Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!apple!usc!sdd.hp.com!caen!sol.ctr.columbia.edu!src.honeywell.com!msi.umn.edu!cs.umn.edu!thornley From: thornley@cs.umn.edu (David H. Thornley) Newsgroups: comp.ai.philosophy Subject: Re: Searle's Chinese Room Keywords: Strong AI, Turing test Message-ID: <1990Nov30.232758.19231@cs.umn.edu> Date: 30 Nov 90 23:27:58 GMT References: <16197@mentor.cc.purdue.edu> <3952@media-lab.MEDIA.MIT.EDU> <10297@jpl-devvax.JPL.NASA.GOV> <1990Nov16.171041.14144@ncsuvx.ncsu.edu> <1990Nov19.192555.29337@cs.umn.edu> <1990Nov21.181445.11552@n Organization: University of Minnesota, Minneapolis - CSCI Dept. Lines: 51 In article <1990Nov27.231501.1621@ncsuvx.ncsu.edu> fostel@eos.ncsu.edu (Gary Fostel) writes: > >Is the Turing Test well defined? Thornley, in answer to some of my >objections to serious use of the Turing test was that the person >applying the test needed to: > > Use a little common sense. > >This is delightful. Is "common sense" related to intelligence? Well, >if so than the key to Thornley's defence of the Turing test might be >that the tester needs to be intelligent. How intelligent? I'm reminded >of the old saw: > > Common sense is neither common nor sensible. > >My criticism of the Turing test is directed at it's use as an operational >definition of intelligence. It ends up being circular since it can not >be used without slecting a judge to make the decision and the judge must >be intelligent themselves. > >A valid SCIENTIFIC test can not rely on the judegement of the experimenter. True. So what's wrong with Turing's suggestion of "an average interrogator will not have more than 70 percent chance of making the right identification after five minutes of questioning"? This is a specific proposal, if you will grant a little leeway on the word "average." (In psychology departments, it seems to be the average introductory psych student.) When I read a paper claiming that a system has passed the Turing test, I want to read how the interrogators were selected, what sort of relevant backgrounds they had, and a statistical breakdown of success rates. I also want to see confidence ratings, and various other assorted details. I then intend to study the article carefully. I certainly would not take a claim seriously without such background. In particular, if someone were to claim that their husband/wife could not tell the difference, and that is how the machine passed the test, I would not be impressed. In the meantime, please feel free to propose any selection process you please for the interrogators (I insist on multiple interrogators). I don't think you will, in good faith, come up with one so weird that I will refuse to consider it a Turing test. However, I claim that it makes sense to speak of the test in general terms, with the understanding that we will nail it down when necessary. Describing something as "passing the Turing test" is something like referring to "the capacity of short-term memory" in that we don't quite define it as we say it, but we can use it in the lab just fine. DHT