By Duncan Geere, Wired UK

Are you human or a machine? Prove it, by passing the Turing Test a test of the ability of a machine to exhibit intelligent behavior.

Turings original imitation game had nothing to do with artificial intelligence. It was a simple party game with three players a man, a woman, and a judge of either sex. The judge sits in a room apart from the man and woman, and has to guess which is which from nothing but written communication.

The standard interpretation of the Turing Test today, however, replaces one of the participants with a machine which has to imitate intelligence. In this case, the judge has to decide which of the pair is the person, and which is the machine. The computer is successful, and passes the test, if as Turing puts it the interrogator decide[s] wrongly as often when the game is played [with the computer] as he does when the game is played between a man and a woman.

Theres a bit of debate over whether the computer and the person are both supposed to try to trick the interrogator into making an incorrect decision or not. In the original imitation game, one of the pair tries to trick the judge while the other does not, meaning that both will be pretending to be the same gender. The common interpretation of the Turing Test today, however, is one of imitation rather than trickery.

One aspect of the test that Turing never made clear is whether the judge should know whether theres a computer in play at all. Serious attempts at passing the test would almost certainly require a double-blind control, where the judge repeats the experiment multiple times sometimes with a pair of humans, sometimes with the human and the computer, and sometimes with two computers.

While the Turing Test has been lauded for its simplicity and its ability to test across a wide range of intellectual tasks (natural language, reason, knowledge and learning can all be tested), it has also been criticized for a number of reasons. Firstly, the Turing Test doesnt directly test intelligence. Instead it merely tests how much a computer can behave like a human being.

Thats an important distinction because some human behavior is unintelligent, and there are plenty of intelligent behaviors that humans dont do. If, for example, a computer solved a mathematical problem that humans dont have the intellectual capability to do, then it wouldnt make it unintelligent but it would make it fail the Turing Test.

A second issue is that simulated intelligence isnt the same thing as real intelligence. A machine that can pass the Turing Test could just be following a large list of mechanical rules. As such, the Turing Test doesnt test whether a machine can genuinely think. The counter-argument to that is that humans could well just be following a large list of mechanical rules, and then youre deep into philosophy of consciousness and intentionality.

See the original post: