grady@apple.UUCP (10/05/87)
O.K. So the Turing test can tell you when AI's are indistinguishable from humans. But what we'll need in the 90's is a way to tell when AI's are uncontrovertedly superior to humans -- this test, I propose to call the CyberTest. The criterion the CyberTest uses is simple: when an AI is so persuasive, intuitive, and eloquent that EVERY person is convinced of its superiority, then, in fact, it will be. This is a good test because the sufficiency of it cannot be questioned. (Since everyone is convinced, no one would think to ask the question.) The hard part, of course, is convincing everybody of the AI's preeminence. Isn't there always at least one old redneck who refuses to accept anything? Well, thats why I call the criterion the "CyberTest," because in that case -- when no one's looking -- the AI ensures unanimity by excising the holdout. -- Grady Ward
cmaag@csd4.milw.wisc.edu.UUCP (10/07/87)
In article <6413@apple.UUCP> grady@apple.UUCP (Grady Ward) writes: > > O.K. So the Turing test can tell you when AI's are >indistinguishable from humans. Being fairly new to the genre (I've only read Gibson up to this point), could someone expand a little on the "Turing Test"? Was it actually devised by Turing? Please e-mail if you don't feel it is of general interest. Chris. ======================================================================= Path: uwmcsd1!csd4.milw.wisc.edu!cmaag From: cmaag@csd4.milw.wisc.edu bitnet: cmaag%csd4.milw.wisc.edu@wiscvm.bitnet {seismo|nike|ucbvax|harvard|rutgers!ihnp4}!uwvax!uwmcsd1!uwmcsd4!cmaag =======================================================================
RLWALD@pucc.UUCP (10/07/87)
In article <6413@apple.UUCP>, grady@apple.UUCP (Grady Ward) writes: > The criterion the CyberTest uses is simple: when an AI is so >persuasive, intuitive, and eloquent that EVERY person is convinced of >its superiority, then, in fact, it will be. This is a good test because the >sufficiency of it cannot be questioned. And then we can make it watch while we delete it. CyberSnuff? :-) -Nexus -Rob Wald Bitnet: RLWALD@PUCC.BITNET Uucp: {ihnp4|allegra}!psuvax1!PUCC.BITNET!RLWALD Arpa: RLWALD@PUCC.Princeton.Edu "Why are they all trying to kill me?" "They don't realize that you're already dead." -The Prisoner
vnend@ukecc.UUCP (10/08/87)
In article <6413@apple.UUCP> grady@apple.UUCP (Grady Ward) writes: > O.K. So the Turing test can tell you when AI's are >indistinguishable from humans. But what we'll need in the 90's is a >way to tell when AI's are uncontrovertedly superior to humans -- >this test, I propose to call the CyberTest. > The criterion the CyberTest uses is simple: when an AI is so >persuasive, intuitive, and eloquent that EVERY person is convinced of >its superiority, then, in fact, it will be. This is a good test because the >sufficiency of it cannot be questioned. (Since everyone is convinced, >no one would think to ask the question.) >-- Grady Ward But all (ha!) that this requires is an expert system for convincing humans, a talking head that scores high on retoric. The only area that it would really be superior in would be convincing people (not easy, but...) I'm afraid that the Cybertest is going to require a little more than this, as some of us will require more than persuation, intuition and eloquence to say that someon or thing is superior. It's gonna take *proof*! Nice idea though. How would you have it go about proving its superiority as opposed to just talking about it? -- Later y'all, Vnend Ignorance is the Mother of Adventure. cbosgd!ukma!ukecc!vnend; vnend@engr.uky.edu; vnend%ukecc.uucp@ukma.BITNET Also: cn0001dj@ukcc.BITNET, Compuserve 73277,1513 and VNEND on GEnie I may be smart, but I can lift heavy things.
