BOOK REVIEW / SCIENCE : Building a Better Mousetrap--the Brain : BRAINMAKERS: How Scientists Are Moving Beyond Computers to Create a Rival to the Human Brain <i> by David H. Freedman</i> ; Simon and Schuster, $22, 214 pages

It never occurred to me a decade ago when I was writing about the roles parenting and sexual selection play in evolution that these theories would one day be adapted by a different group of researchers--people interested in building machines that think.

Nor did it occur to me then that building such machines would be possible. I was living in a world apart from men such as MIT’s Marvin Minsky, who were already devoting their careers to the idea that artificial intelligence (usually called AI) was not only possible but desirable. That was the ‘80s. Since then, I have been inundated by a veritable tidal wave of books, articles and TV specials describing computers that can not only direct industrial robots--machines that repeat the same action--but can, it is hoped, learn to adapt to new situations. Theoretically, such autonomous machines will be able to navigate the craters of distant moons and make independent decisions in ways analogous to human thought.

In “Brainmakers,” Boston-based science writer David Freedman sums up the facts and hopes surrounding what he sees as two generations of AI research. The first generation, he explains--which was led by Minsky, among others--got off to a swift start, then lost momentum and ground to a halt when it became clear that abstract logical computer programs could not approach the flexibility and speed of the human brain. Computers could play excellent chess, but they could not initiate projects. The first generation, Freedman says, struck out. The next generation, what Freedman calls “nature-based” researchers, includes molecular biologists, neuroscientists and physicists, as well as computer scientists. They base their efforts, in part, on the model of neural networks--the way individual neurons fire multiple connections inside the brain.

Freedman believes that they will succeed where their predecessors failed because they acknowledge complexity as the key to how nature works and use what they believe is natural (an interpretation open to debate) as their model.

Fleshing out these new approaches, Freedman describes the work of several varied research projects in New England, California and Japan. In lean prose, he provides vivid descriptions of the major players--their biographies, their laboratories, their machines.

Profiling scientists trying to build a microelectronic brain on the scale of quantum mechanics, Freedman explains: “The main motivation for moving to the molecular scale is not merely to speed up existing approaches, but to be able to create new approaches using the same building blocks that nature does . . . biomolecules.”

Freedman introduces us to UCLA’s Chuck Taylor, a biologist doing nature-based research on a larger scale. Taylor is using the tenets of Darwinian evolution to “breed” robots. His aim is to bridge the gap between what Freedman calls the idealized world of the computer and the hopelessly complex world of real-life animals.

Taylor and his colleague David Jefferson, a computer scientist, have started with a herd of 20 robots which, they hope, will grow to 100 in the next year after they follow the model of biological evolution, “killing” off those that perform poorly and mixing the best traits of the brightest.

All of this leads to some pretty basic questions. How smart will these artificial brains become? If they evolve as predicted, will they develop a sense of their own identity and mortality, as the late Isaac Asimov projected in his fiction (and demand a code of ethics and perhaps civil rights protection)? Freedman does not tackle this.

But he does pose the question: Why build artificial brains at all, when we have so many good human brains around? The practical reasons are obvious. Autonomous robots will be able to work where humans cannot--in radioactive infernos or on distant moons. And in the laboratory, artificial brains can provide models for studying natural ones.

But scientists seldom work with practical goals in mind. Freedman understands this and offers an explanation akin to what climbers say about Mt. Everest: They scale it because it is there. Brainmakers build artificial minds because they can. Where this will lead, they acknowledge, is unpredictable. That’s what makes it so exciting.