Big Dimwits and Little Geniuses

By Frederic Golden

Yesterday's klutzy machines have become today's micromarvels

The first electronic digital computer in the U.S., unveiled at the University of Pennsylvania in 1946, was a collection of 18,000 vacuum tubes, 70,000 resistors, 10,000 capacitors and 6,000 switches, and occupied the space of a two-car garage. Yet ENIAC (for Electronic Numerical Integrator and Calculator) was, in retrospect, a dimwit. When it worked, it did so only for short bursts because its tubes kept burning out. Built to calculate artillery firing tables, the half-million dollar ENIAC could perform 5,000 additions or subtractions per second. Today almost any home computer, costing only a few hundred dollars, can outperform poor old ENIAC as a "number cruncher."

Computer designers have obviously come a long way. But behind their spectacular achievements is a colorful history, one involving so many characters, so many innovations and such wrenching efforts that no single person or even country can claim authorship of the computer.

In a sense, humans have been computing--manipulating and comparing numbers or anything that they may represent--since they first learned how to count, probably with pebbles (the word calculus stems from the Latin for stone). At least 2,500 years ago, the Chinese, among others, discovered that they could handle numbers more easily by sliding little beads on strings. Their invention, the abacus, is still in use.

In 1642, perhaps pained by the long hours his tax-collector father spent doing sums, a 19-year-old French prodigy named Blaise Pascal made an automatic device that could add or subtract with the turning of little wheels. But the clerks who spent their lives doing calculations in those days viewed Pascal's gadget as a job threat, and it never caught on. A short time later, the German mathematician Gottfried Wilhelm Leibniz added the power of multiplication and division. Said he: "It is unworthy of excellent men to lose hours like slaves in the labor of calculations..."

But such mechanical contrivances were no more than calculators. They could only do arithmetic, and very clumsily at that. The first man to conceptualize a true computer, one that would be able to do math and much more, was an irascible 19th century English mathematician named Charles Babbage. Incensed by the inaccuracies he found in the mathematical tables of his time, the ingenious Babbage (father of the speedometer, the cowcatcher for locomotives and the first reliable life-expectancy tables) turned his fertile brain to creating an automaton that could rapidly and accurately calculate long lists of functions like logarithms. The result was an intricate system of gears and cogs called the Difference Engine.

Babbage managed to build only a simple model because the craftsmen of the day were unable to machine the precise parts required by the contraption. But the temperamental genius soon had a bolder concept. He called it the Analytical Engine. Even more complex than its predecessor, it had all the essentials of a modern computer: a logic center, or what Babbage called the "mill," which manipulated data according to certain rules; a memory, or "store," for holding information; a control unit for carrying out instructions; and the means for getting data into and out of the machine. Most important of all, its operating procedures could be changed at will: the Analytical Engine was programmable.

Babbage worked obsessively on his machine for nearly 40 years. Presumably he was the world's first computer "nerd." Until his death in 1871, he ground out more and more sketches. The Analytical Engine became hopelessly complicated. It required thousands of individual wheels, levers and belts, all working together in exquisite precision. Few people understood what he was doing, with the notable exception of Lord Byron's beautiful and mathematically gifted daughter, Ada, the Countess of Lovelace, who became Babbage's confidante and public advocate. When the government cut off funds for the Analytical Engine, she and Babbage tried devising a betting system for recouping the money at the track. They lost thousands of pounds.

The Analytical Engine was never built. It would have been as big as a football field and probably needed half a dozen steam locomotives to power it. But one of its key ideas was soon adapted. To feed his machine its instructions, Babbage planned to rely on punched cards, like those used to control color and designs in the looms developed by the French weaver Joseph Marie Jacquard. Ada poetically described the scheme this way: "The Analytical Engine weaves algebraical patterns just as the Jacquard loom weaves flowers and leaves."

