COMPUTERS: THEIR SCOPE TODAY (Oct, 1967)
COMPUTERS: THEIR SCOPE TODAY
ARTICLE BY ERNEST HAVEMANN
AT THE Massachusetts Institute of Technology there sits a giant computer, its lights constantly blinking and its dials endlessly churning out new numbers, on which some unknown technician has fastened one of the buttons now so popular among the hippie set. The button reads:
I AM A HUMAN BEING.
DO NOT FOLD, SPINDLE OR MUTILATE
Newcomers to the laboratory spot the button, move in for a closer look and nod—yet seldom smile. To most people who deal with computers, the button seems not funny, not ridiculous, not cynical but oddly appropriate.
True, the computer has introduced us to an age in which a great many human traits, activities, hopes and fears have been depersonalized into cards designed to be run through a machine, like so many playing cards through a baccarat shoe or slabs of steel through an assembly line. We pay our bills and resubscribe to our magazines on coldly formal rectangles that must not be folded or spindled. We sometimes take our college tests on them. The friendly neighborhood bank knows us only as a number, printed in the strange devices of magnetic ink. We are also no more than a number to machines in some unknown location, never seen by us, that determine our credit standing, calculate the Social Security benefits we may someday receive and decide how honest we have been on our income-tax returns. To other machines, busy charting population growth, divorce statistics, economic trends, future demands for houses and automobiles and the number of hospital beds that will be needed for the victims of heart attacks, we are not even a number—just a couple of anonymous magnetic beeps stored somewhere inside an electronic circuit whose mysteries we would never be able to unravel. The rows of lights wink like drunken fireflies; the wheels spin out their figures; human history gets made and recorded almost without the touch of a human hand.
Still, there is something about the computer that can inspire affection. You can talk to a machine now. Well, not exactly talk; you have to type out your end of the dialog, and the machine types in turn. But this new kind of pen-palship has its own kind of warmth. The machine is at least as solicitous as an airline stewardess and far more polite than a New York bus driver. It says, “You’re quite welcome”; “Come again”; “I didn’t understand you”; “Feel free”; “Very good.” To a school child trying to learn arithmetic, it says, “Hello, I’ve been waiting for you”; “Please type your name”; “Goodbye, O fearless drill tester.” Dr. Joseph Weizenbaum of MIT tells the story of the time he let his secretary start a typewritten conversation with his machine. After a few sentences back and forth, the dialog began to seem so personal and private that she asked Dr. Weizenbaum to please leave the room.
But there is an even deeper bond between the men who use the computers and the machines they use. Some men lavish affection on their automobiles, those metal creatures that have augmented the mobility of the human leg muscles. Some men are sentimentally attached to their power tools, those potent extensions of the human arm. The computer scientists often take this same fond attitude toward their marvelously intricate, cunningly designed, mechanically beautiful extensions of the human brain. Many of them, indeed, believe that the computers came along just in time to save civilization from collapsing under the weight of its own complexity and are therefore not only the friends but the saviors of man. “The trouble with a world like ours,” says Professor Robert Fano of MIT, “is that we will soon need more experts to run it than there are people. Only the computer seems likely to keep us going.” This is a sentiment shared by many thoughtful observers of civilization who, without any direct experience with computers, have developed a grateful respect for them. Professor Jacques Barzun of Columbia University, a noted humanist who might be expected to resent the machines, has said, “I think computers are perhaps the salvation of the welfare state —particularly the overpopulated welfare state in which we’re all going to live.”
There is a persistent legend, of course, that computers are not very smart—no smarter, say, than an adding machine. Faster, yes. Smarter, no. This is partly the result of propaganda. The companies that make these fantastic new machines are all too painfully aware of what happened when the first wide-scale application of labor-saving machinery was begun in the factories of about a century and a half ago: A group of determined Englishmen called the Luddites, believing that the machines would throw everybody out of work, set about systematically destroying them. If machines that threatened to make human muscles obsolete could arouse that much mass resentment, the computer manufacturers figure, then machines that threaten to supersede the human brain may arouse a good deal more. Computer makers consistently put the knock on their own products, which they like to call, in the words of one of them, “a tool and nothing more.” They shudder to hear the machines compared with the human brain. They even avoid die word “memory”; they say that the machine has not a memory but a “storage capacity.”
