[from Gary Cziko 940808.0544 GMT]
Bill Powers (940805) writes about natural and artificial evolution:
The problem with Beer's cockroaches is that he was trying to get them to
evolve a nervous system that would steer the cockroaches to a food
source. In other words, he was trying to get blind evolution (with
recombination) to do what we assign to goal-oriented reorganization.
The process he built into the program wasn't really evolution, but
shaping. The program could detect when a random change in the
cockroach's circuitry produced movements slightly biased toward the
distant food source; the cockroaches that did this were rewarded by
being allowed to propagate. So without realizing it, Beer was using the
E. coli type of control system instead of either pure natural selection
or pure recombination.This same approach was used in a couple of other GA processes that I
have seen. I don't know how universal it is, but it's very easy for
someone to include an unspoken goal in such simulations and never
realize that its effects will completely dominate supposedly "random"
processes.
Bill, I think I understand what you're objecting to here. In natural
selection, there can be no selection of organisms who are "moving in the
right direction" unless this moving has some payoff of its own. I agree.
And I have seen and heard of many artificial life models of evolution that
do not make use of the goal of the researcher in this way (Ray and Yaeger
come to mind).
But what is the "random" process you're referring to. In evolution, the
variations either by recombination or mutation or "random" (I prefer
"blind") but selection is decidedly nonrandom, as it is biased toward
survival and reproduction. Couldn't one say that in natural selection,
surivial and reproduction "completely dominate" the process?
RE: recombination
Most artificial breeding programs end up with forms that please the
breeder but are awfully poor prospects for survival (look at show cats,
show dogs, race horses, dairy cattle, and so forth).
Also take a look at the peacock's tail and how hard it is for him to get
off the ground.
Breeding is also an
E. coli type of process: the breeder has some goal in mind and allows
propagation for any offspring that move _toward_ the goal. It's not that
recombination is so much more efficient than natural selection;
But all natural selection involving sexually reproducing organisms makes
very heavy use of recombination. I don't understand the distinction you
are trying to make between recombination and natural selection.
it's
recombination together with a goal-seeking E. coli type of control
system that is so much more effective.
It is thought that dinosaurs got big because bigger bodies better retain
body heat. Any little increase in size could pay off in better survival.
For gazelles, any little increase in speed will help them stay away from
teh jaguar. And jaguars who are just a little faster are better fed. How
is this unlike the artificial E. coli approach to breeding?
Fitness differences will
be far less for recombination than for pure natural selection based on
mutation. It's really only faith that leads evolutionists to see
recombination as similar to evolution; after all, recombination can't
create any new control systems that aren't already present in the gene
pool. Mutation can.
The recent work on genetic algorithms and genetic programming suggests that
it is the recombination of sexual cross-over and not mutation which drives
evolution. Mutation is often not even used in these algorithms.
In either case, goal-directed recombination or mutation will be far more
effective than simple random recombination or mutation. Recombination is
very strongly affected by ordinary behavioral control systems such as
those that produce language or courtship customs and standards of beauty
or virility. If men start thinking that women with large bottoms are
sexier than those with smaller ones, it won't take long for most women
to be born with large bottoms. That's e. coli type recombination, not
random recombination.
In evolution, the randomness (blindness) refers to the variations. In E.
coli movements, the randomness (blindness) refers to the direction of
movement. And sexual selection is as much a part of evolution as selection
for survival. I don't see the point you're trying to make here.
I think that all modeling of evolutionary processes has to be re-
examined to see whether there weren't hidden E. coli type processes
built unwittingly into the simulations.
Bit it seems to me that there are E. coli type processes built into natural
selection as well (e.g., evolution for larger size, longer tail feathers to
attract mates, longer wings for better flight, larger eyes for better night
vision, etc.)
This process is clearly not
understood by evolutionists who object to the Cairnes et. al. findings.
They are still arguing as if internally-directed evolution had to
somehow bias the mutations themselves, making the changes non-random.
They don't yet understand that you can get directed evolution with
COMPLETELY RANDOM mutations (or completely random recombinations) just
by varying the interval between them.
I've seen at least one critic of Cairns (Donald McPhee) who argues very
much along these lines. There is a higher rate of mutation of E. coli when
they are placed in a medium with a "foreign" sugar.
Bill, I appreciate your comments about the artificiality of artificial
selection that selects for movement toward a goal. But I'm having a hard
time (as you can see) following your other arguments. Perhaps you could
clarify a bit for the man on the net fascinated by both evolution and
perceptual control?--Gary