In addition to being one of the fathers of computer science, Alan Turing postulated a very simple test for when computers move beyond calculations and start engaging in what we might consider thought. For Turing, the ultimate test was whether a person, engaged in a text-based conversation with a machine, would believe that it was conversing with another human.
Each year, the University of Reading hosts a competition where software is put to this test, with the winner taking home the Loebner Prize in Artificial Intelligence. This year’s winner, called Elbot, came within one judge of passing the test, but its success may be less important than the underlying technology: Elbot is the product of a company that promises its software can help companies take the requirement for humans out of live chats and e-mail.
Over a dozen competitors took part in this year’s contest, including older favorites like ALICE and Jabberwacky, both of which wound up among the six finalists. Elbot took home the Loebner Prize by convincing three of a dozen judges that it was human; it and most of the rest of the bots received high scores for portions of their conversation.
Typically, fooling 30 percent of people is considered a pass on the Turing Test, so this suggests that the combination of fast processors and sophisticated software is on the verge of passing the test.
Looks to me like Kurzweil will be cashing in on his Turing bet with Mitch Kapor soon.