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

Cummins (
Wed, 7 Apr 1999 22:35:37 -0500

> From: Brian D Harper []

> I'm glad to see that you accept results of computer simulations.

Of course. Without human input, a computer program is just a structure
where natural forces duke it out. Getting a computer to print "hello" is
just as natural as dropping a rock and watching it fall.

> Perhaps the best (or at least best known) simulation of Darwinian
> evolution is Tom Ray's Tierra World. For a summary I invite you
> to point your web browser at:

Several years ago I downloaded and tried to compile the Tierra project, but
I didn't have much luck.

The quote you provided talks about a creature with a genome of size 36 from
a parent of genome size 80 (after millions/billions of generations). I
won't dispute that nature can optimize, simply, and destroy complexity (does
that small creature still have the creator-given ability to read from other
genomes?). And, I'll wager that's what you'd see if you could compare the
original parent with the size 36 creature. The last creature may have do
things a little differently because of changes to its code, but it certainly
doesn't do any new kinds of things.

Without regard to the prize creature of the size 36 genome, but about all
the creatures. There's no real competition between them, thus no way to
test fitness. Creatures die only by failing to copy legal code some
arbitrary number of times. Thus, the result is brute force with no concern
for non-viable (unfit) stages (e.g. if the creatures were furniture, there
would be no problem with a chair having only two legs on the path to
becoming a legless seat). The brute force technique also means that there's
no appreciation of whatever complexity that might appear by luck (brute
force, any combination of instructions will eventually be reached --
millions of mutations to a genome that started and ended in the double-digit
genome size is the example from what you quoted), so whatever comes along
can go just as easily (ie., put it in an environment where it will be tested
for fitness and see how fast it's destroyed, like a sandcastle on a beach).
Brute force doesn't demonstrate that nature has any creative ability.

>Not only does this show an example of increasing complexity
>arising from "random bit flips", the final result is also
>irreducible: " every component of the code must be in
>place in order for the algorithm to function."

Because this minimal size code "evolved" down, not up, being irreducible
isn't a problem (and neither is non-viable intermediate steps, see above).
The issue is how the creature's required functions formed in the first
place -- it was directly created by an intelligent programmer (the program
didn't create the individual functions). And, it again took an intelligent
programmer to appreciate the size 36 genome (i.e. this creature was no more
successful than any other).

Actually, there is a very crude fitness test provided by a program and I bet
dollars to donuts that the size 36 creature would be judged highly unfit.
Because it's irreducible, there's a high probability that "mutations" will
create illegal or non-reproducing code making it quickly win the race for

Any attempt to demonstrate evolution on a computer results in either garbage
or simplification. The Terria, underneath all the smoke and mirrors, is
just another demonstration that nature doesn't create complexity.