Artificial intelligence (AI) gurus found the Turing test intriguing, and in 1964 Joseph Weizenbaum, at the Massachusetts Institute of Technology, wrote Eliza - a program that behaved like a psychotherapist (one using non-directive therapy, mostly echoing what people say and picking up on mentions of parents).
People who didn't know Eliza was a program believed it was a person - a fact that Weizenbaum found worrying, though possibly it tells us more about psychotherapy than AI. Eliza was a milestone. But the question of whether a machine could persuade a sceptical observer it was human remained open.
In 1991 Hugh Loebner, a maker of disco-dancing floors, began sponsoring an annual contest with a $10,000 prize for anyone whose machine passed a limited Turing test: it would have to be proficient in only one subject (such as Star Trek, or cave-diving). A panel of judges talks with a series of pairs of terminals - one connected to a machine, the other to a person - and rates each on their level of "human-ness." Usually, the machines could be spotted easily - they didn't make spelling mistakes.
This year, the contest (held in April) was won by a program written by Jason Hutchens, of the Centre for Intelligent Information Processing Systems at the University of Western Australia. His HeX program "pretended" to be Australian. (Please, no snide comments about that making it easier.) Hutchens's Web page about it all can be found at .
Objectively, Hutchens's program seems rather capable. A sample dialogue:
PROGRAM: Well, should my age really matter? Just for the record I am 24. How old are you, then?
JUDGE: I'm 21.
PROGRAM: If you are 21, it's your business, mate! Would you like to hear a joke? Knock, knock ...
JUDGE: I don't much care for knock-knock jokes, but go on ...
PROGRAM: No, let me tell you one first. Knock, knock.
JUDGE: who's there?
JUDGE: Cindy who?
PROGRAM: Cindy next judge in, please!
CHARLES ARTHURReuse content