In the U.S., a young engineer named Herman Hollerith persuaded the Census Bureau to try the punched-card idea during the forthcoming 1890 census. Such personal information as age, sex, marital status and race was encoded on cards, which were read by electric sensors, and tabulated. Hollerith's equipment worked so well that the Census Bureau's clerks occasionally shut it off to protect their sinecures. Soon punched cards were widely used in office machinery, including that made by a small New York firm that absorbed Hollerith's company and became International Business Machines.

Babbage's dream of a true computer--one that could solve any number of problems--was not realized until the 1930s. In Hitler's Germany, an obscure young engineer named Konrad Zuse, using the German equivalent of an Erector set for parts and his parents' living room as his workshop, built a simple computer that could perform a variety of tasks; its descendants calculated wing designs for the German aircraft industry during World War II. At Bell Telephone Laboratories in the U.S., the research arm of AT&T, a mathematician named George Stibitz built a similar device in 1939 and even showed how it could do calculations over telephone wires. This was the first display of remote data processing. During the war a British group, putting into practice some of the ideas of their brilliant countryman Alan Turing, built a computer called Colossus 1 that helped break German military codes. The British, German and U.S. machines all shared a common characteristic: they were the first computers to use the binary system of numbers, the standard internal language of today's digital computers.

In this they departed from Babbage's "engine." The engine was designed to count by the tens, or the decimal system. Employing ten digits (0 to 9), the decimal system probably dates from the time when humans realized they had ten fingers and ten toes. (Digit comes from the Latin for finger or toe.) But there are other ways of counting as well, by twelves, say, as in the hours of the day or months of the year (the duodecimal system). In the binary system, only two digits are used, 0 and 1. To create a 2, you simply move a column to the left, just as you do to create a 10 in the decimal system. Thus if zero is represented by 0 and one by 1, then two is 10, three 11, four 100, five 101, six 110, seven 111, eight 1000, and so forth.

The binary system is enormously cumbersome. Although any number can be represented, it requires exasperatingly long strings of Os and 1s. But putting such a system to work is a snap for digital computers. At their most fundamental level, the computers are little more than a complex maze of on-off switches that reduce all information within the machine to one of two states: yes (1) or no (0), represented either by the presence of an electrical charge at a particular site or the absence of one. Accordingly, if in a row of three switches, two of them are in an on position (11) and the other off (0), they would represent the number six (110).

In the world of digital computers, each of these pieces of information is called a bit (for binary digit). In most personal computers, bits are shuttled about within the machine eight at a time, although some faster 16-bit machines are already on the small-computer market and even speedier 32-bit machines are in the offing. Clusters of eight bits, forming the equivalent of a single letter in ordinary language, are called bytes. A typical personal computer offers users anywhere from about 16,000 bytes of memory (16K) to 64,000 (64K). But that figure is climbing fast. A few years ago, the standard memory chip, a quarter-inch square of silicon, was 16K. Today it is rapidly becoming 64K, and the industry is already talking of mass-producing 256K chips.

The novel idea of using strings of Is and Os to solve complex problems traces back to another gifted Englishman, George Boole. A contemporary of Babbage's, he developed a system of mathematical logic that allows problems to be solved by reducing them to a series of questions requiring only an answer of true or false. Just three logical functions, called AND, OR and NOT, are needed to process Boole's "trues" and "falses," or Is and Os. In computers these operations are performed by simple combinations of on-off switches, called logic gates. They pass on information, that is pulses of electricity, only according to the Boolean rules built within them. Even a small home computer has thousands of such gates, each opening and closing more than a million times a second, sending bits and bytes of information coursing through the circuitry at nearly light's velocity (electricity travels about a foot in a billionth of a second).

The earliest digital computers were much more plodding. They relied on electromechanical on-off switches called relays, which physically opened and closed like the old Morse code keys. Physicist-Author Jeremy Bernstein recalls that Mark I, IBM's first large computer, assembled at Harvard during World War II, sounded "like a roomful of ladies knitting." It could multiply two 23-digit numbers in about five seconds. Even some hand-held calculators can now do the same job in a fraction of the time.

