X-Message-Number: 4297
From:  (Brian Wowk)
Newsgroups: sci.cryonics
Subject: Definition of Strong AI
Date: 25 Apr 1995 17:20:44 GMT
Message-ID: <3njb1c$>

	I had interesting discussion with an AI researcher this 
weekend that may shed some light on some of the arguments here
recently.  Most people here seem to believe in "Strong AI" in
the sense that human-level intelligence (and maybe even human
consciousness) can be reduced to low-order symbolic manipulations
(manipulations of binary bits, if you will).  In fact this is
NOT what most AI researchers mean when they say Strong AI.
To them, Strong AI is the belief that human intelligence can
be fully simulated by HIGH-ORDER symbolic manipulation.  In
other words, if you start with a sufficiently large set of
abstract concepts (say, the words of the English language)
and write a set of rules and relations between them in a
computer program, you can eventually end up with human
intelligence.  This is something much more specific than
the generic statement that intelligence can be simulated
by computers.  Strong AI is a statement about *a type
of program* that is sufficient for intelligence.

	I write this because I notice Thomas Donaldson
(an AI researcher himself) recently questioned on CryoNet
whether neural nets are computers, and I want to forestall
the flury of responses to his message.  Even Thomas
would agree that neural nets are computers in the sense that
they perform *low order* symbolic manipulations.  They
do not, however, perform the high order symbolic manipulations
that Strong AI people believe are sufficient for intelligence.

	In fact, I was told that one of the most ardent
critics of Strong AI, Patricia Churchland, has no problem
with the belief that neural nets or their derivatives
may come to emulate human intelligence.  The belief that
intelligence cannot be simulated by ANY kind of symbolic
manipulation (high or low order) appears to be quite
rare.  As far as I know, the only people holding such
beliefs are religious theologians and Roger Penrose.

---Brian Wowk


Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=4297