Re: Hardware Progress: $319/GF

From: James Rogers (jamesr@best.com)
Date: Fri May 31 2002 - 14:19:48 MDT


I think your premise is incorrect, though I agree that the amount of
hardware required is probably overstated. The flaw I see is that you are
assuming that all neural functions map to silicon with the roughly the
same efficiency. This is almost certainly not the case. Audio and
visual processing are the same class of problem i.e. they are both very
similar types of signal processing functionality, and I would therefore
expect them to have roughly the same relative relationship to silicon in
terms of processing efficiency. Something like language functionality,
which does not appear to be terribly analogous to signal processing, may
have a substantially different relative processing efficiency for better
or worse.

My intuition would be that signal processing type functionality is
actually one of the least efficient processes in the brain with relation
to its silicon counterpart, and would therefore stand to gain the most
by mapping it into silicon.

-James Rogers
 jamesr@best.com

--------------------------------------------------------
On Fri, 2002-05-31 at 12:31, Mike & Donna Deering wrote:
The experts have grossly over estimated the computational requirements
for GAI. We have a way of calculating it precisely. Kurzweil says we
have simulated the human auditory functions and we are very close to a
complete simulation of the human visual function. And we have a good
idea of the neuroanatomy of these systems. Simply take how much
hardware does it take to simulate these functions, how much of the brain
do these functions comprise, extrapolate how much hardware it takes to
simulate the whole brain.



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT