I don’t think it especially supports either, although I do think it’s very interesting – it’s a kind of exploitation of “hive mind”.

  32. 32
    kairosfocus says:

    There is a consistent pattern of conflating hill-climbing within an island of function with the challenge of getting to the island of function.

    Until small changes can get small gains or losses in an existing function, moving about a little bit is of no advantage.

    And the ID challenge is to get to the islands of function. As has been repeatedly pointed out.

  33. 33
    kairosfocus says:

    The message is quite obvious, on the superiority of intuitive, intelligently nudged search over blind search.

  34. 34
    Joseph says:

    Elizabeth Liddle:

    If “microevolution” is possible, and of course it is, and has been observed, then a great many ID arguments simply fail, including the “No Free Lunch” argument.

    Unfortunately for you just baldly declaring it doesn’t make it so. And also unfortunately for you there isn’t any part of microevolution that refutes any part of ID.

  35. 35
    Eugene S says:

    Well, apart from preconceptions we have common sense, haven’t we?

  36. 36

    How so, Joseph? How would microevolution occur if it were true that evolutionary processes were no better than blind search?

  37. 37
    rhampton7 says:

    kairosfocus, RE: (4.2.1.1.4)

    Until small changes can get small gains or losses in an existing function, moving about a little bit is of no advantage.

    This is too vague a statement to be helpful. More importantly, it seems to contradict this statement:

    ID doesn’t claim that neo-Darwinism mechanisms cannot cause small-scale changes in organisms that might represent small changes in specified complexity. But the observation that neo-Darwinism can do some things does not imply that neo-Darwinism can do all things.

    Finally, the paper that I linked to shows the math and the Meyer-Ritchie LOOP programming that supports his work. Chaitin’s website has more that can be freely downloaded: http://www.cs.umaine.edu/~chaitin/

  38. 38

    But the scientists weren’t engaging in “blind search”. Quite the opposite. They seemed to be hampered by their preconceptions.

    And even if they were (which they weren’t) – what is the point of the OP wrt to evolution? Evolution isn’t a blind search.

  39. 39
    Eugene S says:

    Joseph,

    That statement by Dr Liddle is a striking example of confusion between the necessary and sufficient. I did not expect to see this much of a logic flaw.

    Dr Liddle,

    Trouble is macroevolution has not been observed. “No free lunch” has been proposed as a theorem irrespective of ID, to my knowledge. Also, could you be more precise to indicate which arguments of ID fail as a result of microevolution being observable. Thanks.

  40. 40

    There is a consistent pattern of conflating hill-climbing within an island of function with the challenge of getting to the island of function.

    Until small changes can get small gains or losses in an existing function, moving about a little bit is of no advantage.

    And the ID challenge is to get to the islands of function. As has been repeatedly pointed out.

    And I am seeing a consistent pattern of failing to demonstrate that the functional domains populated by living things are, in fact “islands”. The evidence strongly suggests that far from being “islands” they form a connected tree – that islands are not “colonised”, precisely because they can’t be by an incremental process – an intelligent designer, of course could do so.

    It’s because the pattern of living things, over the ages, form a connected tree (Darwin’s “tree of life”) that his theory, which we know works on human time scales, can be extrapolated to biological time, and it’s because we don’t find “islands” that a designer is a less persuasive hypothesis.

    No centaurs, gryphons or crocoducks, and, pace Behe, no Irreducible Complexity, at least none that seems persuasive enough to offset the overwhelming pattern of connectedness.

    Unless you are talking about the “island” of life itself. But then we wouldn’t be talking about Darwinism.

  41. 41

    Where is my “logic flaw” Eugene?

    “No Free Lunch” is a perfectly respectable theorem that says something like evolutionary searches are no more successful than random searches when averaged over all possible fitness landscapes.

    But we know, from our observations of “microevolution” (by which I mean, evolution on an observable time-scale, I’m not sure what you mean) that over the kinds of fitness landscapes that populations adapt to, evolutionary search works far better than random search.

