You're Not Very Smart After All
February 18, 1950 - The Saturday Evening Post
2001: A Space Odyssey, released in 1968 and based at least in part on Arthur C. Clarke's 1948 novel The Sentinel, was more than just a science fiction movie. It was a reflection on the public's and even some of the scientific community's trepidation over the potential power of run-amok computers to be used for or even themselves commit evil (e.g, HAL 9000). Fear of the unknown is nothing new. Noted mathematicians and computer scientists quoted in this 1950 article from The Saturday Evening Post worry about robots (aka computers) "going insane" or being used by the likes of Hitler and Stalin to dominate the world with totalitarian rule. Others, however, have a more optimistic outlook: "The men who build the robots do not share these terrors. Far from destroying jobs, they testify, they will create new ones by the hundreds of thousands, just as the industrial revolution eventually did. Moreover, most of the robot builders would make book that in time 'thinking machinery' will bring about a happier, healthier civilization than any known heretofore. What the odds on Utopia ought to be, however, not even the robots themselves can estimate."
February 18, 1950
The Saturday Evening Post
These articles are scanned and OCRed from old editions of the The Saturday Evening
Post magazine. Here is a list of the
The Saturday Evening Post
I have already posted. As time permits, I will be glad to scan articles for you. All
copyrights (if any) are hereby acknowledged.
You're Not Very Smart After All
By John Kobler
Now the scientists have come up with "mechanical brains" - electronic monsters that solve in seconds a problem that would take you hours. They're human enough to play gin rummy, even have nervous breakdowns.
OUT of scientific laboratories from New York to Moscow there is emerging in ever-increasing numbers a series of wonder-working robots whose power for good or evil, for creativeness in peace or destruction in war, exceeds that of supersonic flight and nuclear fission. Indeed, scientists working in both of those fields, among many others, continually look to the robots for the answers to their thorniest problems. Yet for all their fabulous potentialities the robots merely count and measure.
They are the gigantic computing machines with the bizarre names - SSEC, Eniac, Edvac, Binac, Mark I, II and III, Rudy the Rooter, to list a few and they can solve in infinitely less time than it would take Albert Einstein merely to state them almost any practical mathematical problem and many problems in pure mathematics. Although they have been developed chiefly in the United States, scientists on both sides of the Iron Curtain are now producing them. Recently, Pravda announced that Russia's two top-priority targets of scientific research were atomic energy and computing machinery.
So strikingly do the mechanisms of these robots suggest to some observers the workings of the human brain and nervous system that they are often called "mechanical brains." This infuriates a good many of their creators, notably Prof. Howard Aiken, of Harvard's Computation Laboratory. "They can't think any more than a stone," Aiken states flatly. "They're timesaving tools, pure and simple. There is no substitute for the mathematician, and there
Jack Manning Photos
Its panels of electronic tubes blinking and clicking like mad, International Business Machines' SSEC goes to work on a problem. It costs $300 an hour to run and is booked solid for six months ahead.
Harvard's Professor Howard Aiken is infuriated by suggestions that any robot computer can think.
MIT's Professor Norbert Wiener finds a startling similarity between robots and the human brain.
IBM's President Thomas J. Watson reassures us that machines won't replace mortal scientists.
never will be."
Another school of mathematicians, however, whose most eloquent spokesman is MIT's brilliant, eccentric Prof. Norbert Wiener, does not hesitate to draw startling parallels between the robots and humans. Like humans, Wiener points out, the robots remember, choose, correct their own mistakes. Dr. Claude E. Shannon, of the Bell Telephone Laboratories, has shown how a computer can play chess; Dr. J. W. Mauchly, of Philadelphia, has trained his Binac to play gin rummy. Doctor Shannon puts it :this way: "The machines will force us either to admit the possibility of mechanized thinking or to further restrict our concept of thinking."
Whatever the essential physiology of the robots, it is certain that their computing capacities surpass those of any human being. Consider the behavior of one of these prodigies, Aiken's Mark II, in action:
From the Air Force at Wright Field recently came a request to interpret the performance data of a new four-engine bomber. The end object was to enable the pilot to complete a round trip from air base to target with the optimum consumption of fuel. Expressed another way: given his altitude, load, number of engines functioning and other variables, how fast should he fly to get the best mileage per gallon? This involved finding equations between all variables which would be applicable under all flying conditions.