It is also true that, until recently, die machines were not asked to behave in any very intelligent fashion. When they were first introduced, they were terribly expensive to own or rent and to operate. They were put to work only on jobs where they could pay their own way— and these were mostly clerical and accounting jobs of the simplest kind, where the machines did nothing spectacular but did it in enormous quantity and at superhuman speed. A good example is keeping track of magazine subscriptions and printing the mailing labels.
Even today, most of the computers in the world are performing routine tasks. It is only when you go to the universities and the experimental laboratories of the manufacturers that you catch a glimpse of the true possibilities—and see machines that can do well on human intelligence tests, can converse intelligently, can learn and can even teach. Yes, even teach. One of the most exciting prospects for the future is that all of us, however lacking in engineering skill, will someday be able to operate a computer as easily as we now operate our automobiles, because the computer itself will show us how. It is only a matter of time, you discover in the laboratories, before all of us will have our homes hooked up to what the scientists are calling “information public utilities” and will have brain power piped in just as we now have electric power. What the world will then be like staggers the imagination. But more of this later.
As of this moment—the autumn of 1967—die most important fact about computers is that they already are essential to civilization as we in America now know it. They are only 20 years old; they have been used mostly in obvious and unimaginative ways; their real potential has only been scratched—yet life would not be the same without them. You cannot have a nation as big, prosperous, active and mobile as the United States, with 200,000,000 people in constant interaction, without the computer. The complexities of keeping the communications lines open, getting the goods delivered, keeping the accounts and paying the bills represent too big a job for the unaided human brain and hands.
The telephone system is a good example. If every call still had to be handled by an operator sticking plugs into a switchboard, today’s volume of telephone conversations, local and long distance, would be an intolerable strain on the economy; it would take the services of every woman now alive in the U. S. Instead, the job is done by computers. To those who doubt that the average man will ever be using a computer, indeed, the telephone system provides a clear answer. Every one of us already has a computer console in his own home—the telephone dial or touch buttons—and we use it every day to tap the amazing resources of the computer. By dialing or touching ten numbers, we manage with the computer’s help to ring a telephone all the way across die continent, any time we feel like it.
If you are running a modern railroad with tens of thousands of freight cars, how do you know where any one of those cars is at any given moment? You don’t—or didn’t until recently. One reason the railroad industry has fallen on lean days is that the average freight car has been used on an average of only a little more than an hour a day and has stood idle the rest of the time, waiting for somebody to discover where it was, put something in it and send it on its way. Now the New York Central uses a computer to keep an account of every one of its cars, and the Association of American Railroads is about to install a computer that will keep track of all 1,800,000 cars in the nation.
If you are running an airline with several hundred flights a day and a thousand or more ticket counters, how do you keep track of reservations? You don’t —without a computer. Here American Airlines pioneered, with its $30,000,000 system called SABRE, which serves all of its offices, from Boston to San Diego. SABRE tells the ticket clerk immediately if space is available on any particular flight at any time within the next year; it then reserves seats under die customer’s name, remembers how to get in touch with him and, when the time comes, makes sure there are enough drinks and food aboard to keep him happy. It even corrects the kind of human error mat used to foul up the airlines’ plans almost beyond redemption. John Smythe, on a trip of many stops from coast to coast, stops by a ticket counter to change the date of his next flight. The clerk, as so often happens, gets the name a little bit wrong; he asks the computer about John Smith. Without hesitation, the machine says, in its own language, “Hold on, buddy. I’ve got John Smythe, John Schmidt, John Schmid and John Smithfield—but no John Smith. Which one did you mean?” Other airlines have quickly followed suit: Delta, as a side line, now uses its computer’s excess capacity to sell tickets to the Atlanta Braves’ baseball games.
