The Loebner Prize

News

Not a member yet? Register for full benefits!

The Loebner Prize

The Loebner Prize for artificial intelligence is an annual contest, designed to put artificially intelligent chat programs through their paces. It uses the Turing Test for AI, as conceptualised by Alan Turing in the 1950 paper "Computing machinery and intelligence". In the decades since then, the Loebner prize is the first attempt to take the Turing test seriously, as a benchmark for AI.

Turing Test

The Turing Test itself is simple in concept. The original test involved three rooms, three parties, two humans, one an artificial intelligence. Originally, the parties communicated by teletype; nowadays they use chat software and a local area network.

The judge, or panel of judges sit in a room away from the two participants, able to communicate with them, only via a textual interface. Of the two participants, elsewhere in the building, one is human, the other an artificial intelligence persona. Both attempt to hold conversations with the judge(s), via the textual interface, attempting to convince them they are talking to a human through the link. The point of the test, is to leave the judge(s) in serious doubt as to which is the human, and which is the AI.

The AI passes the test, if the judge(s) cannot render a verdict absolutely, as to which is the human, and which the AI.

Hugh Loebner started the Loebner prize itself, in 1990. Loebner agreed with The Cambridge Centre for Behavioural Studies to underwrite a contest on the Turing test. With a grand prize of $100,000 usd and an engraved solid gold medal for any AI who was able to convince a panel of three judges, the Loebner prize has understandably, acted like a rocket to the chat-AI community.

Each year an annual prize of $2000 and a bronze medal is awarded to the most human-like computer entry for that year, keeping a steady flow of applicants.

As of writing, the 2006 Loebner prize has recently passed, and no entry won or came close to the gold medal.