    And nobody in the ID movement seems to deny this. So we immediately have, in front of us, a fitness landscape that can be climbed faster by evolutionary processes than by random search. The “Free Lunch” is provided by the simple fact that a random change to a genotype that already works doesn’t take the phenotype very far in terms of fitness. “NFL” would only apply if a random change to a genotype is no more likely to move the phenotype a small distance than a large one, and that is simply not the case with populations of biological organisms – if it were the case, not even microevolution could occur, and yet it does.

  42. 42

    Greg Chaitin’s Life as Evolving Software video 1:18:48
    Lecture given at UFRGS in Porto Alegre, Rio Grande do Sul, Brazil

    Few people remember Turing’s work on pattern formation in biology (morphogenesis), but Turing’s famous 1936 paper On Computable Numbers exerted an immense influence on the birth of molecular biology indirectly, through the work of John von Neumann on self-reproducing automata, which influenced Sydney Brenner who in turn influenced Francis Crick, the Crick of Watson and Crick, the discoverers of the molecular structure of DNA. Furthermore, von Neumann’s application of Turing’s ideas to biology is beautifully supported by recent work on evo-devo (evolutionary developmental biology). The crucial idea: DNA is multi-billion year old software, but we could not recognize it as such before Turing’s 1936 paper, which according to von Neumann creates the idea of computer hardware and software.

    http://www.youtube.com/watch?v=RlYS_GiAnK8

  43. 43
    dmullenix says:

    Why are you trying to shoehorn evolution into this? There’s no mention of it in the original article and evolutionary methods were *never” used or mentioned. It’s strictly chemistry and molecular biology. A bunch of intelligent amateurs using some neat game-style programming did a better job than a smaller number of intelligent experts using conventional messages. And in the real world, proteins fold due to electrostatic attractions.

    You have evolution on the brain.

  44. 44

    Elizabeth: “evolutionary search works far better than random search.”

    What does this even mean? Are you saying that the event of change is not random, i.e., it is following some law-like process or is directed in some way?

    If by evolution not being random, you mean simply that once the random event changes the organism, the result of that event may be that the organism experiences a fitness advantage, then you haven’t shown anything about evolution being more effective than random search in finding the change in the first place. The question about the random search applies with equal force to a pure blind search, as it does to the front half (i.e., the change part) of what you are calling “evolutionary search.” You are dealing with a blind search in either instance.

    Now, if you what you mean is that getting to point Z directly by a random search is less likely than getting to point Z, through points A-Y, *each of which requires a much smaller search space than Z and each of which confers some kind of fitness advantage*, then sure, we all agree that it might be less daunting than getting to Z directly by random. But that climb up mount improbable must be tempered by several facts, among others: (i) you are still dealing with a random search at each step of the way, (ii) the multiple probabilities of the various random searches at points A-Y have to be taken into account, (iii) points A-Y have to confer some kind of fitness advantage, (iv) the fitness advantage has to be sufficient to overshadow competing changes, or to some other way get fixed in the population.

    That such a fitness landscape actually exists (which Darwin certainly hoped, in terms of his ever “plastic” organisms), is conjecture, not fact. Sure, we can get a finch beak to oscillate around a norm, but there is zero evidence that we have an available climb up mount improbable to the finch itself.

    BTW, I hope you were typing quickly and that your definition of microevolution is not simply “evolution on an observable time-scale”. That would be an example of equivocating two different meanings of the word evolution. Rhetorically convenient, of course, because if we see small scale, even cyclical changes, we can imagine that we have found evidence for more significant changes. But unfortunately it doesn’t hold.

  45. 45

    No, I was not “typing quickly” – what do you mean by “microevolution”?

