Re: G. Chaitin on AI

From: Eliezer S. Yudkowsky (
Date: Sat Mar 02 2002 - 08:38:46 MST

Ben Goertzel wrote:
> But the question is: is the problem *primarily* one of engineering, or
> *primarily* one of mathematics, or *primarily* one of neuroscience, or
> *primarily* one of cognitive psychology, etc.

Remember the illustration on the cover of "Godel, Escher, Bach"? The
"trip-let", the solid object that cast three different letters as shadows?

When I look at the Bayesian Probability Theorem, I also see the social
process of science, the emotional frustration a rationalist experiences on
arguing with a Bible-thumper, an elegant but nonhuman way to implement
negative and positive reinforcement in goal systems, and the glue that binds
a model to external reality. For me it all goes under the name of "Bayesian
Probability Theorem" because that theorem is the purest expression of the
underlying essence that gives rise to all these effects.

Surely you've already had some experience with this in Webmind/Novamente?
I.e., you argued over whether it was engineering or math or neuroscience or
cognitive psychology, and then you implemented it and found out it was all
the same thing? I'm not necessarily saying that I would expect this to be a
routine experience, but weren't there at least one or two memorable

And no replies saying, "Yes, but there were incredible engineering
difficulties in implementation, and after we found the deep underlying
principle we had to specialize and differentiate it again in order to make
it work in different domains." I know about that part.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT