depends on context.The literature is full of amusing misreadings by machines. In his book Mind Matters: Exploring the orld of Artificial Intelligence, James Hogan tells how in one project in the 1960s, the metaphorical phrase “Time flies like an arrow,” perfectly clear to you and me, was sadly misunderstood by a computer; one of its interpreta-
tions, for arrow.”
instance, was “Time
Efforts to enable computers to be programmed in natural lan- guage and to translate human languages continue, although they are not yet perfect.With large speech databases and fast processors, ma- chine conversation using word recognition and synthesis is becoming routine in such applications as travel booking. What works over a telephone also works in a mobile unit, and so the Sony QRIO robot has language capability. But these systems can hold only limited con- versations, a far cry from the generalized and diverse humanlike re- sponse the Turing test is meant to uncover.
In 1950, Alan Turing predicted that a computer would pass his test by the end of the twentieth century, but we are still far from developing a synthetic intelligence that can persuade us of its own personhood.The best-known attempt to determine how close we are to this goal is the yearly competition sponsored by Hugh Loebner, a hardware manufacturer who developed an interest in the problem and offers a substantial prize for the computer program that best meets the Turing criterion. Rather than using voice communication, these Loebner Prize events test linguistic intelligence by using keyboards for the human-computer interaction, as Turing envisioned.
The Loebner competitions began in 1989, and initially—espe- cially in 1991—attracted luminaries of the AI world. But the conver- sational ability of the artificial minds was disappointingly poor in those first years and has not improved much since; Loebner calls the level of performance “gruesome.” Some of the AI community has repudiated the competition, protesting that it is conducted in a way that renders