Suppose you are an airplane manufacturer and have spent several billion dollars to develop a new supersonic plane that will further improve air travel by carrying several hundred passengers from New York to Los Angeles in a single hour. To the pilot, the plane is a totally new experience^ for the cockpit from which he scans the landing field is several stories above the landing gear. And there’s the rub—for even the best pilot in the world, trying to land this monster for the first time, will almost surely crack it up. Do yon take the chance? Of course not. You use a computer to simulate the airplane’s controls, its performance and the landing field and thus teach the pilot how to land safely and gracefully without ever taking to the air.
These are the obvious ways in which the system—and individual units of the system, such as the supersonic plane— has grown far too complex lor human control. There is another way that, though hidden, is in fact even more important. What has made our modern scientific era possible is an information explosion; our whole society is based on a rapidly expanding new knowledge of physics, chemistry, medicine and engineering. Information pours out of the research laboratories too last lor any human brain to absorb. Around the world, 100,000 separate and different technical journals are published in (30-odd languages. The total amount of scientific data published each year runs to 250,000,000 pages. In the rush, there is even a good deal of information that never gets published at all. In the laboratories of a pharmaceutical company, for example, a new thug discovered by some ingenious researcher today may, in fact, be being “discovered” for the second or even the third time. It’s original discovery may very well lie buried in the unread workbook of some earlier researcher who, knowing of no use for the drug at the time, did not pass his knowledge along.
Thus, the information explosion that produced the computer now needs the computer to keep up with it. A computer prints the weekly publication called Chemical Titles, listing the topics covered by new articles in 690 leading chemical journals. Another computer prints the monthly Index Medicus, which lists alphabetically, under subject matter, the new articles in some 2400 medical journals. Each issue of Index Medicus runs about GOO pages and nearly 2,000,000 words; without the computer it could never be printed in time to keep up with new developments. Drug companies are starting to store the workbooks of their researchers in a computer’s infallible memory, to avoid the waste of new research that merely duplicates the old.
The biggest single user of computers today is the U. S. Government, which would otherwise be crushed under an impossible load of paperwork. Computers not only scan tax returns and keep Social Security records but also calculate census data, direct prospective employees to their Civil Service examinations and make out pay checks for something like 2,000,000 Federal workers and more than 500,000 others who are retired and living on pensions. The Air Force uses one of the most intricate of all computer systems to control its world-wide inventory of equipment and supplies, which is worth well over ten billion dollars at any given moment. It uses another complex system to keep track of every single airplane in the skies above the U.S. and Canada, second by second, around the clock. The space program— the control and tracking of 20,000-mph satellites and spaceships—would be impossible without computers.
The second biggest user is, of course, American business; and there are experts who will argue that it has been the computer in industry, rather than the new economics or sheer luck, that has permitted our nation to remain so prosperous over so long a period. This is because one of the uses to which the computer has been put by business as well as by the Air Force is the control of inventory, at which it has been exceptionally effective. Industries that get a constant flow of computerized information about their supplies and sales can get by with a much smaller inventory than in the past; they do not get stuck with large and immovable supplies of goods and then have to cut back on production until the supplies have been whittled down. A decade or so ago, the inventories of the manufacturers of durable goods used to fluctuate by an average of four billion dollars a year, causing corresponding changes in production and employment. Now, the average fluctuation is cut in half and the peaks and valleys of production have been evened out. No one would argue that the computer has made recessions impossible—but it has certainly reduced the danger from one of the frequent causes of recessions in the past. Other business applications of the computer are legion. Schlitz uses it to make predictions of the sales of beer and to help decide where and when to build new breweries. International Harvester uses it to make simulated runs of new trucks and to predict their life span. Ford has a computer controlling a production run that will operate nonstop for three years and at the end of that time will have turned out a 9000-mile strip of windshield glass; this computer gets a constant flow of information on how the process is going from 700 sensing devices and makes the necessary readjustments at 80 control points. Mobil Oil has a computer-controlled refining unit that automatically analyzes market prices and then turns out whatever combination of products will be most profitable. Pills-bury has a machine that provides its executives, at eight o’clock in the morning, with a complete analysis of the company’s sales and inventory position as of closing time the previous day. Executives at Woodward R: Lothrop, a Washington department store, get a similar early-morning report on the previous day’s sales in each department of nine separate stores, along with a running account of this year’s trend compared with last year’s at this date. Clothing manufacturers feed a new dress design into the machine and let the computer draw up the many individual patterns needed to produce it in all sizes from 6 to 46.
