RE: Evolutionary computation (was: Where's the Evolution?)

Cummins (
Tue, 20 Apr 1999 12:49:19 -0500

> []On Behalf Of Brian D Harper

> If you want a really good yet enormously simple example of
> doing new things you can try Langton's ant.

The mechanism to create complexity that evolutionists have proposed is
mutation+selection. The mutation is random and the selection is based on
competition and environment.

None of the computer programs that have been mentioned even attempt to do
that. That PLGA was based on intelligent selection (a predefined goal).
The Terria project also lacked natural selection. The ant program might
form some sort of ordered pattern, but there's no competition and you didn't
mention anything about mutation. You concede that the outcome is
deterministic (thus does not demonstrate any creativeness).

When there is no natural selection, there is no such thing as anything being
unfit. That alone is why all these programs fail to even begin to
demonstrate evolution or that nature has any significant ability to create

> I hope you won't take this the wrong way :), but the above
> description of Tierra World is just wrong.

You could have expected me to want to know "how so."

> >Because this minimal size code "evolved" down, not up, being irreducible
> >isn't a problem (and neither is non-viable intermediate steps,
> see above).
> What do you mean by evolving down? BTW, you are the one who says
> evolution *has* to go toward increasing complexity. Whether
> down or up (whatever that means) it is still evolution.

Note the quotes. Besides, if it's not an increase in complexity, then it's
not impressive.

> Let's recall, however, what Ray said " has packed a much
> more complex algorithm into less than half the space". According
> to Ray, it is both shorter *and* more complex.

That has nothing to do with the irreducible nature of the code. It's not
significantly more complex, unless Ray thinks cutting the fat is an increase
in complexity. I mean, it doesn't do anything fundamentally different than
its ancestor a million generations ago.

> It was more successful than those that died.

The only way to die is to fail to copy some number of times, not that they
competed against each other.

Doesn't a double-digit size genome and millions of mutations (virtually none
of them eliminated due to selection) make any demonstration of "complexity"
unimpressive to you? Now, you want to tell me specifically what the most
complex creature could do that the first one couldn't? Just copy in a
slightly different style ("unrolling the loop")?

Instead of stones for bread (the other computer programs that supposedly
show evolution), how about someone set up a program that has some real
competition between creatures (e.g. a modified Quake 3 engine with bots that
fight each other, those that survive create mutated copies of its script).
Turn off the graphics and let it run for a month. You think after a billion
generations we might end up with sexual reproduction and other wonders of