Newsgroups: comp.ai,comp.ai.philosophy
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: How is AI going?
Message-ID: <1996Feb19.034134.18851@media.mit.edu>
Sender: news@media.mit.edu (USENET News System)
Organization: MIT Media Lab
References: <824619785.278snx@ocean.southern.co.nz> <8j68d4kzj0.fsf@lethe.nynexst.com>
Date: Mon, 19 Feb 1996 03:41:34 GMT
Lines: 52
Xref: glinda.oz.cs.cmu.edu comp.ai:37144 comp.ai.philosophy:37944

In comp.ai article <8j68d4kzj0.fsf@lethe.nynexst.com>,
Manfred Marriott (manfred@ocean.southern.co.nz) wrote:

>: I've been reading comp.ai.* for a while now, but I've yet to see 
>: information suitable for the layman about what AI will achieve in the 
>: future, and more importantly, when.

fawcett@nynexst.com (Tom Fawcett) replied:

>I suggest you dig up some of the comical predictions made back
>in the 60s and 70s about what AI will achieve in the future, and when.
>They're very entertaining.  Do you believe we're any better at
>predicting the future now?

This raises an interesting historical question.

I agree that we can't predict the future very well because in some
parts of AI we still don't know how hard the problems are.  However,
I'd like to see your catalog, because I have a feeling that the
oft-cited allegation that AI researchers produce a great deal of
'hype' might actually be a myth propagated by the vitalists.  Possibly
you, yourself may have become pawn of that still-influential
philosophy.

More specifically, it seems to me that the skeptics mostly repeat the
second-hand misquotations of reporters. To be sure, there is the
well-known statement in which H. Simon once predicted a world-champion
chess player in a decade.  He was off by a factor of 4 or so, but
surely historians will agree that he was more right than wrong, since
at the same time Dreyfus was publishing claims that machines would
never play even good chess.  (In the same paper, Simon also predicted
that AI programs would produce difficult mathematical proofs within a
decade, and that turned out right with a smaller time-error.  In fact,
MACSYMA was solving century-old symbolic problems only a half-decade
later.)

A second source of complaint has been from frustrated investors.  It's
true that most "AI companies," including those that built pioneering
special hardware, eventually failed on Wall Street.  However, this was
not due to the failure, for example, of the expert-system crowd.
INstead, it turned out that the expert systems (and other AI
technologies) rapidly spread beyond the control of those companies, so
everyone made mony on them except the annoyed early investors.  There
has in fact been no "AI winter" except for those first investors.

So lets see your own collection of first-hand "comical predictions" by
other established researchers.

Of course, the list you show us should not include "machines will
someday think" predictions, because those cannot be falsified.  Nor
should it include the enthusiastic predictions of peripheral groupies:
all fields have 