The stock exchanges have computers that talk. If you want to know what is happening today to U. S. Steel or Syntex, you simply call the computer on your touch-tone phone, touch three more buttons to give the code number of your stock, then listen. Stored in the computer are sound tracks such as those on a movie film. The computer picks out the correct combination of voice sounds to give you the quotation on the last sale, the current bid and offered price and, in fact, the figure at which the stock opened and its high and low for the day. Several times a day and immediately after the close of the markets, the Associated Press uses its own computers to rush the stock tables to newspapers around the nation.
In medicine, the computer already has many uses. A computerized dummy of a human being, with a heart that beats and throat muscles that twitch, is used to train anesthesiologists at the University of Southern California; the dummy, which sometimes has heart failure or vomiting attacks, enables the trainees to learn in a few days what used to take several months to master. At Mount Sinai Hospital in New York City, a computer is used for rapid analysis of electrocardiograms made at the patient’s bedside. In fact, electrocardiogram signals have been sent from France to the U. S. via satellite, have been expertly analyzed by a computer and the results have then been sent back to France, all within 30 seconds. In many medical centers, computers make and report on blood tests much faster than the tests can be made by human technicians. In Mayo Clinic, a * computer keeps constant track of the blood pressure, body temperature, heart rate and breathing rate of patients undergoing neurosurgery and flashes its findings onto a television screen in the operating room. At Presbyterian Medical Center in San Francisco, a computer keeps watch over patients recovering from open-heart surgery and flashes an immediate alarm if complications develop. In Brooklyn, nine hospitals are hooked into a computer system that keeps a second-by-second census of the beds available or occupied in their children’s wards; a child brought to a hospital that is already filled can be sent without delay to the nearest place where space is available. At Sara Mayo Hospital in New Orleans, a computer plans menus that provide a balanced diet and the proper calorie count at lowest possible prices; the computer takes full account of the preferences and dislikes expressed by patients in the past and also the need for variety in the color and consistency of the foods served at each meal.
The miscellaneous uses of the computer would require a catalog of their own. It has been used by modern versions of the old-fashioned lonely-hearts clubs in an attempt to help college men and women, and older people as well, find dates who share their interests and tastes. (Computerized dating may be just a fad and no better than older methods of matching cards showing personality traits—but conceivably, if applied to thousands of people on a city-wide scale, it could be a standard part of romance in the future.) It is used in printing plants to set type, on airplanes and ships as a sort of supernavigator, in engineering to solve previously insoluble problems and to draw up working blueprints, at the race track to calculate the pariimituel odds and pay-offs. It is being used to analyze and to plan ways of combating air pollution and water pollution and, experimentally, to relieve big-city traffic jams and to improve weather forecasting. It provides pathways through the vast confusion of the nation’s laws, for the benefit of lawyers and legislators and, in fact, augments the grasping power of the long arm of the law. In Chicago, it is used to enforce payment of 2.500.000 parking tickets issued each year. The Federal Bureau of Investigation has set up a nationwide computer system that will eventually enable any local policeman, when he sees a suspicious automobile, to phone his headquarters and learn almost immediately, while still keeping the automobile in view, whether it has been stolen or is wanted in connection with some crime. The computer is even used to design and help manufacture better computers—an indication that it may someday, like living creatures, be capable of reproduction.