    It’s not a word with a well-defined technical meaning. However, whenever observable evolution is mentioned, people here say “oh, that’s micro-evolution, not macro-evolution” so presumably people accept that evolution occurs on observable time-scales

    As for evolution not being a random search – no, it isn’t. I think part of the problem here is the “search” metaphor, but it’s used on the NFL theorems so we are sort of stuck with it.

    A “random search” for a solution would be one in which the “searcher” searches the entire search space at random, the previous “pick” having no relationship to the next. An evolutionary search doesn’t do this – each pick is constrained by the previous one.

    This won’t do any better than random search if the solutions are randomly scattered through search space. But if the solutions are clustered in search space, it will.

    And, in biology, they are. Which is why microevolution works.

  46. 46
    Eugene S says:

    Elizabeth,

    If a smaller world effect is observed, it does not necessarily mean it can be extrapolated into the bigger picture, esp. taking into account strong evidence against it. Your reasoning has a flaw of the type: (1) All bullies in this school are boys; (2) Mike is a boy, so he is a bully.

    If you can’t see this, I am afraid I can’t help it any more than by this comment.

  47. 47
    Eugene S says:

    “So we immediately have, in front of us, a fitness landscape that can be climbed faster by evolutionary processes than by random search.”

    Not immediately. The precise picture is that this landscape is not everywhere defined. It is only defined for tiny islands of functionality in the search space. The bigger problem which you either superficially do not notice or unwilling to recognise (unfortunately, having seen your responses, I am afraid I must admit the latter to be the case) is how you drive your system towards those islands where microevolutionary adaptations are made possible. The search space is so enormously vast that within timespan bigger than that of the entire universe it is not possible to plausibly get any biologically feasible solutions without a priori intelligent fine-tuning of vital parameters.

    Imaging a pile of junk, will it ever become a computer without intelligent intervention? But in this case the probabilities are mucho-mucho higher than for life to emerge spontaneously…

  48. 48
    Eugene S says:

    In addition to parameter tuning. I am working in the area of combinatorial optimisation. And I know from my own experience that the problem of parameter tuning is intractable. It quickly becomes really hard to tune the behaviour of a complex enough system as the number of parameters grows. And quite often there may be no feasible solutions. In other words, quite often even intelligent intervention on our part can’t help it, to say nothing of a blind search, whether or not it is accompanied by the natural selection filter: the search space is too large.

    To me, the fact alone that biological systems do allow solutions with incredibly high numbers of parameters is itself a miracle.

  49. 49

    No, I know it doesn’t, Eugene. However, I was not making that extrapolation.

    What I was doing was saying that if Dembski’s application of NFL theorem is correct, then microevolution shouldn’t be possible either, yet clearly it is.

    So why should we think that Dembski’s application of the theorem to biological evolution is correct?

    Explain why the NFL theorems don’t apply to microevolution but apparently do apply to macroevolution.

  50. 50

    Please justify your claim that “this landscape…is only defined for tiny islands of functionality in the search space”.

    The reason you think I “do not notice” (I’ll draw a veil over your other alternative) the problem of “how you drive your system towards those islands” is that I’m not convinced that any islands are actually reached!

    They don’t have to be (though they may exist – an extra pair of hands would be nice, but we are are extremely unlikely to get them – wings would be more probable). All that matters is that there is useful connected land in some direction (and the directions are extremely numerous).

    I think the diffulty is that people forget that while mutations randomly occur anywhere in genotype space (to be precise, because “random” can mean too many things – can occur anywhere in the genome with fairly equal probability), they don’t occur randomly in phenotype space. A mutation is far more likely to result in a phenotype that is very similar to the parental phenotype than one that is very different. In other words, while the “genotype space” explored may be “rugged”, that’s not what matters, because natural selection occurs at the level of the phenotype, and in phenotype space, the fitness landscape is very smooth.

  51. 51

    To me, the fact alone that biological systems do allow solutions with incredibly high numbers of parameters is itself a miracle.

