The Loebner prize of $100,000 to be awarded to the first computer to rigorously pass the Turing test went yet again unclaimed this year.

The Turing test, for those who don't know, is a litmus test for determining whether a machine possesses an intelligence comparable to a human being. It basically goes like this. You set up an interface through which humans can communicate with either a machine or another human being, and make it one that hides whether a person or machine is on the other end (say a keyboard/CRT terminal). In other words, the only possible source of information about whether the communicator on the other end is in the content of the communication itself. Then, human beings are asked to communicate through the interface, and are told that either a human or machine is on the other end, and that their job is to figure out which within a specified time, based soley on conversation they have. If, after the specified time their determinations aren't better than tossing a coin, then, Turing said, for all practical purposes that machine is intelligent in the same way that a human being is, because you can't tell the difference. Note that this doesn't assume the machine is conscious, just intelligent.

Anyway, ever last one of this year's entries was blown out of the water by the expert panel. As Julie Flaherty of the New York Times paraphrased, "They couldn't fool any of the people, any of the time".

Interestingly one of the most common techniques used to trip them up was to tell a joke based on a construction o language. You know, like "A buhddist when up to a hot dog vendor and said, 'Make me one with everything'".

So Buck up Mr Kasparov. Deep Blue and his silicon bretheren are still just Idiot-Savants, good for a challenging game of chess now and then, but pathetically boring dinner companions.