Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!yale!mintaka!mit-eddie!media-lab!minsky
From: minsky@media-lab.MEDIA.MIT.EDU (Marvin Minsky)
Newsgroups: comp.ai
Subject: Re: Buzzword
Message-ID: <5491@media-lab.MEDIA.MIT.EDU>
Date: 11 Mar 91 05:47:51 GMT
References: <1991Mar6.192044.8055@csn.org> <3713@intvax.UUCP>
Reply-To: minsky@media-lab.media.mit.edu (Marvin Minsky)
Organization: MIT Media Lab, Cambridge MA
Lines: 19

>From article <1991Mar6.192044.8055@csn.org>, by tesar@tigger.Colorado.EDU (Bruce Tesar):
>>I'm looking for a buzzword that describes problems that at first sound
>>easy to do, but when it actually comes down to writing an AI program, the
>>problem turns out to be very difficult to solve.

I've watched this thread for a while.  In a certain futuristic sense,
all problems so far solved by writing AI programs were very easy -- by
historic standards.  Thus the probllem of indefinite formal
integration was unsolved for some 300 years until decisively solved by
Moses-Engelman-Martin-Risch-Caviness, etc.  

But as I suggested in Society of Mind, the problem is that it is the
"humanly" easy problems that turn out hard -- like learning to talk,
or to see, or to do "simple, commonsense reasoning".  These are the
problems that use lots of brain.

So, seriously, that buzzword could do more harm than good unless it
has some sort of time-scale or dating parameter.  The point is
"difficult for whom" at what state of technology.  Or, to specify to
who the problem looks difficult.