This past weekend Dartmouth hosted the annual Loebner Contest, a competition which challenges computer programmers from around the world to create programs whose communication skills are indistinguishable from those of the humans they emulate.
The contest was part of a conference called "Turing 2000: The Future of the Turing Test" which stemmed from the 1950 predictions of British mathematician and computer science visionary Alan Turing, who said that it was only a matter of time before computers could be programmed to "think" like humans.
Although Turing predicted that in 50 years, "an average interrogator will not have more than a 70 percent chance of making the right identification after five minutes of questioning," humans continued to beat out the computers at this year's conference.
According to Dartmouth professor of philosophy James Moor, who coordinated this year's event, this year's programs have not made any dramatic improvements on past programs.
"I think it will be a long, long time before any computer program succeeds in fooling someone after a five minute conversation. Right now, machines often miss the point of the question and don't respond in a relative way," Moor said.
Moor also said that humans have a broad ability to learn and understand things, whereas current computer programs only give coherent answers in restrictive and formal contexts.
At the conference in Hanover this year, judges attained accuracy levels of 91 percent after five minutes and 93 percent after 15 minutes.
The contest featured six computer entries from around the world, with names like "e-Brain," "ChatRascal" and "Algernon." The winner, named "Alice," was a program designed by Dr. Richard Wallace of San Francisco.
In addition to the six computer entries, the contest also featured four human beings, three of whom were mistaken for a computer by at least one judge.
Questions asked by judges ranged from, "How are you?" to "What is a lie?"
One question that was particularly effective in differentiating computers from humans was, "What does the letter 'M' look like upside down?"
After the conversations, which lasted approximately 15 minutes each, judges were asked to rank the 10 contestants (six computers and four humans) in order from most human to least human, as well as rate them on "responsiveness" and "human quality."
Turing's test set up a text-only dialogue between an interrogator, and either a concealed human or a computer program designed to "think" like a human.
The interrogator poses questions to both the human being and the computer in an effort to differentiate between the two. If they were found to be indistinguishable by means of textual exchange, then Turing surmised that the computer could be labeled "intelligent."
The annual Loebner Contest is the official administration of the Turing Test. Its founder, Hugh Loebner, said he will reward a prize of $25,000 to the first computer that is indistinguishable from a human being after a five minute textual conversation.