Re: Is the complexity of evolutionary change explainable?

dkomo wrote:

hrvoje-d wrote:

To Dkomo:


The explanation of the anology
--------------------------------------------

I don't need to get a 100 billion mutations on that 16 000 bits - there
are 3 billion nucleotides in the DNA! The number of bits per individual
is 160 (or 320 if you consider that they occur in pairs (an rough
analogy to homolog chromosomes). This is because the individual has 10
chromosomes each that controls one characteristic (a variable, a
number) that has to adapt (so that it comes close to a maximum of a
function).

Here is my logic:

To say the main point in a few words: I am not looking at mutations of
the entire DNA, but the segment of it which is, for example responsable
for two things (comparing humans and early primates):

Changing the shape of a hand for it to adapt from a walking tool to a
tool used in manipulation (making the tools for hunt). There are many
things to be changed in a hand to make that transition: for example
more nerve endings to make it controlable.

Now, I think its ok to assume a continuing population of 10 000 of
individuals that are adapting in such a way, and also lower the
mutation rate in the algorithm from 10^-3 to 10^-5 (1 in a 100 000) -
there will be no differences in mutation effects. Again, I assume that
there is a limit of recombinaton, so this transition has to occur by
mutation (only slightly).

Now, I'll make a transition from nucleotides to genes with this
assumption "the effective mutation rate of a gene is 10^-8" This is
debatable because the mutation rate af a gene is 10^-5 if it has 1000
nucleotides, but still - a single nucleotide in it is mutated of 1000
of them (10^-5 is the probability of it happening).


I was busy so I didn't have a chance to reply yesterday. It looks like you are making an argument for rescaling the total amount of mutation to account for the fact that the evolution of the human only used a small
^ ^
correction: should read "human hand"

part of the total genes in the human genome. And you're trying to map that rate of mutation to the 10 chromosomes per individual in your GA. I pretty much agree with that, without agreeing with all the specific details. I'll just point out that the estimated total of 200 billion mutations for the human genome since 1,000,000 BC I came up refers to the coding/regulation part of the genome, and doesn't include the "junk" DNA. So this is about 2% times 3 billion base pairs.

But none of this is that important anyway. After looking at your web page at http://www.ffdi.hr/~hdagelic/evo and thinking about your GA a little more, I concluded that your casting biological evolution as similar to a GA optimization problem is incorrect anyway. Evolution doesn't find global fitness peaks in genome space, only local peaks! And these peaks can be quite low in fitness and prevent a global peak from being found. All that matters is that the population fitness *is high enough to prevent the species from going extinct.* It's not the same thing as what a GA does, which is to converge to an maximum/minimum.

So you have already shown that mutation and crossover actually do work. All you need to show is that when you stop the iterations on your GA program, the average fitness of the population then is greater than the average population fitness you started with. And study this over different mutation rates and iterations. It doesn't matter in the least whether the GA finds a global maximum or comes close to it.

Let's look at this in another way. You have, in common with some intelligent design advocates that have posted to talk.origins before, committed the "specification fallacy" in regards to the human hand. You have assumed that the hand was a target for the evolution that occurred, in analogy to the way one of your faces on the web page represents the "best" face to which the GA is supposed to converge. This is a complete misunderstanding of evolution. The human hand emerged from an almost endless amount of trial and error over millions of year. There are a myriad number of other hands that could have evolved, some better, some worse, that would have been "good enough" to allow us to survive as a species. And there are probably a large number of "better" hands that could have emerged as well, that we'll never see, because our hands are trapped on local fitness plateaus.

That's what evolution is. It does as little work as it can get away with. It's a tinkerer, but not much of an optimizer.

[snip]

Another thing - this problem is simple, the hand isn't.

LOL. Actually, your problem is hard, but evolution of the hand is relatively simple.


--dkomo@xxxxxxxx


.

Relevant Pages

  • Re: Is evolution more then mutation and selection?
    ... I recommend rereading Darwin, as well as more ... modern books on evolution. ... The shifts in fitness ... ] mutation tend, on the whole, to be ...
    (sci.bio.evolution)
  • Re: Part 1 (of 3): What are major aspects of evolutionary theory?
    ... >> in phenotype, hence in fitness. ... >of the time giant steps can be ignored as a course of evolution. ... >all it takes is one such giant mutation to succeed every few million ... >already-exiting variation is tested and only the best is retained. ...
    (talk.origins)
  • Re: Is the complexity of evolutionary change explainable?
    ... there will be no differences in mutation effects. ... I'll make a transition from nucleotides to genes with this ... It looks like you are making an argument for rescaling the total amount of mutation to account for the fact that the evolution of the human only used a small part of the total genes in the human genome. ... And these peaks can be quite low in fitness and prevent a global peak from being found. ...
    (talk.origins)
  • Re: Part 1 (of 3): What are major aspects of evolutionary theory?
    ... of the time giant steps can be ignored as a course of evolution. ... all it takes is one such giant mutation to succeed every few million ... So maybe the correct metaphor is a fitness ... already-exiting variation is tested and only the best is retained. ...
    (talk.origins)
  • Re: Fraud of Evolution sidebar
    ... " The Human Genome Project, a joint international effort to unravel the ... Since at most two nucleotides can change per generation, and evolution ... mutation in them is necessarily more than a billionth of the genome. ...
    (talk.origins)
Loading