From: Dan Fabulich (daniel.fabulich@yale.edu)
Date: Tue Feb 29 2000 - 22:13:33 MST
> I was very much thinking of something beginning in the unconscious, with
> only the final phases emerging to consciousness.
Again, I'm skeptical about theories like this. How could we have concrete
evidence that this is what actually happens in the human brain?
> As to whether the mechanism operating in the unconscious to produce new
> ideas has the form of natural selection, the proof would be if AI
> researchers discovered that they needed to build structures that embodied a
> natural selective process in order to see creativity operating in their
> systems.
This would provide *some* evidence for your theory, though it
certainly wouldn't clinch the matter. You began w/ a very strong
proposition, remember:
---
> There is a proposition that all creative processes have natural
> selection as an underlying mechanism: 1. Mutation or a random
> information source external to the system. 2. A progressive winnowing
> of the information for compatibility or usefulness. 3. Amplification
> and combination of the survivors to compete at a higher level.
---
It's one thing to say that SOME creative processes have natural selection
as an underlying mechanism. I take this claim to be *fairly*
uncontroversial. (Though I could easily imagine some stick in the mud
insisting that if it's not a product of Free Will, then it's not really
Creativity. Doesn't convince me, but it does convince a lot of anti-AI
thinkers like Searle. Anyway, I find Genetic Algorithms impressive enough
that I'm willing to call them "creative" processes.) But your claim is a
lot stronger than that; it claims that ALL creative processes have natural
selection as an underlying mechanism.
Don't forget, also, that AI researchers can't "discover" that they *need*
to use a natural selection process (say, GAs) to get creativity. All AI
research can ever show is that an AI which employs natural selection in a
certain way can pass the Turing Test with regards to creativity, and that
no other mechanism we know of will do the job. Again, this is no small
bit of evidence, but this is quite far from a proof.
It's also pretty far from brain science. A lot of empirical theories have
been posited for how the brain does what it does; the memetic view has
been a sufficiently hard-to-map way of talking about brains that no
structural parallel to memes has ever been empirically established.
(Again, I welcome you to try; you'll certainly get published for it if you
succeed.) Unless and until you can say "look at these data: THIS is where
natural selection happens in the brain!" then your proposition will always
remain "just a theory."
I don't mean to say "just a theory" the way so-called Creationists claim
that evolution is "just a theory." A *number* people on this list and in
the scientific world at large believe that the brain is capable of doing
uncomputable stuff. (Penrose is, at present, the most popular such
writer.) So even if you *could* somehow prove that a computer needs
natural selection before it will yield creativity (which you can't),
objectors could still point out that there might be something uncomputable
going on in the brain, in which case, the brain is not a "computer" in
that sense, and all bets are off.
To clarify, you might try to argue this way:
1) Natural selection works in AIs and we don't know of any other way to
get creativity
2) Therefore natural selection is the only way for a computer to have
creativity
3) Therefore natural selection is the only way to have creativity at all
4) Therefore the creativity present in the brain has natural selection as
its mechanism
*I* think most of these lines are right, even if the moves aren't very
plausible, but a lot of people would disagree with me. Worse, this
argument smells like that sort of science which is called "deductive"
science by its friends, "theoretical" science by neutrals, and
"non-experimental" science by its polite enemies.
Anyway you look at it, you're still stuck with the problem of
unverifiability at every line except the first, and possibly the last. I
say "*possibly* the last" since I can't think of any good way to verify
line 4 empirically, but I could imagine some experiment which would
establish the point convincingly. In contrast, 2 and 3, while they agree
with my intuitions, would only be a reasonable inference to me given 4; 1
is thus way way short of proving 4.
-Dan
-unless you love someone-
-nothing else makes any sense-
e.e. cummings
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:27:06 MST