Aiken entrusted the preparation of the problem to one of his brightest disciples, Peter Young, who is so accustomed to thinking in digits that he has been known to state his age as "twenty-two point seventy-five." Young began by supposing the plane to be on the ground, with no load and two propellers turning. He then rearranged the variables in every practical combination: altitude still zero, still no load, but three propellers turning, and so on up to maximum performance. All told, he correlated 100 items of data. To do so and translate them into the only language Mark II understands - punched tape - took Young two days. Had he attempted instead to solve the problem himself with pencil and paper, he would have had to work steadily around the clock for one month. Mark II ground out the results - 7920 of them - in thirty-six hours.
They rolled off a typewriter-like part in long sheets. When reinterpreted in the form of a graph and installed in the instrument panel of the bomber, they will tell the pilot from minute to minute his exact fuel potential. For example, at 5000 feet, with a load of 70,000 pounds and all four propellers spinning, he will know that to obtain optimum efficiency - in this case one eighth of a mile per gallon - his speed should be 160 miles per hour. "A trivial problem," says Aiken.
Another problem, which cannot be considered trivial, was fixing the position of the moon at any time, past or future, with high accuracy - perfect accuracy is not possible by any method. This was the first challenge to be taken up by International Business Machine's SSEC - Selective Sequence Electronic Calculator - which has the highest capacity and production rate of any calculator now in service - when that mammoth robot moved into its soundproofed, air conditioned chamber in the company's Manhattan headquarters two years ago. It was a problem in pure science, although knowing the approximate positions for the current year is a practical necessity for navigators. The American Nautical Almanac publishes them regularly. But formerly to calculate the current positions would occupy two mathematicians at the Naval Observatory, using what were then the fastest calculators, every working day the year round. SSEC computed more than eight positions an hour. One machine hour corresponds roughly to ten years of paper-and-pencil work.
Today, Government agencies and the armed forces, industrialists, economists and sociologists are feeding problems to the robots as fast as they can digest them. The Mark trio, which cost more than $1,000,000 - a "megabuck" or "kilogrand," as mathematicians say facetiously - work twenty-four hours a day, seven days a week. SSEC, costing $300 an hour to run, is always solidly booked six months ahead.
One of the trickiest tasks, and until recently a top-secret one, to which a robot has ever been assigned was working out equations for the guidance of antiaircraft fire during World War II. Using MIT's Bush Differential Analyzer-designed by Dr. Vannevar Bush - Wiener and several other mathematicians devised an apparatus to be built into antiaircraft range finders which would locate and track enemy planes and calculate the trajectory of the bullets faster than either bullets or planes could travel. This entailed prediction. The fire-control apparatus, in itself a computer, aimed the gun not directly at the plane, but at the next point where the plane might be, taking into account its speed, the wind velocity and other variables.
To improve firing accuracy still further, Wiener proposed adding to the computer's intake a subtler kind of data - the probable behavior of the pilot himself.
"The more a plane doubles and curves in flight," Wiener reasoned, "the longer it remains in a dangerous position. Other things being equal, a plane will fly as. straight a course as possible. However, by the time the first shell bursts, other things are not equal, and the pilot will probably zig-zag, stunt or in some other way take evasive action.
"If this action were completely at the disposal of the pilot, he would have so much opportunity to modify his expected position before the arrival of a shell that we should not reckon the chances of hitting him to be very good. On the other hand, the pilot does not have a completely free chance to maneuver at will. For one thing, he is in a plane going at an exceedingly high speed, and any too sudden deviation from his course will produce an acceleration that will render him unconscious, and may disintegrate the plane. Moreover, an aviator under the strain of combat conditions is scarcely in a mood to engage in any very complicated and untrammeled voluntary behavior, and is quite likely to follow out the pattern of activity in which he has been trained."
Accordingly, the escape tactics of thousands of fighter pilots were analyzed, reduced to equations and incorporated into the same fire-control aparatus. This, of course, could not enable antiaircraft range finders to predict with 100 per cent accuracy the tactics of any individual pilot, but it did immeasurably narrow the margin of probability.
Wiener has since become so terrified by the possibilities of his own war work that in 1947 he refused to address a symposium at Harvard on computing machines, on the ground that they were being used for war purposes. "I do not intend," he declared at the time, "to publish any future work of mine which may do damage in the hands of irresponsible militarists."
A great many adaptations of the robots' answers have been and still are military secrets even to the mathematicians in charge. The Harvard group recalls the day shortly after Mark I got cracking when a problem arrived from the Army which seemed to make no sense. The figure apparently represented an attempt to release an immense output of energy from a tiny input of matter. Only after Hiroshima did Harvard realize that it had been dealing with the mathematics of the atom bomb.
