Mysteries of the Brain That Baffle Science
“A bold new breed of supersmart computers is taking shape in artificial intelligence laboratories,” says High Technology. They are second-generation “expert” systems, which—like their first-generation counterparts—will have the specialized knowledge of human experts encoded in their data banks. The newer systems, moreover, will have some problem-solving abilities not found in the older versions. But will they be able to think?
Creating a computer that thinks has been the dream of computer engineers ever since the mid-1950’s, when artificial intelligence became a well-defined field of computer science. But so far the dream has not materialized. “We don’t have programs that are truly creative, or truly inventive, or can understand the complexities of somebody’s reasoning,” admits Roger C. Schank, director of the Artificial Intelligence Project at Yale. In fact, Psychology Today sums up over 25 years of research this way: “Every human infant can do three things that no computer is yet able to do—recognize a face, understand a natural language and walk on two legs.”
Computers are simply left behind by the capabilities of the human mind. Why? For one thing, the most advanced computer microcircuitry is rudimentary compared to the interconnections of an estimated 100 billion (100,000,000,000) neurons—nerve cells—that are found in a normal human brain. According to one theory, the brain’s retrieval system is based on a network of connections and “this rich network of connections in human memory is one of the most profound differences between humans and machines. The brain’s ability to search for information through its millions of neurons simultaneously looks positively uncanny.” Further, says Science, “the brain makes millions or billions of neuronal calculations simultaneously and in parallel; our current generation of serial, one-step-at-a-time computers are hopelessly outclassed.”
True, some computers can perform difficult mathematical calculations in a fraction of the time it would take the smartest mathematicians. Advanced computers can even beat most people at chess. But the machines have serious limitations. “An inspired chess-playing program might be able to trounce a good player,” states a recent article in The New York Times Magazine, “but change the rules a little . . . and the machine will be at sea, while the human player will manage to cope.”
What gives humans this advantage? We reason and make analogies. We look at a problem from many different angles, distinguishing important data from what is irrelevant. Further, we have no difficulty dealing with language concepts or with learning from experience. In short, we have “common sense.” The frustrating experience of trying to duplicate this “common sense,” says Science, has given scientists “a certain humility, an appreciation of how awesomely complex the most ordinary human act can be—and of just how much a computer (or a human) has to know before it can do much of anything.”
Scientists admit that there will be no major breakthroughs soon in producing artificial intelligence, despite the increased capabilities of upcoming computer systems. Part of the problem is that we simply do not understand our own thinking process well enough to create a model of it.
“Aha!” we say when a good idea comes to mind. But just how we got the idea remains a mystery.