FRANK ROSE's Into the Heart of the Mind seems an unabashed effort to imitate Tracy Kidder's The Soul of a New Machine, exploring the field of artificial intelligence as opposed to computers in general. Like Kidder, Rose attempts to humanize high technology by telling a tale of personalities and environment. Alas, like Kidder, he lacks a novelist's flair for doing so and relies on obsessive detail as a substitute.
To describe the AI labs, he begins with a dry, factual synopsis of the weather outside their windows. To convey the character of computer programmers, he invariably lists the colors of their clothes, as in: "Joe arrived breathless from the Xerox room, closing the door behind him. Today he was wearing brown polyester slacks and a yellow shirt with wide brown stripes. He seemed very animated." Readers would surely not tolerate such pedestrian scene-setting in a work of fiction; perhaps it's more acceptable in a book of this type because it demystifies a disturbing subject simply by making it dull.
To be fair, there are sections of lucid and authoritative exposition buried among the platitudes. The persistent reader can piece together a fair summary of the history of artificial intelligence, and its two current areas of endeavor: pure research, with the grandiose goal of programming a computer to function in exactly the same way that our minds function; and applied research, which uses orthodox programming techniques to make computers imitate a few limited skills such as deductive logic or pattern recognition.
Applied AI is already being used in industrial robots, cruise missiles, mail-sorting machines, and oil-prospecting programs, to name a few rando examples. It will soon impinge on domestic life in the form of chattering toy robots and "friendly" appliances, exacerbating our already tense love- hate affair with technology.
THERE IS certainly scope for a book about these everyday gadgets. But Into the Heart of the Mind focuses instead on the abstractions of the pure reseach, which shows no imminent signs of impinging on anyone or anything, mainly because it has made negligible progress toward its goal. The human brain contains perhaps 100 billion neurons, in a web so complex that each neuron may well be linked with a thousand others. No one knows how the cognitive areas work: how vision is interpreted, memory is stored, or thought is processed. No one even knows what thought is.
Consequently, computer simulations of the brain are attempts to imitate the unknown. Researchers must work by inference, studying human behavior in hopes of deducing the thought processes that direct it. Computer programmers thus end up as social psychologists, a role for which few of them have much natural aptitude. Their approach is relentlessly mechanistic -- almost a parody of behaviorism -- and even their simplified models of social action and reaction, goals and avoidance, achieve meager results. Rose describes months of tedious work to give a computer the initiative to solve the elementary problem of how to go out into a rainstorm and pick up a newspaper without getting wet. (Answer: put on a raincoat.) Teaching the computer what rain is, what a raincoat does, what "outside" means, what movement entails, and the relationships between these elements, turns out to be horrendously complex. In fact, as Rose describes, pure AI remains bogged down in philosophical debates about the best way of thinking about the process of thinking about thinking.
The Soul of a New Machine derived structure and focus from its own simple mandate: to chronicle the development of one computer from conception to completion. Alas, pure AI research lacks such a tangible theme or product, and Rose's book is unstructured and diffuse as a result, meandering from one academic discussion to the next. The resulting sense of inconclusiveness is exacerbated by the author's somewhat puritanical refusal to enliven the wealth of detail with any analysis or comment. The reader is left without guidance, drifting through a gray world of philosophical speculation and incomplete program code, populated by scientists who seem affable yet as unexciting as astronauts in a NASA press release.
One longs for the humor, sharp perceptions, and forthright opinions of, say, Tom Wolfe. Not until The Right Stuff did we discover the real human human interest in the space program. To judge from the recent spate of dull books about daily life in data processing, we may have to wait another decade or so to discover the true colors of computer scientists.