From: mjg223 (mjg223@is7.nyu.edu)
Date: Sat Feb 26 2000 - 02:16:09 MST
Well, here's a Yea.
I'm as firmly in the bottom up camp as one can be, possibly to a
fault. It's my sincerest dream that, once we've got it all figured out,
we'll be able to describe the mind on one side of one sheet of paper.
An abstract model for a universal learning machine... and all this junk
about which part of the brain does what will be about as relevant to a
real understanding as control register documentation is to Turing
computability. The idea that the mind is just an ad hoc junk yard of
semi-vestigial crap horrifies me.
The ant colony optimization article you mention is a perfect example:
put a bunch of idiot state machines together and perfection
emerges. And not in a mysterious way either - the ant foraging
algorithm is as elegant, intuitive, and esthetically satisfying as any
idea I can think of.
I don't know if you need external noise to break symmetry. I suppose
so. When I was a kid, I'd wonder what it would be like to meet an
exact copy of myself. We wouldn't be able to start a conversation:
We'd just mirror each other perfectly, our mouths perfectly
synchronized, as we tried exactly the same strategies to bootstrap our
way out. I decided the solution was to go talk to someone else, and
ask them to pick one of us - a plan we'd conveniently come up with
simultaneously. It's an interesting idea, but the real world is such
an imprecise place that I doubt that it's of practical importance. The
symmetry in any real system butterflies away before it's a
problem. More broadly though, I would agree that exposure to novelty
is essential to keeping the brain healthy, to stop it settling into
some comfortable local maximum. It's like simulated annealing - you
jostle the system every time it looks like it's converging. More
broadly still, it seems to me this is the purpose of education.
Anyway, I appreciated your post. You raise some interesting issues -
thanks for perturbing my lazy grey matter.
-matt
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:27:02 MST