Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!uhog.mit.edu!news!ml.media.mit.edu!minsky
From: minsky@ml.media.mit.edu (Marvin Minsky)
Subject: Re: Does AI belong in neural nets? (Marvin Minsky quote)
Message-ID: <1996Mar3.025523.17325@media.mit.edu>
Sender: news@media.mit.edu (USENET News System)
Organization: MIT Media Lab
References: <DnLF5n.G7n.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk> <4h80e2$8cc@news.cis.okstate.edu> <313880CC.4267@ccs.neu.edu>
Date: Sun, 3 Mar 1996 02:55:23 GMT
Lines: 48

In article <313880CC.4267@ccs.neu.edu> cloder@ccs.neu.edu writes:
>I'd like to post an quote from a good paper about this (long-running) 
>division between neural networks/natural intelligence and symbolic 
>artificial intelligence by Marvin Minsky. I think it sort of puts things 
>in perspective - Minsky's been around for a long time (he's seen both 
>sides of the neural net vs. symbolic AI war). This is probably a good 
>example of the Hegelian Dialectic at work.

I dunno about Hegel, but it would certainly be a good thing to shake
off this foolish dumb-bell division and consider a more rich
representation of the issue.  I'm sure we all could use better
cognitive architectures, in which some systems "manage" the activities
of others, using knowledge about adapting various strategies for
thinking to particular kinds of problems.

Anyway, polarizations can be productive.  When I finished my thesis on
neural nets at Princeton, someone in the Math department gave it to
von Neumann and asked if it was mathematics.  He answered, "if it
isn't now, it will be someday."

Chan Loder suggests reading

>"Logical vs. Analogical or Symbolic vs. Connectionist or Neat vs. 
>Scruffy"
>       in Artificial Intelligence at MIT., Expanding Frontiers, Patrick 			
>							H. Winston (Ed.),
>       Vol 1, MIT Press, 1990. Reprinted in AI Magazine, 1991 
>
>This article is also available in plaintext format from:
>
>	ftp://ftp.ai.mit.edu:/pub/minsky/SymbolicVs.Connectionist

Also, you might read the short essay called "Causal Diversity Matrix"
in the same ftp folder.  This attempts to say something about when to
use nets and when to use symbols.  In many areas one wants to use
both: consider how much more flexible can be Fuzzy Expert Systems than
simple rule-based ones.  The paper itself is purposely vague, and I'd
like to see ideas about how to add other dimensions, and how to
improve the general approach.

Marvin Minsky

    _____________________________________________________________
  "Don't pay any attention to the critics. Don't even ignore them."
                                            ---------  Sam Goldwyn

