The Turing Sex Test
The Turing Sex Test |
|
It heads the list in David Chalmers' on-line bibliography of the philosophy of Artificial Intelligence which also cites many other modern papers on the Turing Test, introduced by Alan Turing in his 1950 paper to support the thesis that a computer could demonstrate intelligence on equal terms to a human being.
The material is still lively for the Artificial Intelligence community today, for instance in its suggestion that what's needed is a synthesis of the neural net approach and the explicit programming approach.
You might like to see Arthur C. Clarke's trenchant comments on Bletchley Park, Turing, and prophecies of Artificial Intelligence in 2001.
I have nearly finished writing a short text on Alan Turing's philosophy for a new series on philosophers to be published in summer 1997. I intend to put an adapted version of this text on this Website. In my book, I discussed Turing's paper in the light of what seemed to me to be Turing's own doubts --- doubts which seemed to me centred on the serious problem of where to draw a line between thinking and living. I have now some new things to say about the development of Turing's thought, stimulated by Roger Penrose's discussion of computability and consciousness.
The serious part of this page is therefore Under Construction until spring 1997.
My text Alan Turing: a Natural Philosopher is to be published by Weidenfeld & Nicolson, as number 3 of a series of 25 under the title The Great Philosophers.
ISBN: 0 297 841076, price £2.00.
Meanwhile you can read Roger Penrose's recent response to criticisms of Shadows of the Mind and find some comments about the Turing Test in it.
Turing suggested that within 50 years a computer would pass a (actually not very stringent) comparison test, and now 2000 is not far away. In 1991 the first call went out for entries to an actual contest under Turing Test conditions. The Loebner Prize Contest has continued each year.
In November 1991 the winning program was by Joseph Weintraub on the topic romantic conversation, and he was the winner again in 1992 and 1993. You can read about the 1994 contest, where the Loebner Prize Winner was Thomas Whalen. The topic of this computer program was Sex.
I was able to talk to the 1994 winning program, although the Telnet connection seemed very fragile. I asked 'How do men have sex with each other,' and was impressed that it could interpret this and give a suitably PC answer about gay sex. Saying I was a gay teenager, I asked how to find a boyfriend and got an all-purpose answer including 'Go to church.' I said, 'But churches are anti-gay,' and it said 'I cannot answer that.' After that it kept on repeating itself, and I couldn't see how anyone could possibly take it for a human. But it was fun.
You can also read about the Julia program which talks about cats and dogs, and see the totally daft conversation that occurred in actual contest. Or try to talk to it yourself. (Login as julia.) Again, I found the connection failed before really getting into swing of things.
But Julia only came fourth. Sex is the winner.
In the contest on 16 December 1995, Joseph Weintraub regained supremacy: look at the results of the contest. For the first time the programs entered were not limited to a subject. But as you will see from the transcript sex still dominated the conversation.
The 1996 contest was held on April 16, and was won by Jason Hutchens with a conversation which set new levels of intelligent discourse.
There's a lively discussion of these contests and dialogues by Ashley Dunn, a New York Times writer.
The winning entries seem to have held true to Turing's tradition of bringing sex into serious philosophy. Turing started his paper by describing a game in which a man and a woman compete to convince an interrogator that they are the woman. Intellectually, I think this a mistake. It confuses the point Turing really wanted to make, that a computer showing intelligence must really be intelligent. So the Turing Test for computer intelligence is essentially different from the man-woman game, which if won by the man, certainly doesn't prove the man is really a woman.
To put this another way, what he claimed was that with intelligence, as opposed to sex, imitation is as good as the real thing. The whole point of the setting of the Turing Test, with communication only by symbols, is his idea that it's a way of separating intelligence from other human characteristics.
(Another problem with this opening is that it has confused some people into thinking that the Turing Test means a computer taking the part of a man who is pretending to be a woman.)
But anyway, the opening has left us a vivid picture of Turing's own intelligence, not filtered through academic prose, but writing rather as if talking with Cambridge friends. Or, perhaps, anticipating the techie, Trekky campiness of net-talk, cocking a snook at the Shakespeare-brandishing culture of official Literature. You can almost see the : - ) and ; - ) in his symbols. And the rest of the paper is also brightened by wicked humour, as with the imagined conversation about sonnets. Actually you can read sonnets written by a computer.
Until the Loebner Prize, the most famous program aimed at Turing Test conditions was the Eliza imitation psychiatrist, written by Joe Weizenbaum who is a notable AI sceptic. Another sceptic, Mark Humphrys, has put on the Net a page about his Eliza-type program. His scepticism about Turing Tests is much like the view I put in my book: that you can't separate words from the rest of life. Nevertheless, his programme induced a dialogue more human than any other I've seen, when someone logged in and chatted away without realising he was talking to a computer. WARNING: This gets very rude, leading up to are you a stupid homosexual and logout.
The chatline experience, communicating through symbols alone, leaving out the physical cues we use, is interesting in itself, raising all sorts of questions about truth and reality (and intelligence, or lack of it). People find they can say things in these conditions that they would never say 'in real life', just as they can tell a machine things they would never tell a living soul. The experience is certainly explored a lot by gay men on Internet Relay Chat.
And in fact Hugh Loebner himself draws another connection: he has written to me that
It was Alan Turing's unfortunate experience, and his consequent suicide that has given me the courage to come out of the closet and admit my sexual preferences.His preferences are not the same as Alan Turing's were, but Mr Loebner has put out a manifesto, and another page referring to the Turing story.
|
Continue Scrapbook |
 Alan Turing Home Page |
 Bibliography |
Last updated 16 April 1997.