    Not a miracle at all, it’s the very high numbers of parameters that makes solutions virtually inevitable – it means the fitness landscape is very high dimensioned, so there is a high probability that somewhere in adjacent space, along some dimension, there will be a “bridge” to somewhere useful.

    Again, that’s because what is being explored is phenotype space not genotype space, and in phenotype space, variants are adjacent to each other.

  52. 52
    bornagain77 says:

    Programming of Life – Biological Computers – Ch. 6
    http://www.youtube.com/user/Pr.....Rooe6ehrPs

  53. 53
    Eugene S says:

    Elizabeth,

    It is a common mistake to assume that we can extrapolate a function over singularities.

    You have been pointed many times towards the literature where these claims are substantially justified. Refer to papers by Douglas Axe for example. In a nutshell, consider the number of all thinkable amino acid residue sequences (for a protein domain it is 20^150, and it is a huge number indeed, I can assure you) and compare it with those sequences that bear any biological meaning. Axe finds out that the ratio of functionality to no functionality is 1 in every 10^74 sequences (for any functionality whatever), if I remember rightly. To compare, the number of atoms in the observable universe is 10^80.

    E.g. Behe points out that a retina tissue needs a complex set of protein-protein intercations to enable the necessary curvature alone, to say nothing of any other parameter. This is all horrendously complex.

    These difficulties only escalate as we go from protein to cell, to tissue, to body because the number of parameters grows. Combinatorially, it is not at all clear to me that this problem can be easily dismissed as one becoming close to trivial insolubility as the number of parameters grows. What if I say that it might be closer to the phase transition than to the trivially insoluble edge? In any case, it must be rigorously demonstrated that parameter tuning moves towards trivial insolubility. I don’t see what makes you think we can easily assume this to be the case. We do not have the right for an a priori dismissal like that. All possibilities must be slavishly tried if we are talking about a blind search without intelligence.

  54. 54

    It is a common mistake to assume that we can extrapolate a function over singularities.

    Please explain what you mean by this. I am not aware that I was doing so.

    And you seem to be missing my point. Evolution doesn’t have to explore “all thinkable amino acid residue sequence” and, of course doesn’t. It explores (this metaphorical language is hampering us I think) only adjacent sequences to ones that work.

    And ones that work tend to be next door, phenotypically, to ones that also work.

    Approaching the problem with combinatorics is misleading you, just as it has misled Dembski. You are missing the point about the adjacency of solutions in phenotype space, also my point that having lots of parameters makes brings more parts of search space adjacent to a population at any given time, not fewer, thereby increasing the probability that a viable move will be found.

    In a high dimensioned fitness landscape, viable parts of it are more connected, not less.

  55. 55
    Eugene S says:

    8.2.1.2.1 by Elizabeth Liddle

    Elizabeth,

    I am sorry to have supposed deliberate dismissal of opponents’ claims on your part. But certain things that have been said a number of times appear to be going unnoticed by you for some reason. Maybe it’s me not noticing your answers. Sorry.

    What precludes macroevolution from happenning in practice is the amount of information that needs to be generated/injected into the system to make a leap feasible from one island of functionality to another. I do not agree with your claim that the search space is continuous. It is not. There is simply no time in the history of the universe to make macroevolution plausible. Experimental findings I referred to prove my point.

  56. 56
    Eugene S says:

    8.2.1.2.3 by Elizabeth

    Elizabeth,

    “It explores (this metaphorical language is hampering us I think) only adjacent sequences to ones that work.”

    This is microevolution. I agree with this.

    The bigger question is how to get to those sequences that work, to start with. You are not addressing that question. Once you provide substantial evidence it can happen spontaneously, you will have dealt with it adequately. So far, you haven’t, unfortunately.

  57. 57
    Eugene S says:

    BTW, as regards solution rarity vs. solution isolation of functional islands in the search space, Douglas Axe deals with it nicely on the webpage of his Biological Institute. He empirically proves that islands of functionality do exist on the protein level.

