From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Thu Apr 12 2001 - 21:18:45 MDT
Damien Broderick wrote:
>
> At 10:22 PM 4/12/01 -0400, Eliezer wrote:
>
> >I can't tell what the problem is from your post. Could you be more
> >specific?
>
> No, but I can be more general:
>
> http://www.bcp.psych.ualberta.ca/~mike/Pearl_Street/PSYCO452/Week6/week6.pdf
>
> e.g. (discussing how autoassociative networks can overcome
> underdetermination [and help explain overdetermination]).
Damien,
I'm already familiar with Hopfield networks, the analogy to spin glasses
in solid-state physics, and so on. I think that they're fine for a
monolayer problem but lack the holistic structure needed to, e.g.,
visualize a triangular lightbulb. See CaTAI section 2.3.4, "Concept
application and combination".
Also, AFAIK, current Hopfield models use a single state which iteratively
relaxes, rather than a set of superposed states; the actual physics for
most phenomena resembles a probability fluid forming puddles in local
minima of a potential energy surface. (JoSH once told me about a
generalization of the Hopfield network which used analog computation (!)
to near-instantly find and combine the local minima in dynamically
changing potential energy surfaces, but I forget what that was called.)
Now that I've shown off my erudition and proven that I know the background
material, would you please be more specific?
-- -- -- -- --
Eliezer S. Yudkowsky http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:06:57 MST