At present, IBM mathematicians are baffled by the 'Purport of what they have named "Problem Hippo." The statement of it covers thirty-six pages, the solution calls for 9,000,000 operations, and it will keep SSEC ticking away for 150 hours, or the equivalent of 1500 years of man-hours. The address of the sender is Los Alamos Scientific Laboratory.
Occasionally somebody hands the robots a problem that stymies them. Such a one was forwarded not long ago to SSEC by the Adjutant General's office, which wanted an analytic expression of qualifications for military personnel. Thousands of recruits had been quizzed before and after service. The Army proposed to establish mathematically what questions put to the recruits on entrance into service had been predictive of their future success or failure as military men. To untangle that one would have taken SSEC 150 years.
And then there are the people who submit problems so far beneath a robot's talents that it would not deign to wink a single tube at them. During the recent Pyramid Club madness a reporter wanted the same robot to compute the number of days one club would need to run to exhaust the population of the world. Robert R. Seeber, Jr., co-inventor, with Frank E. Hamilton, of SSEC, explained to the reporter that this was like asking a Big Bertha to shoot a sparrow. With pencil and paper he whipped out the answer in ten minutes - thirty-two days.
What is the anatomy of the robots and how do they work? Their complexity lies mainly in the vast numbers and interrelations of their parts, the miles of wire, the tens of thousands of tubes. The basic principles are comparatively simple. There are two great families of mathematical robots: the digital calculators and the analog machines. The first, with which this report is primarily concerned, compute in individually distinct digits. In other words, they count. The second, of which the Bush Differential Analyzer is the best known, compute in physical quantities such as length, angle, electric current, water pressure. They measure. The analog machines are faster, but their precision is limited. For the upper spectrum of mathematical shadings the digital calculators are required.
In appearance, a digital calculator SSEC, for instance - is a large chamber one or more of whose sides are glass-enclosed panels of electronic tubes. When SSEC is at work, the panels blink furiously with a click-clacking sound, a galaxy of noisy glass stars in a glass sky. Standing in this chamber with the IBM motto, THINK, emblazoned over the doorway, visitors sometimes remark that they feel, not like a man with a brain inside him, but like a brain with a man inside it.
The men who tend SSEC vigorously agree with IBM's President Thomas J. Watson that" no machine can take the place of the scientist; this machine only leaves him more time for creative thinking." At the same time they display an almost emotional attitude toward it, patting it when it functions smoothly, chiding it when it falters. "We think of it as having temperament," one of the scientists confesses, "a woman's temperament."
The robots have five main groups of organs: An input system - the "eyes," so to speak, which read the problem and the instructions for solving it. Computing units - the inner "brains" which perform the actual mathematical operations. Storage cells or" memory" of two kinds, one which remembers intermediate results until they are to be combined with the body of the problem - as when you say "put down two and carry the one" - and a permanent memory containing logarithms and functional tables. A central control or "nervous system," to route the traffic of numbers from one set of tubes to another, keeping the operations in the right sequence. An output system, or "voice," that delivers the final solution. These five organs are fundamentally mechanized versions of the same ones you use when tallying a bridge score or checking your bank balance.
For the robots, which, after all, are not quite so bright as you, the job has to be facilitated by several ingenious short cuts. Here is one of them: the most fiendishly intricate problems that scientific genius might dream up can be reduced to the four elementary operations of schoolroom arithmetic: addition, subtraction, multiplication and division. And these can be further reduced to two, for multiplication is merely repeated addition, and division merely repeated subtraction. So no matter how knotty the problem, the robot need only add or subtract at any one stage.
Another short cut is its language the punched card or perforated tape, to mention only two dialects in use. A card or tape wide enough to carry five positions in a row offers thirty-two different possible meanings. Thus, the first position can be blank or punched, two possibilities; combinations of first and second positions give four possibilities; and so on up to thirty-two.
The robots' panels frame cells or banks of tubes, each tube corresponding to a position on the cards. Eniac, a ten-digit calculator, has cells of ten columns, ten tubes to the column. The first column represents digits, the second tens, the third hundreds, and so on. The bottom tube of each column represents 0, the second 1, the third 2, and so on. Suppose the number to be indicated is 6,487,399,961. As the card is fed into Eniac's input system, electrical pulses light up Tube 6 in the tenth, or billion, column, Tubes 4, 8, 7 in the hundred-million, ten-million and million columns, and so on.
To follow a simple operation from start to finish, take 268 times 64. The first step is up to the mathematician, who must break up the problem into a kind of pidgin mathematics - the additions and subtractions that the robot can readily handle. Furthermore, the problems as originally propounded by the sender are rarely free from errors in statement, and these errors must be weeded out. The robot can do only what it's told, and if its orders contain nonsense, it will grind out nonsense. In a difficult problem these preliminaries call for a very high order of thinking, which is one reason why both Aiken and Watson insist that no robot will ever replace human brains.