  58. 58

    I am sorry to have supposed deliberate dismissal of opponents’ claims on your part. But certain things that have been said a number of times appear to be going unnoticed by you for some reason. Maybe it’s me not noticing your answers. Sorry.

    No problem. I appreciate your graciousness.

    What precludes macroevolution from happenning in practice is the amount of information that needs to be generated/injected into the system to make a leap feasible from one island of functionality to another. I do not agree with your claim that the search space is continuous.

    OK, but in that case, that is the argument we need to be having, not an argument about NFL. NFL is irrelevant to the question of the nature of the search space, as it is about averages over all possible search spaces. And we agree, in the case of micro evolution, that we do not have to consider all possible search spaces.

    It is not. There is simply no time in the history of the universe to make macroevolution plausible. Experimental findings I referred to prove my point.

    There certainly would not be enough “time in the history of the universe” for living things to be found by random search. That is why the theory of evolution posits that the search space is continuous – or, in Darwin’s metaphor, that life forms a huge family tree.

    This is what I find a bit frustrating about these discussions – people keep saying how improbable evolution is because it can’t reach islands of functionality, as though “evolutionists” are proposing that it can. We aren’t. We are saying that what it reaches, and has reached, aren’t islands.

    Darwin’s theory of evolution does assume one starting island, that of self-replication with heritable variance in reproductive success, but posits that from that point onwards everything is connected.

    So challenging the hypothesis of connectedness is a legitimate way of challenging evolution IMO. What are straw men are the ideas that “evolutionists” think that evolutionary processes are a “random search”, i.e. obviously wrong because “random search” doesn’t have time to work (which is true, but evolution is not a random search); or that evolutionary processes don’t work any better than random search in phenotype space (they do, or we wouldn’t observe “micro” evolution).

  59. 59

    Well, we simply don’t (yet) know how the whole thing got started. Darwin didn’t attempt to explain that – his theory only applies once you have self-replication with heritable variance in reproductive success.

    So that remains a problem, but it’s an OOL problem, not an evolutionary problem, although the two do overlap to some extent, as to solve the OOL problem you have to push back the earliest, Darwinian-capable entity.

    However, once you have that minimal, Darwinian-capable entity, there is no reason that I can see to postulate any other disconnection, and genetics (developmental genetics in particular) seems to me to confirm the incremental relatedness of living things rather than disconfirm it.

    But clearly you disagree 🙂

  60. 60
  61. 61
    Eugene S says:

    Elizabeth,

    It depends greatly how you define relatedness. If in terms of common ancestry, I strongly disagree for obvious information theoretic reasons (with which you just agreed above). In this terms, there is no connected tree of life. If relatedness in understood functionally then yes, I agree. The latter type of relatedness, in my opinion, speaks out loud for commonality of purpose and of design ideas behind living things.

    As far as NFL is concerned, the way I interprete it is that overall for optimisation problems, no winner algorithm exists performance-wise. This appears to be in disagreement with what you are saying about genotype/phenotype search spaces. The way I construe NFL is that it does not really matter on the whole as far as search performance is concerned, which algorithm to choose. To get some benefit you always have to pay: no magic tricks are possible, including spontaneous complexity leaps between taxa.

    That there are complexity gaps between different taxa I think is clear enough. The world is decrete, not continuous as was assumed at the times of Darwin. The Darwinian rendering of life as something plastic does not hold any more on the macro scale and, yes, there are islands of functionality.

  62. 62
  63. 63
    Eugene S says:

    no.4 links to the home page of the BioInstitute. It is better to use this link instead.

  64. 64

    It depends greatly how you define relatedness. If in terms of common ancestry, I strongly disagree for obvious information theoretic reasons (with which you just agreed above).

    Well, no, I didn’t agree! What I mean by “relatedness” is simply differing by no more than a few sequences.

    If you can show that organisms that once lived, or currently live, must differ by more than a few sequences in the genotype from its immediate ancestors, then yes, we have an island.

