Random Mutations: vjtorley

vjtorley, at UD, writes a post entitled It’s time for scientists to come clean with the public about evolution and the origin of life that includes this:

Edward Frenkel, a professor of mathematics at the University of California, Berkeley, recently reviewed a book titled, Probably Approximately Correct: Nature’s Algorithms for Learning and Prospering in a Complex World (Basic Books, 2013) by computer scientist Leslie Valiant, in a report for the New York Times (Evolution, Speeded by Computation, September 30, 2013). The following excerpt conveys the gist of Dr. Valiant’s conclusions:

The evolution of species, as Darwin taught us, relies on natural selection. But Dr. Valiant argues that if all the mutations that drive evolution were simply random and equally distributed, it would proceed at an impossibly slow and inefficient pace.

Darwin’s theory “has the gaping gap that it can make no quantitative predictions as far as the number of generations needed for the evolution of a behavior of a certain complexity,” he writes. “We need to explain how evolution is possible at all, how we got from no life, or from very simple life, to life as complex as we find it on earth today. This is the BIG question.”

Dr. Valiant proposes that natural selection is supplemented by ecorithms, which enable organisms to learn and adapt more efficiently. Not all mutations are realized with equal probability; those that are more beneficial are more likely to occur. In other words, evolution is accelerated by computation.

The criticisms being made here of the Darwinian theory of evolution are pretty devastating: not only is it far too slow to generate life in all its diversity, but it’s also utterly incapable of making quantitative predictions about the time required for a structure of known complexity to evolve, by natural selection. And there’s no reason to believe that the “nearly neutral theory of evolution” espoused by biologists such as Professor Larry Moran would fare any better, in this regard.

 

Dr Torley is a scholar and a gentleman and someone for whom I have a great deal of personal respect. In fact I owe him more than one debt of personal kindness.  But that does not mean that I think his ideas are correct, and I submit he is profoundly wrong here in an extremely useful way.  Unusually, the passage he cites is very specific about the kind of randomness that cannot be the kind of randomness that would produce Darwinian evolution: equally distributed.

Leslie Valiant is absolutely correct: equally distributed mutations (any mutation as probable as any other) could not possibly result in Darwinian evolution.  And Darwin (who to my knowledge did not even use the word “random” nevertheless “equally distributed”) proposed nothing resembling such a thing.  And nor has any neo-Darwinian either.

In other words, “random equally distributed mutations” is a straw man of epic build.  The unusual thing about Dr Torley’s quotation is that it is explicit.  The “equally distributed” part is more usually taken as read into the meaning of “random” when “random” is held up for ridicule as an important factor in the evolution of complex life, as we can tell from the persistent avoidance of tackling the eleP(T|H)ant in the room.  Even though Dembski tells IDists to use “the relevant chance hypothesis” all calculations for the alphabet soup of FIASCOs that I have seen, including Durston et al, assume blind draw from a distribution that may or may not consist of an equiprobable distribution of bases, but does consist of an draw from an unordered bag of the things, a process that will make any drawn sequence of given length as likely as any other.

And of course, the way that mutations occur is nothing like that. A mutation is most likely to be extremely similar to the sequence from which it mutated, slightly less likely to be a little more different, and extremely unlikely to be very different.  Moreoever, as, by definition, the parent sequence belongs to a cell or organism capable of reproducing, i.e. is part of a viable genome, any mutation is far more likely to be viable than not. In other words, mutations, far from being “random, equally distributed” variants, are colossally biased in favour of viability.  Which means (and it is what we observe) that the vast majority will be neutral or near-neutral with respect to reproductive success, with some being slightly better, and, at least in a well-adapted population with a genomes that are already near-optimal, probably slightly more being slightly worse.

Even more importantly, the fact that the vast majority of mutations will be near neutral with respect to reproductive success means that most populations will be extremely rich in variants with more or less equal, and near optimal, viability in the current environment.  This in turn means that the population is extremely well-equipped to face any change in that environment – which has a rich pool of variance to sample from in order to move the population to a new optimum.

And this is why both drift (the generation of a rich pool of near-neutral variants) and Darwin’s process (heritable variation in reproductive success aka natural selection) are such a powerful pair of factors (not that they are really separate factors; we can regard biased sampling, i.e, as special case of random sampling, just as the distribution of throws of an eccentrically weighted die is a variant on the distribution of throws from a centrally weighted die, but still “random”) in accounting for adaptive evolution, and why, contrary even to Darwin’s prognosis, adaptive evolution is so rapid, as the Grants discovered with Darwin’s own finches in the Galapagos.