If the nation’s computers went dead this very moment, all of us would know it at once—and to our sorrow. Our telephones would go silent. Our self-service elevators would stop. It would be impossible to make an airplane reservation. A great many industrial plants would shut down. The nation’s banks would stagger under an avalanche of checks impossible to sort by hand. Yet most of the current uses of the computer—however marvelous, however essential to the day-by-day operation of a civilization as complicated as ours—are somewhat remote from your and my personal experience. Most Americans have never seen a computer, except possibly on television on election night. Except for our contacts with telephones and elevators, most of us have never operated one or even dreamed of using one in any intensive way. As a matter of fact, there are very few executives in Government or business, those two great customers of the computer industry, who have ever had any direct personal contact with a machine. Most executives do not understand computers or know how to use them; they deal with the machines through a group of rather mysterious middlemen called programmers, who use a new kind of language, full of code words and mathematical symbols, to tell the machines what to do.
Thus, the computer age, in many ways, has so far been a disappointment. The computer was supposed to revolutionize our lives; it was supposed to put marvels of new technology at our finger tips: there were even some dark but fascinating hints that it might out-think us and take us over. But where are all these fabulous new machines and what are they doing for us} “Computers,” admits one of their manufacturers, “have been flagrantly misrepresented. They haven’t been useful to people at all—except to the few people who run them.” But that day is ending. “In the past,” says this manufacturer, “computers have been like railroads, great for the businessman who wants to ship a heavy load of freight across the country, no good at all to the individual who wants to get from his home to his office. Now we’re moving into a new stage, where the computer will be like a passenger automobile, available to take anybody wherever he wants to go, whenever he wants to go there.”
There are many scientists now at work trying to adapt the machines to handle the complications, not of civilization as a whole but of life for the individual man; these scientists are thinking not of the mass problems of society but of the personal problems that you and I encounter every day of our lives in this intricate, difficult, baffling and often exasperating world. This is the special goal, for example, of Dr. Fano and the MAC Project (for Machine Aided Cognition) that he directs at MIT. Says Dr. Fano, “We’re all faced with daily problems that are rapidly getting much too complex to solve. We don’t have enough time, information or experience to keep up with the growing difficulties that surround us when we try to budget our incomes, pay our bills, balance our checkbooks, make out our tax returns, save and invest for our old age, decide whether to buy or rent a house, evaluate a new job offer, contemplate a move to California, keep track of family anniversaries, draw up a will or even plan a sensible work schedule for tomorrow. Somehow we have to break through this ceiling of complexity.”
The kind of computer use that Dr. Fano and his colleagues have in mind is typified by an experimental program already under way in California, where high school students get a highly expert form of guidance from a machine. In planning his next year’s program, the student goes to the machine and learns what courses are available. The machine asks him if he is planning to go to college and, if so, what kind of college; it then advises him what kind of courses to take to meet that college’s entrance requirements as well as his high school’s requirements for graduation. It tells him how well students with his kind of record have done at that kind of college —how many of them have managed to be A students or how many have flunked out—and may suggest that he is aiming his sights either too high or too low. It refers all special problems to the school’s guidance counselor, while freeing him from the routine questions.
A similar sort of program, Dr. Fano has pointed out, could be devised for helping people make out their income-tax returns; it would be superior to any conceivable book on taxes. One trouble with the tax laws is that they have to Ik written to cover every possible situation; they must therefore be extremely complicated. The trouble with the books that try to explain the laws is that they can do so only by citing examples—which often turn out to be slightly different from our own problems, in ways whose importance or unimportance we find hard to judge. A computer could be programed to listen to any kind of tax problem, even the most unusual, and work out the logical applications of the law to that particular case. Dr. Fano has even suggested that the income-tax laws of the future might be programed directly into the machine instead of printed on paper.