    But the challenge for those claiming islands is to show that those islands are real. I don’t see that “information theoretic” principles tell us that those islands must exist. I’m not sure what you are saying here, in fact.

    As far as NFL is concerned, the way I interprete it is that overall for optimisation problems, no winner algorithm exists performance-wise. This appears to be in disagreement with what you are saying about genotype/phenotype search spaces. The way I construe NFL is that it does not really matter on the whole as far as search performance is concerned, which algorithm to choose. To get some benefit you always have to pay: no magic tricks are possible, including spontaneous complexity leaps

    between taxa.

    I’m not sure what you are saying here, either. What do you mean by “no winner algorithm exists, performance-wise?”

    The NFL theorem tells us that if a variant drawn at random from the total “variant space” is has the same probability of fitness as a variant drawn from a nearby space, then evolutionary algorithms won’t “perform” (i.e. find the high-fitness variants) any better than random search, which is sort of obvious – if a fitness landscape consists of a sea of spikes, evolutionary search won’t help. However, that’s not the form the fitness functions “searched” by phenotypes actually takes. Most mutations make very little difference to the reproductive success of the phenotype, which means that the “reproductively successful” parts of phenotype space are clustered together, and the fitness landscape, far from being spiky, is very smooth. Not only that, but it’s high dimensional, so there is a high probability that some reasonably high fitness part of the landscape will be nearby along some direction.

    So, sure, evolution would have to “pay” for “complexity leaps between taxa”. So, if there are indeed “complexity leaps between taxa”, evolutionary algorithms won’t cross the gap. But that, as I understand it, is somewhat different from Dembski’s NFL argument which seems to say that evolutionary processes don’t work any better than random search even when the fitness landscape is smooth. In fact, we know they do (hence “micro” evolution).

    So what about these “leaps between taxa”? You say:

    That there are complexity gaps between different taxa I think is clear enough. The world is decrete, not continuous as was assumed at the times of Darwin. The Darwinian rendering of life as something plastic does not hold any more on the macro scale and, yes, there are islands of functionality.

    And this is exactly the point at issue. Good to have it stated clearly (because, as I said, the NFL and the UPB are irrelevant to the evolutionary argument which posits a fitness landscape in which neither apply).

    And this is why, IMO, Behe has the only decent ID argument – that there are indeed “islands” (Meyer, of course, argues that the earliest Darwinian-capable entity is also such an “island”).

    But in my view it is by no means “clear enough”! Indeed, it is the contention of all evolutionary biology that it there is are no “complexity gaps” and it was Darwin’s thesis – indeed, it was his explanandum – seeing evidence, in taxonomy, of a tree, he sought to explain the tree, or, at least, saw that if there was a tree, self-replication with heritable variance in reproductive success was a viable explanation for biological variety and adaptation.

    So it all hangs on these “gaps” 🙂

    And the trouble with any argument based on the existence of “gaps” is that gaps can be filled, and indeed, frequently are. It is, literally, an argument from ignorance “we don’t know how X got from A to B, therefore design”.

    But it’s important, IMO, to understand that that is what the argument is. “Evolutionists” aren’t saying: “Oh, sure, random search could get across that gap” – they are saying “there are no gaps”.

    We all agree that random search can’t cross the gaps!

  65. 65
  66. 66
    uoflcard says:

    It absolutely is, Elizabeth. That is blatantly obvious. What is not blantantly obvious is how neo-Darwinian mechanisms cause complex, seemingly far-sighted evolutionary steps. Seeing that you deny intelligence to exist at all, even in humans (as your flabbergasting denial of it in this case demonstrates), it is not that surprising that you cite neo-Darwinian mechanisms as the cause. You are simplying inferring the best known cause. We are doing the same thing, except we believe intelligence to exist (as evidenced by almost every second of our existence)

  67. 67
    uoflcard says:

    I am not an expert, so correct me if I’m wrong, but I thought everyone had recognized that the “continuous search space” idea is simply false. Observed in the fossil record as well as present-day evolutionary leaps. Wasn’t that the whole point behind James Shapiro’s new book? That the evolution we witness is nothing like the “smooth tree” Darwin fantasized about and legions have worshipped since?

