BRAINMAKERS

A lively, clear, and comprehensive survey of who's who and what they're up to in the world of Artificial Intelligence (AI to cognoscenti) from science journalist Freedman. The term ``artificial intelligence'' calls up visions of the advanced computers and human-like robots familiar from hundreds of science fiction movies. The state of the art in AI today is probably best symbolized by Rodney Brooks' robot Attila, a six- inch metal ``cockroach'' scrambling over and around obstacles in the MIT AI labs founded by Marvin Minsky some 30 years ago. Not that the field has stagnated; in fact, the software-oriented approaches explored by Minsky and his fellow pioneers (John McCarthy, Seymour Papert) are considered somewhat old-hat by today's AI researchers, who have branched out into dozens of new directions. Japanese teams, with a mandate to develop a ``sixth generation'' computer, are exploring how living neurons actually transmit information, hoping to build up a detailed ``wiring diagram'' of the human brain. At UCLA, Chuck Taylor and David Jefferson are following an evolutionary strategy: creating computer programs with small random variations, then choosing the most successful in each batch as the ``parents'' for a new generation, in imitation of biological natural selection. Even the ``neural network'' approach to AI, which mechanically emulates the pattern recognition capabilities of the human brain (and which was temporarily discredited by Minsky and Papert in a 1969 critique), has been revived with some success at Stanford and Caltech. All these researchers have made interesting progress; Freedman lucidly summarizes the methods and results of their investigations, as well as the criticisms of such skeptics as John Searle and Roger Penrose. While much remains to be done to match even the more conservative prophecies of AI pioneers such as Minsky, the accomplishments to date, are often fascinating and shed light not only on the future of computers, but on the nature of intelligence itself.