Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!mit-eddie!media-lab!minsky
From: minsky@media-lab.MEDIA.MIT.EDU (Marvin Minsky)
Newsgroups: comp.ai.philosophy
Subject: Re: Help!
Message-ID: <4168@media-lab.MEDIA.MIT.EDU>
Date: 25 Nov 90 23:03:28 GMT
References: <1990Nov25.223132.24431@nntp-server.caltech.edu>
Reply-To: minsky@media-lab.media.mit.edu (Marvin Minsky)
Organization: MIT Media Lab, Cambridge MA
Lines: 32

In article <1990Nov25.223132.24431@nntp-server.caltech.edu> morphy@nntp-server.caltech.edu (Jones Maxime Murphy) writes:
>Can someone tell me if(and where) there are archives of discussions that have
>taken place in this group on Godel's icompleteness theorem and it's impact on
>AI?
>Jones

Interesting subject.  But has anyone really got anything useful out of
it?  It seems to me that all the discussions have been irrelevant
because of two peculiar assumptions.

1. There has been a general assumption that human reasoning is logical
-- and hence is limited by the limitations implied by Godel's theorem.
This is strange because people do not normally assert proofs.  They
make assertions.  Nor do they claim, normally, to make these on the
basis of a completely consistent logical or formal system.  Now you
might argue that whatever the brain is, it might well be approximated
adequately well by some formal system, hence is subject to those
limitations.  But Godel's theorem doesn't apply to all formal systems,
only consistent ones.  Sometimes, a person may detect an inconsistency
and withdraw an assertion, but that's beside the point, because there
is no reason to suppose this can be always done, etc.

2. As for Godel himself, he did claim to have a peculiar nonlogical
ability to sense mathematical truth.  Heaven knows, he didn't make as
many mistakes as some people -- but this kind of mathematical ESP
should be discussed in sci.skeptic or psi.(para)psychology, shouldn't
it.

In other words, none of this makes sense to me.  I see Godel's theorem
as asserting that consistency is incompatible with heuristics, in the
sense that there is no way to ensure only true assertions in suitably rich
systems.  Very sad, but has nothing to do with our rich, heuristic,
fallible brains.