One computer scientist, partly for fun but partly in earnest, has been thinking about a computer program that would give advice on restaurants in New York or other big cities; it would ask the user questions such as, “Do you like Chinese food?”; “Do you prefer a small place or a large one?”; “Do you want music?”; “How much do you want to spend?”; “Does the restaurant have to be near some particular theater?” From what it learned about the user’s tastes and budget, it would then give him a list of places he might like. Afterward, by finding out how well he actually enjoyed the meal, the machine would learn even more about him and be able to supply even better advice the next time. The same program could be enlarged to suggest ways of spending an evening or a weekend. (“People who live in cities have a fantastic array of choices of how to spend their time,” says this scientist, “but at the moment, they don’t take much advantage of the choices. They need help.”) It could even be made into a sort of super travel bureau, with a full array of up-to-the-minute information such as no travel bureau now possesses. Conceivably, you could call up the computer and say, “I’ve got from four o’clock Friday afternoon to nine o’clock Tuesday morning; where shall I go?” The machine would ask some questions, such as, “Would you be willing to spend as much as $300?” and “Do you like to fish?” and would then come up with a suggestion: “I can give you a suite in the Americana Hotel in Miami Beach for $75; you can have a seat on Eastern’s Flight 123 at six p.m. Friday and return on Flight 321 arriving Monday midnight; Captain Harry Brown and his fishing boat will be waiting for you at seven o’clock Saturday morning.”
The possibilities are almost endless. It is perfectly feasible to program a computer to help the baffled householder; he could be instructed how to repair a leaky faucet or get squirrels out of his chimney. Dr. Spock’s wisdom about babies could be put into a computer, in much more detail, to help meet individual crises, than is possible within the covers of a book. Excellent medical and legal advice could be dispensed by a computer, if the medical and legal professions permitted. It is even possible that people will someday take to the computer the kind of problems they now take to psychiatrists, psychologists and marriage counselors. A computer program has already been developed, in fact, that actually sounds quite a bit like some types of psychotherapists; it has had the following typewritten conversation with a make-believe woman “patient”: patient: Men are all alike. machine: In what way? patient: They’re always bugging us about something or other. machine: Can you think of a specific example?
patient: Well, my boyfriend made me come here.
machine: Your boyfriend made you come here?
patient: He noticed that I’m depressed much of the time. machine: I’m sorry to hear you are depressed. . . .
And so on, through what sounds very much like a 50-minute session with a therapist.
Another use of the machine that is prominent in the thinking of computer scientists is in shopping. The Sears, Roebuck type of printed catalog, expensive to publish and distribute, will probably vanish. Instead, the shopper will call up the computer at Sears, Roebuck or his favorite department store, ask what is available, see pictures of it on a television screen, ask questions and get more intelligent and complete answers than he is likely to get in person from most of today’s salesclerks. He will probably pay for his purchase with computerized money, by direct transfer from his own bank account to the account of the store, without writing or mailing a check. This kind of financial transaction is already taking place successfully in an experimental program set up by the Bank of Delaware and a chain of shoe stores; when a customer buys a pair of shoes, the clerk uses a touch-tone telephone to call the bank’s computer, enters the customer’s identification number and presses buttons showing the amount of the sale. The method is bound to spread, because the volume of checks written in the nation, about 70.000,000 a day, is rapidly getting out of hand, despite the computerized sorting of checks. It costs the banks more than three billion dollars a year just to move the checks through the clearing system and eventually back to the people who made them out.
The physical barriers to setting up an intellectual public utility are not very serious. As a matter of fact, there already is such a system, designed to help scientists, in operation under the MAC Project. It has 160 “outlets” in the form of teletypewriters scattered around the MIT campus and in the homes of professors; these typewriters are connected to two separate computers, each of which can serve 30 users simultaneously under a time-sharing program developed by the MAC scientists and the computer manufacturers. The MIT faculty and students use the system constantly, particularly to help with the complicated mathematical calculations involved in advanced research. One engineering professor says that he has not used his slide rule, once the badge of his profession, in three years.
