Re: Behaviorism vs. evolutionary psychology

On Jul 13, 3:03 pm, Don Geddis <d...@xxxxxxxxxx> wrote:
JGCASEY <jgkjca...@xxxxxxxxxxxx> wrote on Thu, 12 Jul 2007:

If you want to "evolve" the sequence 123456789 there may
be something that can generate every possible 9 place
number and test it for success. But as that number gets
bigger the chances become exponentially unlikely of
success.

You've mentioned things like this kind of "trial and error"
before. But that's not how evolution works.

The result of evolution may be a very, very unlikely thing,
in terms of the space of all possible arrangements of molecules.

But evolution isn't randomly assembling huge structures of
molecules, and then hoping one of them luckily turns out to
be a human being.

Instead, it's a very long series of small, incremental
improvements.

In fact, the silly (anti-evolution) Intelligent Designer
attempt to use an argument like this, by trying to find
biological structures that could not possibly evolve in an
incremental fashion, but need to exist as "all or nothing".
Their work is silly, but they are at least correct that IF
it were possible to prove such biological structures
existed, THEN the theory of evolution would have a hole.

Because evolution does NOT proceed by random trial and
error of assembling huge structures.

Instead, it is very slow and gradual transformations from
one form to another, with every intermediate form also
having some survival advantage.

That's why the odds aren't as bad as you suggest. Because
it isn't random ("Monte-Carlo") search through the space;
it's hill climbing with a fitness function to guide you.
Genetic mutation gives you some nearby point in the species
space, and the species follows the direction that seems to
lead upward (in the local fitness space).


I don't think you read my post very carefully. The paragraph
you quote was the problem. I then went on to explain as you
have repeated the way it must have happened with evolution.

However the point of the post was with reference to AI and
how incremental improvements were also how we build up our
abilities to do intelligent things. In other words how to
make reinforcement learning work in practice.

--
JC


.

Relevant Pages

  • Re: Defining&testing "macroevolution" (was: Testing the Laws of Intelligence)
    ... as evolution is strictly defined. ... For example, in the case of a ring species, we make it ... clearly two populations definition are ... cross-breed success relative to mono-breed success, ...
    (talk.origins)
  • Re: Behaviorism vs. evolutionary psychology
    ... can generate every possible 9 place number and test it for success. ... But that's not how evolution works. ... by trying to find biological structures that could not ... I am not a vegetarian because I love animals; I am a vegetarian because I ...
    (comp.ai.philosophy)
  • Re: 7 Habits of Highly Effective Anti-Evolutionists
    ... so looking for weaknesses in evolution may impede your own success (as ... =A0Modern creationists allow species to evolve within 'created ... agreed with part of evolution. ... creationist success - the number of different species was too large to ...
    (talk.origins)
  • Re: If you believe any of your theories, youve turned science into a religion.
    ... Evolution, if true, should be happening today. ... Keep in mind that the idea of a "species" ... see where the divergence was fairly recent, ... very success may doom us as a species within a short time. ...
    (sci.astro.amateur)
  • Re: comparing emotions
    ... evolution and clearly beliefs,ideals and fantasies have contributed to ... the success of the species and it's collective survival. ...
    (talk.origins)