Newsgroups: comp.ai,comp.ai.nat-lang,comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!newsfeed.internetmci.com!swrinde!cssun.mathcs.emory.edu!news-feed-1.peachnet.edu!paperboy.wellfleet.com!news3.near.net!inmet!news
From: Arnold Chien <chien>
Subject: Re: Wanted--judges for 1996 Loebner (Turing Test) Prize
X-Nntp-Posting-Host: corvette.camb.inmet.com
Content-Type: text/plain; charset=us-ascii
Message-ID: <Dpnpn0.EHK.0.-s@inmet.camb.inmet.com>
Sender: news@inmet.camb.inmet.com (USENET news)
Content-Transfer-Encoding: 7bit
Organization: Intermetrics, Inc.
References: <4kbcim$jjq@HOPPER.ACM.ORG>
Mime-Version: 1.0
Date: Wed, 10 Apr 1996 17:23:23 GMT
X-Mailer: Mozilla 1.1 (X11; U; SunOS 4.1.4 sun4c)
X-Url: news:4kbcim$jjq@HOPPER.ACM.ORG
Lines: 25
Xref: glinda.oz.cs.cmu.edu comp.ai:38159 comp.ai.nat-lang:4775 comp.ai.games:4407

loebner@ACM.ORG (Hugh Loebner) wrote:
>On   5 Apr 1996 in Message-ID <4k49t9$5ip@gondor.sdsu.edu>
>Robert Epstein wrote:
>
>[snip] 
>>We do not believe that
>>Loebner's new contest can legitimately be called a Turing Test,
>
>         Why  not?  Please be specific.  How does it fail?
>

I am not Robert Epstein, but may I offer an answer, perhaps in addition to Mr.
Epstein's?  As I understand it, the Loebner competition has a judge (or judges)
decide among competing programs.  In contrast, Turing envisioned a judge
deciding between a program and a human being hidden out of sight, typing
responses on a terminal.  A program passes the test if after extended
interaction - conversation on any subjects of the judge's choosing - the judge
cannot tell program from human.  That's roughly the idea, give or take a few
details.  Clearly a program might win the Loebner competition, i.e. by
appearing to judges as better than any other program, without being able to
pass the Turing test, i.e. by appearing to judges as indistinguishable from a
person.

Arnold Chien

