**The Best of Creative Computing Volume 2 (published 1977)**

*by Dr. Bertram Raphael*

Misleading Myths Many people believe that computers are inherently stupid, and think that even a suggestion that computers might be made smarter is ridiculous. This belief is so widespread that most people never even consider the many ways in which smarter computers might help them. Misconceptions about a computer's limitations seem to be based upon two widely accepted but basically untrue premises. Let us examine these myths. By pointing out some of their fallacies, perhaps I can open your mind to the fascinating prospects for smarter computers. THE ARITHMETIC MYTH. A computer is nothing but a big fast arithmetic machine. Computers are arithmetic machines, certainly; almost every computer has wired into it ability to add and subtract. But are they "nothing but" arithmetic machines? Certainly not. Take the reference manual for any computer, and scan through its "instruction set": the collection of basic operations it has been designed and wired to perform. You will see a few, perhaps as many as ten or twenty, operations that bear some close resemblance to arithmetic-e.g., ADD, DlVlDE, FLOATING SUBTRACT, MULTIPLY STEP, and so on-but you will also see many, perhaps one or two hundred, operations that have relatively little to do with arithmetic-eg., STORE, LOAD, TEST, SHIFT, READ, WRITE, REWIND TAPE, SKIP, MOVE, MASK, MATCH, TRANSFER. and so on. Much of the time that any computer works on any problem, the computer is positioning, comparing, moving, choosing, copying ..., but it is not doing arithmetic, Rather than calling a computer "nothing but a big fast arithmetic machine," it is much more accurate to say that a computer is a big, fast, general-purpose symbolmanipulating machine. THE STUPID COMPUTER MYTH, A computer is an obedient intellectual slave that can do only what it is told to do. This second myth is even more persistent than the first one, and even more damaging in the way it tends to constrain our thinking. Suppose I gave you the pieces of a jigsaw puzzle and told you, "by the way, these pieces cannot be fitted together." Would you try very hard to fit the pieces together? Why should anyone try to build a smart computer, if he is told over and over again that computers are inherently stupid? The stupid-computer myth has been repeated and generally accepted for more than a hundred years. In 1842, after Professor Babbage of Cambridge designed his Analytical Engine, a large-scaled mechanical digital computer (which unfortunately was never completed), his friend Lady Lovelace wrote, "The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform," There is no question that Lady Lovelace's argument, and all the subsequent versions of the stupid-computer myth, are true, in a certain literal sense: a computer must be given its program of instructions, and it will always do exactly what those instruc This article consists largely of material from the book, THE THINKING COMPUTER: Mind Inside Mattel, by Dr. Bertram Raphael which will be published early in 1976 hy W.H. Freeman and Company. As novel sources of information, amusement, or artistic experiences, the potential for us to benefit from thinking computers is limited only by our imaginations. tions tell it to do (unless, of course, one of its circuits fails). And yet this basic truth is not a real restriction on the intelligence of computers at all. The claim that a computer "can only do what it is told to do" does not mean that computers must be stupid; rather, it clarifies the challenge of how to make computers smarter; we must figure out how to tell (i.e., program) a computer to be smarter. Can we tell a computer how to learn? To create? To invent? Why not? I'd bet even Lady Lovelace would have agreed that the task of figuring out "how to order" a computer "to originate" something would be a fascinating and meaningful research challenge. Progress in "artificial intelligence," the study of how to make computers smarter, is now enabling computers to apply a wide range of problem-solving methods; to communicate in ordinary English; to perceive the physical world; and to combine such abilities into flexible systems that perform useful tasks. The following paragraphs review some of this progress. Problem-solving Methods How do most people solve common, everyday problems? Suppose Mr. Pollack is driving to a ski resort in his little foreign car. On the way he encounters a snow Storm, and finds he must mount his brand new tire chains on the wheels of his car. This problem-how to mount the chains on the wheels-can be divided into many little subproblems. Do the chains go on the from or the rear wheels? Should they be wrapped around a wheel by jacking the wheel off the ground, by driving the car onto the chains, or by figuring out how to use the funny little "mounting tools" that come with the chains? Which side of the chains should be up? How does the peculiar linking mechanism work? And so on. Mr. Pollack must solve these problems as quickly as possible, so that he can accomplish the task without freezing his fingers and soaking his clothes, and so that he can still get to the ski area without missing too much of the day's activities. Well, exactly how is this kind of problem usually attacked? By encoding the known facts into mathematical axioms, and using theorem-proving methods? Not likely! Instead Mr. Pollack (and millions of others) use informal problem-solving methods. Informal problem-solving methods are especially intriguing because of their extreme generality. Problem-solving methods that most scientists develop work only in highly specialized, highly technical areas: e.g., a method for solving second-order linear differential equations with constant coefficients, or a method for estimating the distances of stars less than twenty light years away. Informal problem-solving methods, in contrast, seem to be applicable to a wide variety of problems, most of which may be brand new to the person using the methods: even though Mr. Pollack may never have been called upon to mount tire chains before, he need not be at a total loss as to how to proceed. ls there some standard way of viewing any task, that enables people to apply their reasoning abilities with flexibility to any problem that arises? Psychologists have developed various approaches to explaining such complex cognitive behavior. One approach that has been embodied 43