Throughout the afternoon, while the typed conversations between programs and judges flickered on large screens in an auditorium at the Boston Computer Museum, a group of scientists and philosophers grappled with the meaning of the event.
"If a computer were to pass an unrestricted Turing Test," said Daniel Dennett, a professor of cognitive studies at Tufts University, "it will raise a number of deep civil rights questions. For example, it would be a crime to unplug such a machine because it would be a person in a legal sense. Turing should be thought of as a pioneer in machine rights."
First prize went to a commercial software program called PC Therapist III, which was designed to engage its questioner in a whimsical conversation. The program was written by Joseph Weintraub, the president of Thinking Software Inc., a small artificial intelligence software company in Woodside, Queens.
At one point, another program suggested to a judge: "Perhaps you're not getting enough affection from your partner in the relationship."
"What are the key elements that are important in relationships in order toprevent conflict or problems?" the judge responded. Thinks It Doesn't Think. . .
"I think you don't think I think," answered the terminal.
"What kind of a response to a question is that?" asked the judge. "Nextquestion: Do you think romantic relationships are unrealistic, doomed from the start?"
"I do tend to walk around with my head in the clouds," the terminal said.
"Are you a romantic?" the judge queried.
"I take in whatever you ask," came the reply.
"Have you experienced problems in romantic relationships?"
"I try to be agreeable," responded the terminal.
Many of the watching experts, including scientists, philosophers and psychologists who assembled here, said they were surprised by the outcome. Computer experts said they could easily tell the human from the computer respondents. Unlike the untrained observers, the experts spotted the "common sense" mistakes that were an immediate giveaway, as when the computer would respond to a question other than the one asked.
There were several other confusing results. One of the human confederates, Cynthia Clay, who was judged "most human of all contestants," was nevertheless judged to be a computer by two judges.
But the deeper question of whether computers will ever be able to convincingly mimic human beings and what it will mean for the human spirit if they can, remained tantalizingly distant.
"These were real judges," said Oliver Strimpel, executive director of the Boston Computer Museum, one of the sponsors. "Maybe the Turing Test isn't as difficult as we first thought." Hard to Tell Difference
Mr. Turing believed that by the end of the century it would be possible to program computers to play an "imitation game" well enough so that an average human judge would not be able to determine if a conversation typed at a computer terminal was with a human "confederate" or a computer.
That assumption has become the subject of a bitter debate between scientists and philosophers who have taken sides on the question over whether the human mind will ever be reduced to a set of computer programs.
The first modern version of the Turing Test, made possible by the establishment of a $100,000 prize by the philanthropist Hugh Loebner and with the support of the National Science Foundation and the Alfred P. Sloane Foundation, was not the true test that Mr. Turing envisioned. Because of limitations in the programs' abilities, each was confined to discussing a narrow topic, such as women's clothing, romantic relationships or Burgundy wine. For winning the first contest, Mr. Weintraub was awarded $1,500.
"We're witnessing history today," said Keay Dewdney, a computer scientist at the University of Western Ontario and author of the Computer Recreations section in Scientific American. At the Start: Oops!
Still, like many of the mythical confrontations between human and computer in science fiction literature and cinema, the first interim Turing test involved a technical glitch.
The start of the contest was delayed almost 45 minutes while techniciansscrambled to sort out a problem that kept telephone lines connecting two of the human confederates to the contest busy. When it did start, the number of terminals and screens was reduced from the original 10 to eight.
Several members of the Loebner Prize committee said they thought that the glitch made judging harder.
Joseph Weizenbaum, a Massachusetts Institute of Technology computer scientist, said that he was struck by how similar PC Therapist was to a program that he had written several decades ago called Eliza. Eliza became widely popular for engaging questioners in a simple nonsensical conversation. He said that such programs were very difficult to trick because it was difficult for humans to assess the nonsensical replies from the program.
After the contest, Mr. Dewdney said that he had quickly decided that there were six programs and two humans behind the terminals. But he was surprised when he heard that people were changing their minds about several programs late in the contest. A Shakespeare Curve
In particular, at a terminal which was discussing Shakespeare's plays, a great deal of confusion was inspired because the terminal kept changing its answer to the question, "What is your favorite play?"
One of the human judges, Lou-Anne Kroutil, a student and part-time auto mechanic, said she had been fooled by PC Therapist when it responsed with a nonsense answer.
"It typed something that I thought was trite, and when I responded it interacted with me in a very convincing fashion," she said.
Mr. Weintraub's program held a number of the audience spellbound. One conversation went as follows:
Terminal: A whim is a sudden capricious and often peculiar behavior. A bit of sunlight in a grim grey world!
Judge: Something like a hunch?
Terminal: Everybody sets out to do something and everybody does something, but no one does what he sets out to do.
Judge: Well anyway I would imagine whimsical converation is like a series of knight's moves in chess -- sort of zig-zag rather than following any particularline.
Terminal: A whimsical person is subject to erratic behavior and unpredictable thoughts. . . .how nice to be unpredicttable!
Judge: Are you kidding? Sure total predictablicity would make the world a dull place, but unpredictability is hardly a virtue!
After the contest was over Dr. Loebner pronounced himself satisified with the results and said that he felt that the Turing Test would now have a life of its own. Dr. Loebner is president of Crowne Industries, an equipment supply firm in Orange, N.J.
He said that he had thought of sponsoring a Turing Test seven years ago to stir interest in designing artificially intelligent programs.
Calling himself a "dyed-in-the-wool," utopian, he said that in bothliterature and science there has been a noble tradition by those whohave tried to build machines with human qualities.
The New York City philanthropist also revealed another, possibly more personal motivation in sponsoring the contest: "I'm in favor of 100 percent unemployment. I've always wanted computers to do all the work."
No comments:
Post a Comment