  68. 68
    Starbuck says:

    Here’s an interesting discussion. My guess is that functional proteins didn’t originally evolve from randomization of large reading frames but from smaller fragments.

    Cordes, M.H.J., Burton, R.E., Walsh, N.P., McKnight, C.J. & Sauer, R.T. (2000) An Evolutionary Bridge to a New Protein Fold. Nature Structural Biology 7, 1129-1132.

    Davidson, A.R., Lumb, K.J. & Sauer, R.T. (1995) Cooperatively Folded Proteins in Random Sequence Libraries. Nature Structural Biology 2, 856-863.

    Davidson, A.R. and Sauer, R.T. (1994) Folded Proteins Occur Frequently in Libraries of Random Amino Acid Sequences. Proc. Natl. Acad. Sci. USA 91, 2146-2150.

  69. 69

    I don’t “deny intelligence to exist at all”. I don’t know where on earth you got that from.

    And I agree that “neo-Darwinian mechanisms” (if I’ve understood correctly what you mean by that) cannot “far-sighted evolutionary steps”. I don’t think they do. I think they produce very near-sighted steps, and it’s the near-sightedness of the steps that is one of the most important pieces of evidence for a near-sighted intelligent system rather than a far-sighted intelligent system.

    We, on the other hand, embody far-sighted intelligent systems, by virtue of our brains, which allow us to make “forward models” of the consequences of our actions, and select those actions that are most likely to achieve our distal goals.

    I would ask that you don’t attribute to me positions I do not hold, and have never indicated that I hold. Not only do I not “deny intelligence to exist at all”, intentional decision making is actually my research area. I’d scarcely study it if I didn’t think it existed.

  70. 70
    uoflcard says:

    But what about modern researchers who WITNESS genomes crossing gaps? If it wasn’t a random search, then what was it?

  71. 71

    Can you tell me what you are talking about?

    Link?

  72. 72
    Petrushka says:

    The Case Against a Darwinian Origin of Protein Folds

    That’s one of those interesting arguments invoking the problem of big numbers. Not only are the numbers big, but he argues that there are no shortcuts to designing proteins.

    Which implies that it is impossible for a “natural” designer to design proteins.

    Which is the argument I raised on the contest thread, and which was rejected.

    So either Douglas is right,and proteins cannot emerge and evolve naturally, and must all be specially created, or he is simply wrong about the landscape, and they can evolve.

    If they can’t evolve, it is certainly not possible for humans or any finite entity to design them. Remember Axe’s observation that there are no shortcuts. It’s evolution or intervention. For every protein.

  73. 73
    Eugene S says:

    Nor any other kind of search can except intelligent interference.

  74. 74
    Eugene S says:

    Elizabeth,

    I strongly suggest you first read the papers I referred you to and think them through before continuing to argue the existence of complexity gaps/islands of functionality in the space of system configurations.

    It is not my idea, I only accept it based on the work of others. I don’t see anything in those papers that would contradict common sense in the way TOE does. Please read the papers before commenting any further.

  75. 75

    Could you tell me, specifically, what you mean by “modern researchers who WITNESS genomes crossing gaps”?

    Meanwhile, I’ll have a look at those papers.

  76. 76
    Petrushka says:

    Wasn’t that the whole point behind James Shapiro’s new book? That the evolution we witness is nothing like the “smooth tree” Darwin fantasized about and legions have worshipped since?

    It could be the point of his book, but it would be a weak assertion, because genomics is beginning to fill gaps. Take a look at the Koonin book that was offered free from Amazon. It isn’t free anymore, but it was offered.

    The tree metaphor doesn’t work for single-celled organisms, because they exchange genetic material.

Leave a Reply