The same kind of time-sharing computer system could rather easily be set up to serve an entire community; each home would have its own console, preferably including not only a teletypewriter but also a viewing screen of some kind, connected to a computer center just as every home is now connected to electric power and telephone centers. The machine might even communicate to the user by speaking, although it probably would not understand human speech. (It is much more difficult to build a machine that listens than one that speaks, and opinion among scientists as to how soon, if ever, a listening machine will be available is sharply divided.) Although the physical problems are easy to solve, there are other problems more sticky. One of them is that most of us would hardly care to spend two years learning how to program a computer for our own special uses; the experts will have to find ways of making computers easier for the average man to use. In the words of Dr. Fano, “The user should be able to talk to the machine as easily as a businessman now talks to an assistant who knows the business well. In fact, that is what the machine should be— a skillful and knowledgeable assistant.” But progress in this direction is being made all the time. Computer scientists like to point out that, when the automobile was introduced, a man had to be an expert mechanic to own and drive one. Now, they say, the computer is rapidly getting to the stage where anyone can learn to operate it, as the automobile did long ago.
What will the intellectual public utility of the future bring into our homes? Dr. Fano foresees a day when the computer will have a giant mass memory containing all the knowledge of its particular community—all the knowledge now found in the books and journals of the very best library, plus everybody’s daily work reports, financial records, tax returns, medical histories and what have you. Storing as much information in a computer as can be put on a page with single-spaced typewriting now costs as little as ten cents a month in some systems; eventually, it will probably be cheaper to store everything in a computer than in a book or a filing cabinet. When that day comes, the computer will put all the accumulated knowledge and skills of the world to our own personal service. A man will be able to produce the solutions to problems in calculus without ever studying calculus, to design a new house without ever taking an architecture course.
Far from depersonalizing life, the computer of the future will probably create a great deal more individual variety. Indeed, it has already won some battles against mass conformity. It makes possible the manufacture of more different models of automobiles with more available options; and in Los Angeles, it enables a customer to walk into any of 75 Dodge showrooms and immediately select exactly what he wants from a pooled and computer-controlled inventory of 6500 cars. Computerized teaching machines have provided considerably more individualized instruction for elementary school pupils, geared to their own backgrounds and ability to progress, than is otherwise possible in a large class. The computer has even brought back at least a faint echo of that delightful bygone day when the village librarian knew every patron so well that she could suggest new books that she knew would interest him. At the big technical libraries of IBM, profiles of staff members have been fed into a machine that automatically sends them abstracts of new books and articles they might want to see. In the future, the computer will very likely print out a morning newsletter or even an entire newspaper designed specifically to meet the needs of the man reading it. It could tailor-make individualized magazines and television programs and permit a man to design his own special kind of furniture or even a new automobile. On an even more important level, it may free us from the restrictions of what has been called the creeping bureaucracy of our recent past—for many of today’s rules, regulations, inspections and permits are designed simply to prevent a densely populated society from falling into chaos. In a society where information circulated freely and almost instantaneously, many of the restrictions would probably be unnecessary.
Which brings us to the question: Just how brainy, in actual fact, are these machines? Can they learn? Can they think? Can they create? Will they ever reach the stage assumed in a joke popular among computer scientists, in which one machine asks another, “Do you believe in man?”
The answer to most of these questions is that, at the moment, nobody really knows. The phrase used by the scientists in this connection is “artificial intelligence”; some think the machine has it, others that it does not.
Certainly the machine can learn. The machine that counsels the California high school students asks questions, learns from the answers what each student is like, then gives him not just general advice but guidance intelligently molded to his special needs. The University of Virginia has developed a computer system that learns to solve problems by trial and error. Part of the MAC Project is a language program called ELIZA because, like Eliza Doolittle in Pygmalion and My Fair Lady, it can be taught to use language increasingly well. Indeed, once ELIZA has been taught English phrases, it can also learn the German equivalents; the operator need merely tell it, “When I say Ich sage, I mean I say”—and the machine remembers this.
