From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!rickert Tue May 12 15:48:39 EDT 1992
Article 5368 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Newsgroups: comp.ai.philosophy
Subject: Re: Physical Symbol Systems Hypothesis
Message-ID: <1992May2.042220.7808@mp.cs.niu.edu>
Date: 2 May 92 04:22:20 GMT
References: <1992May2.031108.7475@beaver.cs.washington.edu>
Organization: Northern Illinois University
Lines: 72

In article <1992May2.031108.7475@beaver.cs.washington.edu> rex@cs.washington.edu (Rex Jakobovits) writes:
>I am looking for arguments against Newell & Simon's Physical Symbol
>Systems Hypothesis (PSSH): "A physical symbol system has the necessary
>and sufficient means for general intelligent action."  Elaine Rich
>declares this to be the underlying assumption "at the heart of
>research in ai".

 You can always try the Chinese Room, although many of us think that is a
bogus argument.

 You raise a very difficult question.  It is a question which is easy
to answer, but one in which the answer is almost certain to be
misinterpreted.

>What are your opinions about this?  Do non-symbolic systems such as
>connectionist nets and Brooksian creatures exhibit intelligence,
>thereby invalidating the PSSH?

  This question gets to the point of the difficulty.  For you declare
connectionist nets to be non-symbolic systems, yet the plain fact is
that almost all connectionism experiments is performed by executing
symbolic programs on normal standard computers.  If it turns out
that connectionist approaches can produce the desired intelligence,
that would surely confirm that a physical symbol system is a
sufficient means.

>                                Is there evidence that the human
>subconscious does not resort to symbolic processing?

  I see substantial evidence of non-symbolic processing by humans.  But then
I am not a psychologist, and it is in the psychologist's domain to
answer that question.  And that is not the same as saying that "the
human subconscious does not resort to symbolic processing."  In my
opinion it is obvious that the subconscious does use symbolic processing,
but not exclusively, and perhaps only incidently.

>                                                      Does this imply
>that the power of symbolic level processing is inherently limited?

  Many of us believe that, given sufficient computational power, symbolic
systems potentially can simulate any process to any desired degree of
accuracy.  I have seen no evidence to challenge this belief.  Thus even if
intelligence has nothing to do with symbol processing there is no reason
to assume symbol processing systems are unable to produce such
intelligence.  It is quite clear, however, that the computational power
required to achieve human level intelligence will be quite high - far
higher than early optimistic views of AI had imagined.

>Can one be skeptical of the PSSH and still believe in AI?

 I think I have already given my answer to this.

 What I haven't commented on is the "necessary" part of PSSH.  I have
indicated why I believe symbol processing is sufficient.  The question
of necessity is much trickier, and it depends on how you define your
terms.

 In my view, a great deal of human intelligence has nothing to do with
symbol processing (although that does not prevent its implementation using
symbolic processing).  For that matter, a great deal of intelligence has
nothing to do with consciousness and thought, either.  The major
mistake of symbolic AI (in my opinion), and perhaps of views of
cognition from other areas also, is to overrate the importance of
symbol processing, consciousness and thought.  This is a natural
mistake to make, since symbol processing, consciousness and thought
are ever present in our conscious thought.

 But, regardless of the relative importance of non-symbolic processing,
it cannot reach the full level of human intelligence until it produces
symbol processing.  So, while I don't see symbol processing as the
key, it must be achieved to implement machine intelligence.  In this
sense it is necessary.


