A man before his time, Alan Mathison Turing is arguably one of the most recognized mathematicians of the 20th century. In 1950, he published a paper that, to this day, sets the standard for artificial intelligence. He proposed a way to determine if a machine has intelligence, and this is now called the Turing Test.
In his 41-year life, Turing accomplished a great deal as a mathematician. Bordering on many philosophical issues, his work is recounted in this online biography (1
). A very good introduction to the Turing Test is given on this site (2
). The text of the original paper written by Turing is provided, as well as an original take on its implications. Another paper about the Turing Test can be downloaded here (3
), but it looks at the test from 50 years after it was first published. The author gives an interesting retrospective on what has transpired since Turing formed his ideas and sets the stage for future work.
The Loebner Prize is an annual competition that implements the Turing Test. It involves a panel of judges who question an entity over a computer terminal. The entity can be either human or a computer program, and it is up to the judges to decide who is not human. The program that gives the most human-like responses wins the contest. If a program can ever be indistinguishable from a human and manages to trick the judges, a grand prize will be awarded (although this has not yet happened). To read the rules of the competition, visit the Web site of the 2002 Loebner Prize (4
), which was held on October 12, 2002. The contest's winning program was dubbed Ella, and an online version of the program can be tried at this site (5
). Ella's responses are usually humorous, and it is surprising how realistic they are. A similar interactive utility is called Mr. Mind (6
), but it reverses the roles. The user is asked to prove to Mr. Mind that he/she is human. By taking this perspective, it is quickly realized how difficult it must be for a computer to respond like a human. One chapter of a book that is scheduled to be published in 2003 is given on the Mindpixel Web site (7
). It argues that a lifetime of human experiences is necessary for a computer to pass the Turing Test, but this can be approximated by a large collection of submissions contributed by the public over the Internet. This, incidentally, is the goal of the Mindpixel Digital Mind Modeling Project. Natural language processing is one important factor for computers to understand, and glean meaning from, human dialog. A research paper that was recently included in the Journal of Artificial Intelligence Research can be viewed at this site (8