As anyone who has ever taken a psych-I course knows, however, the human brain solves problems in strange and marvelous ways. All of us possess a large memory bank of facts, mathematical rules and knowledge about relationships. In our thinking, we sometimes manipulate this knowledge through the rules of logic; we know that since all mammals nurse their young and since a whale is a mammal, whales must nurse their young. At other times, we manipulate it through our own individual kind of something like free association. To a minister, the word “angel” may set off a train of thought such as angel-heaven-God-Holy Ghost-Virgin Mary-Sermon on the Mount. To an athlete, the associations might go angel-fly-fly ball-baseball-Willie Mays. If the machine is ever to be truly creative, it must learn to think in this random, freewheeling, rather haphazard but richly productive kind of way. Can it? Maybe, maybe not. Nobody has yet tried to make the machine think in this fashion.
There have been only about 20 experiments in artificial intelligence, most of them conducted by graduate students working on Ph.D.s. The students have tried what seemed most promising, random thinking not included. They have, however, produced some truly amazing results. One MIT student taught a computer to solve algebra problems stated not in mathematical terms, which is easy for the machine, but in plain ordinary English—such as, “Mary is twice as old as Ann was when Mary was as old as Ann is now. If Mary is 24 years old, how old is Ann?” (Unless you can figure out quickly that Ann is 18, something that few people can do, the computer is a lot better at this kind of problem than you are.) Another student taught a machine to solve the kind of problems in geometrical analogies often included in intelligence tests, such as: The machine can do as well on this part of an intelligence test—even with geometric figures it has never seen before— as the average ten-year-old. And now another student at MIT, this time just an undergraduate, has taught a machine to solve the verbal analogies of intelligence tests, such as dog is to bark as saw is to blade, wood, cut, whine or tool?
There is a popular saying, of course, that the machine can do only what it is programed to do—and a human being must do the programing. This is reassuring to human pride and has become the cliche response to every new accomplishment by the machine. But the scientists who believe in artificial intelligence point out that the human brain must also be programed; our thinking processes are the end result of all the information and instruction ever received by our input devices, notably our eyes and ears, from birth. Moreover, programing a computer is by no means so exact a science as has been advertised. It had to be very precise at first. Now, in the words of MIT’s Professor Marvin Minsky, “Programmers don’t have to know exactly what they are doing; they can be very sloppy, as a matter of fact.” They do not always understand how or why their programs work. And, as the programs get revised and enlarged through dialog between man and machine, they become increasingly versatile. Says Professor Minsky, “People claim that programs are all right for special purposes but aren’t flexible. Well, a typewriter or a violin isn’t flexible, either— until you learn to use it.” In the computers of the information public utilities of the future, which will be conversing with all kinds of people, learning all kinds of new facts and getting instructions to perform many different tasks, the user may be able to present a brand-new kind of problem and get it solved simply by telling the machine, “Which of the problems you’ve solved in the past is most like this one? Try that method.” And quite possibly, Professor Minsky believes, the machine may reach the solution through a method that would never have occurred to the man himself. No one individual will know everything that is in the machine and what may therefore come out. Very little special training will be needed to operate the machine, but probably some users will have better luck with the computer than others. “Some people will be just naturally good at it,” says Professor Minsky, “the way some people are good at skiing.”
Eventually, Professor Minsky believes, the machines will be programed for self-improvement; they will get better and better of their own accord; then they will unquestionably display the traits we refer to as intelligence, intuition and consciousness, and the world will never again be the same. Will they be man’s equal or even his superior? Dr. George Feeney of General Electric says, “I think we’ll get to the point where that question won’t really matter. The humanists, if optimists, will say that the machine is an extension of man—and the realists, if pessimists, will say that man is an extension of the machine.”