The simplified instructions are next translated into punched-hole code, transferred to the cards, and thence to the creature's input system. The switches are flipped - a process which automatically sets up paths of current to the cells. What the punched-card language says goes something like this:
"Store the number 268 in Memory Cell I. Store the number 64 in Memory Cell II. Now take 268 to the Multiplying Unit and 64 to the Multiplicand Unit. Multiply them. Some robots - like Eniac - have built-in multipliers wired to give the product of any two digits; otherwise the robot will add 268 six times, 268 four times, shift the second result over one space in the cell, and add. Deliver the answer to Memory Cell III, then to the printer."
When tussling with a really tough problem, the robot frequently chooses between alternative methods of procedure, for there are more ways than one of skinning a mathematical cat. Its instructions may have said: "If the third intermediate result is bigger than a million, add; if smaller, subtract." If a robot needs a logarithm, it may look it up in its permanent memory, just as a schoolboy consults his book of tables. Eniac, however, computes all logarithms from scratch - it can do it faster that way.
Do the robots pull boners? Lots of them. In fact, two days running without a slip-up is about the record. Tubes weaken, wires short-circuit. A moth once fluttered into Mark II and raised hob with its calculations until the frantic engineers could locate the saboteur. A burned-out tube may produce serious mistakes, but seldom a total breakdown. Usually the robot can correct such mistakes itself, always assuming the proper instructions have been issued to it in advance. One way is by performing all operations in duplicate. If the two sets of results fail to check at any point, a new path of current is set up, causing the robot to retrace its steps and start over from the last checked point. Should the same mistake recur, it may then stop altogether, flash red lights, ring bells, blow horns and otherwise indicate distress until the defective part has been repaired.
The history of man's attempts to invent machines to count for him is millenniums old. The abacus was in use 2500 years ago. It was the ancestor of all digital calculators, as the slide rule, developed in the seventeenth century by a succession of English mathematicians, anticipated the analog machines.
The first calculator to perform a series of operations without human aid, other than its original instructions, however, was conceived more than 100 years ago by a strange, obsessed Cambridge University professor, Charles Babbage. He worked on the design of two machines. His first was the "difference engine," which used, twenty-six digits and was to be used in computing mathematical tables. A considerable portion of this calculator was built, but it was abandoned and Babbage went on to the design of a more ambitious project, the" analytical engine," which was to use punched cards. Design of this second engine was carried out in elaborate detail, but Babbage died before construction was started, and it too was abandoned long before completion. To help him in his work, the British Government granted him substantial sums. In addition, he spent $50,000 of his own, gave up his chair of mathematics at Cambridge, and wrecked his health with overwork. But neither the technical skills nor the materials available in that pre-electronic age were up to the task. Babbage died, broke and disappointed, and the march of the caculating robots slowed to a standstill.
In 1936, a rangy, sharp-eyed young Harvard physicist named Howard Aiken stumbled across some of the forgotten writings of Babbage, and promptly fell in love with the idea of "difference engines." He longed to build one himself, but he could find no backers. His determination hardened, however, when he read this appeal in Babbage's Passages from the Life of a Philosopher:
If, unwarned by my example, any man shall attempt so unpromising a task and shall succeed in constructing an engine embodying in itself the whole of the executive department of mathematical analysis, I have no fear of leaving my reputation in his charge, for he alone will fully be able to appreciate the nature of my efforts and the value of their results.
Aiken knew at once that he was that man, and through him the reputation of "Old Babbage," as he affectionately refers to him, recovered its luster. For further study convinced Aiken that the Englishman had discovered the fundamentals of calculating machinery; only the construction techniques had eluded him. "If Old Babbage had lived another fifty years," Aiken says today, "there wouldn't have been much left for me to do."
It was Watson of IBM, with his long experience in manufacturing business machines, who made the ancient dream possible. IBM scientists, in collaboration with Aiken, provided the mathematical knowledge, its engineers the production know-how, and by 1944 they completed the world's first large-scale automatic calculator. Watson presented it to Harvard, where it was immediately put to work on problems for the Navy, which had meantime commissioned Aiken a commander.
Having since built Mark II and Mark III and set his sights on a Mark IV, Aiken reports that no more robots will be built by his laboratory. "It's time for United' States industry to take over and start producing in quantity," he says.
Already in other laboratories and some commercial plants new robots are being geared to perform feats that will make their predecessors seem like fumbling slowpokes.
