From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!usc!orion.oac.uci.edu!cerritos.edu!arizona.edu!NSMA.AriZonA.EdU!bill Fri Jan 31 10:26:38 EST 1992
Article 3228 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!usc!orion.oac.uci.edu!cerritos.edu!arizona.edu!NSMA.AriZonA.EdU!bill
Newsgroups: comp.ai.philosophy
Subject: Turing Test
Message-ID: <1992Jan28.130209.2486@arizona.edu>
>From: bill@NSMA.AriZonA.EdU (Bill Skaggs)
Date: 28 Jan 92 13:02:08 MST
Reply-To: bill@NSMA.AriZonA.EdU (Bill Skaggs)
References: <11927@optima.cs.arizona.edu> <1992Jan27.201926.20498@gpu.utcs.utoronto.ca> 
 <1992Jan27.191241.8139@husc3.harvard.edu> <1992Jan28.183603.21220@gpu.utcs.utoronto.ca>
Distribution: world,local
Organization: Center for Neural Systems, Memory, and Aging
Nntp-Posting-Host: ca3.nsma.arizona.edu
Lines: 21

  Let's try a different angle.

  Turing, when he proposed his test, thought of it essentially
as a benchmark -- an important and useful goal for AI to aim
at.  He considered "thinking" too fuzzy a notion to be a useful
goal, and proposed to replace it with his test.

  Incidentally (I have just been rereading "Computing Machinery
and Intelligence"), as a stunning example of prescience in a
field notable for appallingly bad predictions, Turing predicted
(in 1950) that in fifty years machines would exist that would
be detected (in the imitation game) less than 70% of the time
in a test of five minutes duration.  This looks to be very close
to the level of performance that will actually be achieved by
the year 2000.

  But my real question is:  are there any other good, simple,
straightforward benchmarks for AI?  

	-- Bill



