home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai
- Path: sparky!uunet!gumby!yale!cs.yale.edu!news
- From: blenko-tom@CS.YALE.EDU (Tom M. Blenko)
- Subject: Re: Loebner Turing test for Human like conversation
- Message-ID: <1992Nov6.042521.10406@cs.yale.edu>
- Sender: news@cs.yale.edu (Usenet News)
- Nntp-Posting-Host: wren.systemsz.cs.yale.edu
- Reply-To: blenko-tom@cs.yale.edu
- Organization: Yale University, Department of Computer Science, New Haven, CT
- References: <1992Nov5.231937.29963@CSD-NewsHost.Stanford.EDU>
- Date: Fri, 6 Nov 1992 04:25:21 GMT
- Lines: 31
-
- Matthew L. Ginsberg writes
- | In my (humble?) opinion, the Loebner test is a joke. The whole
- | *point* of the Turing test, and the reason it is interesting, is that
- | the domain of interaction between the human and the machine is not
- | restricted. By adding an artificial restriction, Loebner (or whoever
- | else is responsible) misses the point entirely.
- |
- | I see no reason to believe that success at the Loebner test is in any
- | way related to making progress on AI's real goals. So here we are,
- | continuing AI's established tradition of generating a tremendous
- | amount of publicity while making no real progress. No wonder the
- | funding agencies are sick of us.
-
- Performing well on this test may not be sufficient for
- reaching "AI's real goals" but it is clearly necessary.
- And to the best of my knowledge, no current systems are
- close to satisfying it.
-
- If one wished to encourage work in the field by presenting
- a sequence of challenges of escalating difficulty, this
- would not be my choice of a place to start. Given the amount
- of papers/conferences/journals/funding in the
- natural language field, however, I certainly don't
- think it's either unfair or a joke.
-
- I think the fault of the funding agencies was not in
- failing to choose the right challenges, it was in failing
- to identify any concrete standards at all for measuring
- progress,
-
- Tom
-