grady@apple.UUCP (10/09/87)
You object to the CyberTest on the basis that persuasion and proof are distinct processes. Presumably, you believe that persuasion and eloquence are simple sophistry, both easily detected and dismissed, while "proof" is a much stronger test against error. May I ask how *you* were persuaded that "proof" is convincing? By an eloquent geometry teacher perhaps? Be warned that that teacher may have been one of my CyberTest agents practicing its ScientificMode persuasion heuristic. Successfully, apparently. I suspect that the superior AI reads both Thomas Kuhn and Karl Popper and knows how to recast argument to suit both the longshoreman as well as the intellectual skeptic, and that if it were still unable to convince you, given *whatever* reasonable or unreasonable canon of proof you employ, then it would not yet be worthy. On the other hand, you may be the last holdout, in which case. . . :-) Grady Ward
samlb@well.UUCP (10/09/87)
In article <3099@uwmcsd1.UUCP> cmaag@csd4.milw.wisc.edu.UUCP (Christopher N Maag) writes: >could someone expand a little on the "Turing Test"? Was it actually >devised by Turing? As far as I know, there is no such thing as a formal "Turing Test" with criteria -- the idea came from a remark of Turing's that a machine would have reached "intelligence" or "sentience" if, when put in a room with a teletype machine (the ultimate in I/O in those days), you couldn't tell whether the entity on the other end of the circuit was a human or a machine. In Gibson, the Turing Test seems to involve finding out whether the AI obeys Asimov's Laws of Robotics or not -- i.e. is ultimately controllable by human beings, rather than self-determining, capricious, and ruthless (like _real_ human beings). The "Turing Commission" people (with some justification) seek to pull the plug on dangerous machine intelligences . . . { Enter asbestos suit } -- Sam'l Bassett, Writer/Editor/Consultant -- ideas & opinions mine! 34 Oakland Ave., San Anselmo CA 94960; (415) 454-7282 UUCP: {...known world...}!hplabs OR ptsfa OR lll-crg!well!samlb; Compuserve: 71735,1776; WU Easylink ESL 6284-3034; MCI SBassett
jr@lf-server-2.bbn.com.UUCP (10/09/87)
In article <1639@ukecc.engr.uky.edu> vnend@engr.uky.edu (D. V. W. James) writes: >In article <6413@apple.UUCP> grady@apple.UUCP (Grady Ward) writes: >> O.K. So the Turing test can tell you when AI's are >>indistinguishable from humans. >>-- Grady Ward > > But all (ha!) that this requires is an expert system for convincing >humans, a talking head that scores high on retoric. The only area that >it would really be superior in would be convincing people (not easy, but...) >I'm afraid that the Cybertest is going to require a little more than this, >as some of us will require more than persuation, intuition and eloquence to >say that someon or thing is superior. It's gonna take *proof*! Now come on. Just look at Reagan's success in selling ridiculous ideas. Couple the cybertron to the right folksy tone and style, and you'll convince the world (well, at least the U.S. population). Garry Trudeau is onto something. >-- >Later y'all, Vnend Ignorance is the Mother of Adventure. -- /jr jr@bbn.com or jr@bbn.uucp
steve@nuchat.UUCP (10/17/87)
In article <3099@uwmcsd1.UUCP>, cmaag@csd4.milw.wisc.edu (Christopher N Maag): > In article <6413@apple.UUCP> grady@apple.UUCP (Grady Ward) writes: > > O.K. So the Turing test can tell you when AI's are > >indistinguishable from humans. > Being fairly new to the genre (I've only read Gibson up to this point), > could someone expand a little on the "Turing Test"? Was it actually > devised by Turing? Please e-mail if you don't feel it is of general > interest. Alan Turing proposed the test which is now named for him in the context of a debate in the mathematics community over just what artificial intelligence _meant_. He did not intend it as a test for AI but more as a definition of it. I don't have that lecture with me, so I paraphrase: The sceptic sits before a teleprinter. The testor is free to attach the printer to a similar device with a human operator or to the mechanism under test. If the sceptic cannot determine which is the machine and which is the human, the machine can be said to be intelligent. It was at one time said that intelligence was the ability to make choices. As soon as digital systems started makeing choices, the definition was narrowed. Each time the definition is met by a computer intelligence is redifined. I wish I could remember some of the other difinitions, but we've been through 3 or 4 widely accepted definitions. Turing's test will probably not be redefined when it is successfully met, but unlike Alan noone today expects that to happen any time soon. _The_Enigma_, a biography of Alan Turing by (first name?) Hodges is recommended. -- Steve Nuchia | [...] but the machine would probably be allowed no mercy. uunet!nuchat!steve | In other words then, if a machine is expected to be (713) 334 6720 | infallible, it cannot be intelligent. - Alan Turing, 1947
brad@ut-sally.UUCP (10/19/87)
In article <407@nuchat.UUCP> steve@nuchat.UUCP (Steve Nuchia) writes: >In article <3099@uwmcsd1.UUCP>, cmaag@csd4.milw.wisc.edu (Christopher N Maag): >> Being fairly new to the genre (I've only read Gibson up to this point), >> could someone expand a little on the "Turing Test"? For your edification (quoted without permission, but for what I hope constitutes "fair use"): " I propose to consider the queston 'Can machines think?' .... Instead of attempting ... a definition, I shall replace the question by another which is closely related to it and is expressed in relatively unambiguous words. The new form of the problem can be described in terms of a game which we call the 'imitation game.' It is played with three people, a man (A), a woman (B), and an interrogator (C) who may be of either sex. The interrogator stays in a room aprt from the other two. The object of the game for the interregator is to determine which of the other two is the man and which is the woman. He knows them by labels X and Y.... In order that tones of voice may not help the interrogator the answers should be written, or better still, typewritten. The ideal arrangement is to have a teleprinter communicating between the two rooms. We now ask the question, 'What will happen when a machine take the part of A in the game?' Will the interrogator decide wrongly as often as when the game is played between a man and a woman?' These questions replace our original, 'Can machines think?'" "Computng Machinery and Intellience," _Mind_, Vol. LIX, No. 286, (1950). Reprinted by permission in _Minds and Machines_ ed. Alan Ross Anderson, Prentice Hall, 1964. >It was at one time said that intelligence was the ability to make >choices. I think part of the point of the imitation game is the demonstration of some ability to consider someone else's position for the purposes of discourse and to respond in a way that shows that consideration (regardless of whether that person's position was taken into account for reasons of deceit or otherwise). > >_The_Enigma_, a biography of Alan Turing by (first name?) Hodges >is recommended. Seconded. >-- >Steve Nuchia | [...] but the machine would probably be allowed no mercy. Brad Blumenthal {ihnp4,harvard}!ut-sally!brad || brad@sally.utexas.edu