Programs for the Future


The Age of Intelligent Machines, by Raymond Kurzweil, Cambridge, Mass.: The MIT Press, 565 pages, $39.95

Before the advent of computers, questions like "What is intelligence?" and "What is consciousness?" were exclusively the domain of philosophy. Artificial intelligence (or A.I.) research, pioneered at MIT in the 1950s, was created precisely to investigate and, ultimately, answer these questions. The Age of Intelligent Machines is an excellent discussion of the roots of A.I., its current capabilities, and its likely future development.

Raymond Kurzweil is well placed to discuss this fascinating field. Himself an A.I. pioneer, Kurzweil developed the first machine that reads printed text to blind people; the first electronic keyboard to accurately recreate the sounds of orchestral instruments; and the first large vocabulary voicewriter. He is also unusual among the technically accomplished in being a clear, readable, and often humorous writer.

Kurzweil's book offers a grand overview of the history and future of A.I. research. More than that, it offers a vision of a weird and wacky world where every appliance has a brain, intelligent houses foresee your every need, and your computer argues with you when you try to buy stock in a company it doesn't like. Weird and wacky it may be; Kurzweil makes a good case for its likelihood.

The holy grail of A.I. research is the development of the intelligent computer: a machine as intelligent, in precisely the same fashion, as a human being. Naturally, that goal has not been reached—and is unlikely to be attained for some decades yet. But by studying intelligent behavior and attempting to reproduce it in computer systems, A.I. research has given us great insights into the nature of human cognition.

One of the strangest realizations has been that many things we think of as difficult are actually very easy; and, conversely, that many things we think are trivial are actually very hard. As early as 1956, the Logical Theorist was able to derive original mathematical proofs, displaying a degree of mathematical sophistication that, in humans, we expect only of the highly educated. Yet a robot that can perform household chores is still considered vastly beyond the capability of modern technology.

A housewife, in some sense, is far more intelligent than a professor of mathematics. The things we think of as conscious reasoning—logical progression from one step to the next—are, in fact, the easiest things to program. Simple pattern recognition—the ability to recognize faces, for instance, or to realize that the cat is not part of the rug and shouldn't be vacuumed—is one of the most difficult.

As Kurzweil shows, artificial intelligence is more than an academic party game in search of an unreachable goal. In the last decade, it has begun to throw off billion-dollar industries as it advances. One of the most interesting is "expert systems"—in essence, massive databases coupled with "rules of thumb" drawn from the knowledge and experience of human experts.

Expert systems are used to diagnose disease, search for minerals, and offer tax-planning advice. American Express uses an expert system to judge the credit worthiness of its charge-card applicants— and the company claims a 75-percent reduction in expensive bad judgments since it adopted the new system. Before the end of the decade, expert systems will probably be at the heart of medical practice, financial planning, and military strategy—to mention only a few fields.

A.I. applications are in their infancy. Within a few decades, we are likely to see books replaced by flat-screen readers and the publishing industry altered beyond recognition: the standard means of interacting with a computer is likely to be English-language conversation: and your telephone will probably be able to translate your English into Urdu for the person on the other end of the line, and his Urdu into English. By comparison, Kurzweil's prediction that, within 15 years, the world chess champion will be a computer is small potatoes.

Kurzweil is concerned with more than technology: his chapter on the philosophical roots of A.I. is particularly interesting. He includes articles by a variety of other writers, including A.I. pioneer Marvin Minsky; Kazuhiro Fuchi, the head of Japan's Fifth Generation computer project; and journalist George Gilder. If Kurzweil has a message, it's that the prospect of machine intelligence, rather than posing a threat to humanity, has the potential to vastly expand our freedom and well-being. The A.I. revolution has begun, and within the decade, Kurzweil believes, it will dramatically affect all our lives.

The Age of Intelligent Machines is not for everyone, not at more than 500 pages of tiny sans-serif type. (The designer could have chosen a less wearing font—though Kurzweil points out that sans-serif is more easily read by machines.) And while Kurzweil eschews jargon, math, and program code, those who think that bit is the past tense of bite and floppy is an adjective that applies to hats would probably be better off finding a primer on computers before tackling the book. Those who are interested in the impact of technological advance on society will find it fascinating.

Greg Costikyan is a writer of fiction and nonfiction who has designed 23 commercially published games.