Valiant is correct – if mutations were “random, equally distributed”, Darwinian evolution would be impossibly slow.  But they aren’t, and so Darwinian evolution is impressively rapid.  The limitation on its net movement (as opposed to its ability to keep finch populations optimised under changing environmental conditions) is not due to the mutations being “equally distributed”, but, ironically, because they are not – because of the strong bias in favour of what works in the current environment, the distribution of variants is necessarily limited, and too rapid an environmental change will tend to deplete the pool of the very range of variation that will confer robustness in the face of further change.  Which is why, of course, that rapid habitat loss causes extinctions, rather than successful adaptation.  But as long as environmental change is slow enough that the pool of neutral variants can continue to be enriched, and there is a reasonable spread of viable variation round the optimum, adaptive evolution can and must occur, and, indeed, can be observed doing so in real time.

60 thoughts on “Random Mutations: vjtorley

  1. coldcoffee: Personally I wouldn’t ban peoplefrom forums unless they spam, but I don’t really know UD rules – I haven’t read the rules!

    Don’t bother. The only rule that is actually in practice is don’t do anything the regulars don’t like.

  2. Apropos of group selection, Dr Liddle remarked:

    Or the capacity for language in groups that have developed a language in comparison to the same capacity in groups that haven’t. That would be a kind of neo-Lamarckian group selection.

    This topic may be worth a separate thread, though I am unqualified to write it.

    However, your statement as it stands struck me as odd. How could we tell if a group has a “capacity for language” unless it, in fact, developed one? The proposition you make seems non-explanatory to me.

    In any case, whether or not group selection exists as a valid evolutionary mechanism, I do agree that language operates in a neo-Lamarckian manner.

    It seems to me to be part of the definition of a language that this must be the case – that is: knowlege and survival techniques can be shared within a group and passed across generations without the knowledge content having to be encoded in genetic material. All that is required is the capacity to learn the shared language symbols.

  3. timothya:
    Apropos of group selection, Dr Liddle remarked:

    This topic may be worth a separate thread, though I am unqualified to write it.

    It is an interesting topic, and I’m not really qualified either! And it would be a difficult hypothesis to test.

    However, your statement as it stands struck me as odd. How could we tell if a group has a “capacity for language” unless it, in fact, developed one? The proposition you make seems non-explanatory to me.

    Think of it in terms of bootstrapping. Let’s say that a lineage of proto-humans forming hunting teams, as some modern primates do, and that this involves, sometimes, signalling to each other that a quarry has been sighted (as actually happens), and that as a result, the beginnings of a rudimentary symbolic language is developed among the troop, in which one call means “meaty monkey” and another means “stay, hidden, I’ve got him in my sights” and another means “go round to the right and creep towards where I’m pointing”. Probably combined with gestures in the early days. Clearly those individuals who pick this system up quickly, are going to tend to be part of successful hunting forays, bring home more meat to their developing offspring, and leave more progeny. So far, so Darwinian. Moreover, troops that develop this kind of system are going to do better than troops than don’t (group selection) and are going to be better equipped to fight (better nutrition) when one troop attacks another.

    But typically genes are shared between neighbouring troops – males are killed and females mated with. So language ability genes are spread across troops, even though a specific call code may not be (we see this in whale pods, for instance – whales have the capacity to communicate by sound, but the sound repertoire of each pod is unique). And some troops will have more effective language traditions than others. A proto-human with good language ability born into a troop with a very rudimentary language tradition won’t get a lot of benefit from it. But a proto-human with good language ability born into a troop with an already quite advanced language tradition will get a great deal of benefit from his/her ability, and the troop itself will get a great deal of benefit either – a rich language with good practitioners. And again, that will benefit the troop in raids on other troops.

    In other words, a language-using troop provides an environment in which language ability is a selective advantage, both at the individual level (the individual will tend to be part of successful hunting trips and bring back good food for his offspring) and at the troop level (troops with good language and good language users will do well against other troops). So you could (and this is very speculative, although not original to me, although I’m afraid I can’t cite my source) get a powerful positive feed-back loop in which language ability (genes that promote good language, and we know they exist) goes hand in hand with language enrichment at cultural level.

    In any case, whether or not group selection exists as a valid evolutionary mechanism, I do agree that language operates in a neo-Lamarckian manner.

    I don’t see why it shouldn’t – a group is a self-reproducing unit, with shared genes, just as a multicellular individual is. The mechanism only requires self-reproduction with variation in reproductive success. And at the group level, a specific language is “heritable”, it’s just that its inherited by example, not by genes. But of course Darwin didn’t even know about genes (and considered that Lamarck’s idea might be a possible source of variation).

