It's always going to be easy to trick a computer to reveal their
identity as such until the computer has full sensory apparatus.
Turing proposals were complex, where the test subject should be asked
whether their counterpart was male or female, rather than human or
computer. You count how often the subject guesses right, and then see
if they score the same if you replace the male with a computer program.
See section one of the paper:
http://loebner.net/Prizef/TuringArticle.html
Also, it seems unfair to grade software on their human-like behaviour
without also grading people on their human-like behaviour. I'm not sure
how well I'd do at that. Maybe we should be testing each other for
computer like behaviour :)
alex
Received on Mon Jun 26 2006 - 10:04:25 BST