In Philadelphia Mauchly and a scientist, J. Presper Eckert, are now building a total of six identical computers for use by such varied organizations as the U. S. Census Bureau, the Prudential Insurance Company and a market-research firm in Chicago.
At the Institute for Advanced Study in Princeton, engineers under the direction of Prof. John von Neumann, one of the world's foremost mathematicians and the No: 1 authority on the laws of probability, are rushing to completion a robot playfully nicknamed "The Maniac" which they expect to forecast weather with a speed and accuracy hitherto undreamed of. Like robot-directed gunfire, weather prediction is based on mathematical probability, the margin of error being narrowed in ratio to the quantity of data that can be' collated. The weather everywhere, past and present, predetermines tomorrow's weather in Chicago. Meteorologists have long understood this relationship and had access to a good deal of the data. Reports pour into the national Weather Bureau in Washington, for example, from some 4000 widely scattered stations at the rate of 600,000 figures a day. But by the time all of it could be mathematically related, tomorrow's weather - in fact, next year's weather - would have come and gone. With the limited data weathermen do have time to assess, they can now forecast only about three days ahead with 60 per cent accuracy. The Maniac should be able to forecast a week ahead with 90 per cent accuracy, and take no more than sixteen hours to do it.
At MIT, meanwhile, the more Wiener studies the robots the more they look like human brains to him. Upon this observation he has erected an elaborate edifice of theory about both brains and machines which some of his colleagues dismiss as a Buck Rogers fantasy and others acclaim as one of the most valuable and exciting ideas of the century. Wiener terms it cybernetics - from a Greek word meaning "steersman" - and he defines it as "control and communication in the animal and the machine."
"Man," he says, "has created these machines in his own image. Since he intended them to replace some of his own functions, it is not surprising that they duplicate some of his own mechanisms. Just as a derrick is a mechanized muscle, so a calculating machine is a mechanized thought process to deal with mathematics."
There is no reason why, Wiener insists, that, in addition to reading, remembering, choosing, correcting their own mistakes, looking up tables, the robots should not develop conditioned reflexes and even learn from experience. He extends his analogy to include "nervous breakdowns." When memory impulses in a man, such as anxiety, fear or guilt, get out of hand and invade the whole brain, preventing it from thinking about anything else, the man is said to be insane. Wiener maintains that robots go insane in very much the same way. An electrical impulse may overshoot the mark and circulate uncontrollably through the whole system. To cure certain forms of insanity in humans, surgeons sometimes excise a portion of the brain, sometimes try to shock the patient back to normality with electricity or drugs. Similarly, says Wiener, when a robot runs amok, its engineers may disconnect part of it or clear its over-burdened circuits by shooting powerful electric currents through it.
The cyberneticians further point out that calculators need not be confined to calculating. They could also operate entire factories. By attaching to them strain gauges, pressure valves and other instruments, mathematical values could be transmuted directly into manufacturing processes. Something like that happens in many a hydroelectric plant situated in areas too remote for easy human access. Such plants regulate their own water height; when in danger, automatically signal the fact. Even Aiken, who rejects the cybernetic theory in toto, says, "The ultimate goal of calculating machines is to design other machines."
The Frankenstein's-monster threat to human security and welfare which Wiener sees in this picture is manifold: if the robots could be used as tools to manipulate a national economy wisely, they could also, in the hands of greedy individuals or totalitarian governments, be used as deadly weapons. It is perfectly conceivable to Wiener that industrial markets might be scientifically rigged, enterprises wrecked, personal liberties curtailed with an efficiency to make a Hitler, Mussolini or Stalin blush.
On the socioeconomic level he warns, "The first industrial revolution, the revolution of the 'dark satanic mills,' was the devaluation of the human arm by the competition of machinery. There is no rate of pay at which a United States pick-and-shovel laborer can live which is low enough to compete with the work of a steam shovel as an excavator. The modern industrial revolution is simply bound to devaluate the human brain at least in its simpler and more routine decisions. Of course, just as the skilled carpenter, the skilled mechanic, the skilled dressmaker have survived in some degree the first industrial revolution, so the skilled scientist and the skilled administrator may survive the second. However, taking the second revolution as accomplished, the average human being of mediocre attainments or less has nothing to sell that it is worth anyone's money to buy."
The men who build the robots do not share these terrors. Far from destroying jobs, they testify, they will create new ones by the hundreds of thousands, just as the industrial revolution eventually did. Moreover, most of the robot builders would make book that in time "thinking machinery" will bring about a happier, healthier civilization than any known heretofore. What the odds on Utopia ought to be, however, not even the robots themselves can estimate.
See all articles from the Saturday Evening Post.