    It seems to me to be part of the definition of a language that this must be the case – that is: knowlege and survival techniques can be shared within a group and passed across generations without the knowledge content having to be encoded in genetic material. All that is required is the capacity to learn the shared language symbols.

    And having the language itself – and being able to teach it, or have someone learn from it, yes.

    It would be quite a complex set of feedback loops, involving genetics, learning, passing on by example and by genetic inheritance, selection at the level of the individual and selection at the level of the troop. But I don’t see why it couldn’t happen. Complicated feedback systems happen all the time, even outside living things!

  4. timothya: In any case, whether or not group selection exists as a valid evolutionary mechanism, I do agree that language operates in a neo-Lamarckian manner.

    I find myself swayed by Jerry Coyne as to whether group selection is a real process and unpersuaded by. David Sloane Wilson

    Regarding the evolution of language, I find Geoffrey Miller’s arguments about sexual selection being a factor amusing. PDF

  5. JonF: The only rule that is actually in practice is don’t do anything the regulars don’t like.

    Yes, I think that’s a factor. I don’t think I was banned for anything I said at UD. I think it was more likely pressure on Barry from Kairosfocus, who was becoming increasingly upset by my very presence at UD, because I hadn’t censored something that had been said about him here. Even though I gave KF every opportunity, including direct invitations, not only to respond here, but to post his own OP, and also explained very clearly that I did not endorse all the opinions expressed here – quite the reverse – I welcome dissenting opinions and the whole point of my blog is to allow dissenting views to be expressed and discussed without fear of being banned or censored.

    But whatever. The nominal grounds were bizarre. I can’t even remember what they were this time. Something that offended Barry’s idea of what was a reasonable opinion to hold, anyway.

  6. Dr Liddle:

    Think of it in terms of bootstrapping.

    Moreover, troops that develop this kind of system are going to do better than troops than don’t (group selection) and are going to be better equipped to fight (better nutrition) when one troop attacks another.

    Possibly so, but the “are going to do better” is surely a non sequitur. It is equally plausible that purely Darwinian selection would lead to better-performing troops (better sense of smell/sight/hearing; better rock-heaving strength etc etc). Perhaps the mistake, if it is one, is to tie language capacity too closely to predation.

    So here is my hypothesis, which in principle, is testable. Superior language capacity will arise among lineages that are not top predatore, but instead among lineages that rely on a range of food opportunities, which in turn are fugitive and require complex gathering strategies.

  7. timothya:
    Dr Liddle:

    Possibly so, but the “are going to do better” is surely a non sequitur. It is equally plausible that purely Darwinian selection would lead to better-performing troops (better sense of smell/sight/hearing; better rock-heaving strength etc etc). Perhaps the mistake, if it is one, is to tie language capacity too closely to predation.

    That’s possible. This is why I think Dawkins’ glib claim that “mutation is random, selection isn’t” is false. Both are highly stochastic, and while can predict, generally, that what promotes reproduction will become more prevalent, and thus evolve, we can’t predict what, any more than we can predict the weather a more than few days ahead. Which is why evolutionary psychology is frustrating as a science – behaviour doesn’t leave fossils, and we can’t re-run the process. But we can conjecture pathways by which language might have evolved, nonetheless.

    An idea that has tended to appeal to me more than the hunting one (although what we now know of primate hunting teams is quite suggestive) is that language was also useful for successful bargaining – which for hunter-gatherers is important, because big game doesn’t keep and is too much for one person. And that might have been women’s role, which just might explain the slightly superiority of women in language. But it’s difficult to do more than speculate.

    heh. My very thought. Yes. Or possibly lots of things. If we have learned anything from the human genome project it’s that things are a heck of a lot more complicated than we thought they might be!

  8. Unfair to Dawkins! Who imagines that selection unerringly picks the most fit individuals in a population? Even the most sleek, alert and sexually attractive antelope might simply be looking the wrong way when the leopard drops out of the tree. Stochastic it may be, but biassed in the direction of reproductive success selection most certainly is.

  9. Of course it’s biased. It’s the idea that you can call mutations random and natural selection not that I take issue with. Both are stochastic, neither are uniform probability distributions. Novel variants are far more likely to work than not, and some kinds of mutations are much more likely than others. Natural selection can be seen as a bias on drift – an extremely slight bias in most cases.

  10. Wait a sec . . . Novel variants are far more likely to work than not? To the extent that they are variants on successfully selected genomes, then that is reasonable, if that is what you meant. Not so sure I agree with the second part it depends on whether you mean “kind” in the biochemical sense of “type of substitution, or in the sense of “more likely to work”.

Leave a Reply