The famous "Turing Test" is usually misdescribed. It's not the problem of making a computer program which can, momentarily or in a narrow field of discourse, convince a naïve person into thinking that s/he's interacting with another human being.
The original Alan Turing parable begins with the challenge of distinguishing a male from a female via teletype interaction: a judge has two contestants to correspond with and must try to figure out their sexes via a typed question-and-answer conversation. One of the pair tries to be helpful; the other is deliberately deceptive. After sketching out that situation, only then does Turing pose the next stage: replace one of the pair with a computer. Can a judge reliably identify which contender is a person and which an algorithm?
That in turn suggests a few other Turing-style conundrums:
Douglas Hofstadter told a story about a time when he was invited to test a program called "Nicoli", and did not realize for nearly the entire interview that it was actually three students on a terminal, pretending to be a computer.
Apparently, earlier on, the class as a whole had carried out a similar test, with just one student in the other room. Even though the student didn't even try to pretend he was a computer, only one classmate suspected that they were not talking to a program.
- RadRob
TopicScience - TopicHumor - 2003-03-20
(correlates: ChatTuringTest, Spectrum of Deceit, ExceptionsRule, ...)