Loebner Prize Chatbots

Artificial Intelligence
Artificial Fiction Brain
Source: Wikipedia Commons. Licensed under Creative Commons Attribution ShareAlike version 2.5.

The Loebner prize in Artificial Intelligence is the first formal implementation of a Turing test with a substantial prize. The ultimate winner of this contest will receive US $100,000 and each year, approximately US $3,000 is given to the creator of the computer program that converses in the most human-like manner.

The Turing test was first described in Alan Turing's 1950 article Computing Machinery and Intelligence. Essentially, human judges engage in written conversation with both human and computer partners through computer terminals. The judges do not know ahead of time which of their partners are human and which are computer programs emulating humans. The computer programs have a goal of convincing the human judges that they are in fact humans, not computer programs. The idea is that human judges are unable to tell the difference between another human and a computer program, the computer program must in fact be, to some degree, exercising intelligent thought.

Although the Turing test is respected as a benchmark of Artificial Intelligence, the Loebner prize itself is somewhat controversial – the acclaimed "father" of artificial intelligence Marvin Minsky believes it to be nothing but an unproductive publicity campaign. This year's contest will be held at Reading University in the United Kingdom on October 12, 2008. You can chat online with some of the six finalists in advance of the contest:

Previous winners of the Loebner prize have been:

Some other memorable chatbots of interest:

  • The classic Rogerian psychotherapist Eliza.
  • Ally

Although not a contestant, another interesting program to chat with is 20Q.net. This program plays a game of 20 (or more) questions with you and is quite often correct at guessing what you are thinking. You can help train it to get better by playing with it.

3 thoughts on “Loebner Prize Chatbots

  1. Interestingly in 1950 Turing had predicted that by 2000 there will be machines that will routinely pass the Turing test! We still seem far from it. But do you think that with clever use of all the data on the web it will be possible to pass the Turing test in the near future?
    None of the systems in the Loebner contest exploit the web very well.
    I would like your comments on my post here:

  2. I can only think that the supply of information that is available on the web would be helpful to training an Artificial Intelligence, but it is definitely not sufficient by itself. I believe the key is to architect a solid model based on cognitive science – the actual programming and training seem secondary in nature.

  3. Check out this Web 2.0 approach to chatbots: http://chatbotgame.com.
    Just as Deep Blue brute-forced it in chess with speed, the idea behind the Chatbot Game is to brute-force it with a huge number of user-submitted Google-like chat rules.

Comments are closed.