ENIAC vastly increased computer speed by using vacuum tubes rather than electromechanical relays as its switches, but it still had a major shortcoming. To perform different operations, it had to be manually rewired, like an old wire-and-plug telephone switchboard, a task that could take several days. The Hungarian-born mathematical genius, John von Neumann, saw a solution. He suggested putting the machine's operating instructions, or program, within the same memory as the data to be processed and writing it in the same binary language. The computer could thus be programmed through the same input devices used to feed in data, such as a keyboard or a reel of tape. The first commercial computer to have such capability was Sperry-Rand's UNIVAC 1, which appeared in 1951 and, much to IBM's chagrin at being beaten, was promptly delivered to the Census Bureau.

Yet even while journalists were hailing the new "electronic superbrains," the machines were already becoming obsolete. In 1947 three scientists at Bell Labs invented a tiny, deceptively simple device called the transistor (short for transfer resistance). It was nothing more than a sandwich of semiconducting materials, mostly crystals of germanium; silicon became popular later. The crystals were arranged so that a tiny current entering one part of the sandwich could control a larger current in another. Hence, they could be used as switches, controlling the ebb and flow of electrons. Even the earliest transistors were much smaller than vacuum tubes, worked faster and had fewer failures. They gave off so little heat that they could be packed closely together. Above all, they were quite cheap to make.

Within a few years, the wizards at Bell Labs built the first fully transistorized (or solid-state) computer, a machine called Leprechaun. But by then Ma Bell, eager to avoid the wrath of the Justice Department's trustbusters, had sold licenses for only $25,000 to anyone who wanted to make transistors, and the scramble was on to profit from them. William Shockley, one of the transistor's three inventors, returned to his California home town, Palo Alto, to form his own company in the heart of what would become known as Silicon Valley. In Dallas, a young, aggressive maker of exploration gear for the oil industry, Texas Instruments, had already hired away another Bell Labs star, Gordon Teal, and was churning out the little gadgets. So were old-line tube makers such as General Electric, RCA, Sylvania and Raytheon. Much of their production went to the Pentagon, which found transistors ideal for a special computing task: the guidance of missiles.

The first computers, even those built with transistors, were put together like early radios, with tangles of wires connecting each component. But soon electronics manufacturers realized that the wiring could be "printed" directly on a board, eliminating much of the hand-wiring. Then came another quantum leap into the miniworld. In the late 1950s, Texas Instruments' Jack Kilby and Fairchild Semiconductor's Robert Noyce (one of eight defectors from Shockley's firm whom he scathingly called the "traitorous eight") had the same brainstorm. Almost simultaneously, they realized that any number of transistors could be etched directly on a single piece of silicon along with the connections between them. Such integrated circuits (ICs) contained entire sections of a computer, for example, a logic circuit or a memory register. The microchip was born.

Designers kept cramming in more and more transistors. Today, hundreds of thousands can be etched on a tiny silicon chip. The chips also began incorporating more circuits. But even such so-called large-scale integration had a drawback. With the circuits rigidly fixed in the silicon, the chips performed only the duties for which they were designed. They were "hardwired," as engineers say. That changed dramatically in 1971, when Intel Corp., a Silicon Valley company founded by Noyce after yet another "defection," unveiled the microprocessor. Designed by a young Intel engineer named Ted Hoff, it contained the entire central processing unit (CPU) of a simple computer on one chip. It was Babbage's mighty mill in microcosm.

With the microprocessor, a single chip could be programmed to do any number of tasks, from running a watch to steering a spacecraft. It could also serve as the soul of a new machine: the personal computer. By 1975 the first of the new breed of computers had appeared, a hobbyist machine called the Altair 8800 (cost: $395 in kit form, $621 assembled). The Altair soon vanished from the marketplace. But already there were other young and imaginative tinkerers out in Silicon Valley getting ready to produce personal computers, including one bearing an odd symbol: an apple with a bite taken out of it. Suddenly, the future was now. --By Frederic Golden This file is automatically generated by a robot program, so viewer discretion is required.