When von Neumann, Goldstine, and Burks began on the IAS computer, von Neumann asked RCA (nearby in Philadelphia) to develop a tube that could be used for memory storage. They did, calling their product the Selectron, but the tubes took too long to develop—they were expensive and complicated—so by the end of 1948 von Neumann had decided to adopt Williams tubes.
Another issue von Neumann and his team addressed was that of translation. Just as Atanasoff had realized in 1939 that not every mathematician was comfortable with base-two numbers, and so the results put out by the ABC were automatically translated into decimal numbers, von Neumann realized that the more powerful and useful a computer might become, the more essential a translating mechanism for input and output would be. And von Neumann wanted his computer to do more than solve math problems—he also wanted it to be able to use language (like Colossus, which could decipher a code more easily than it could perform a large multiplication problem—and we will never know whether von Neumann’s friends on the Colossus project ever chatted with him about what they had done). Unable to get Eckert, von Neumann hired an engineer named Julian Bigelow to put together the IAS computer, thinking that the project would take ten people about three years.
But von Neumann could not work with Bigelow, who, he felt, tended to go down blind alleys, trying things without a good sense ahead of time of how those ideas would work. And Norbert Wiener turned out to be correct about the lack of receptivity at the IAS toward the computer project. It was housed in a boiler room and then an outbuilding, and even then there were complaints about it from the other scholars. Work that was farmed out went to corporations that didn’t know what was really wanted. Von Neumann himself was an ideas man, not a technology man (though when his wife declared that he could not handle a screwdriver, she added that he was good at fixing zippers). Adding to these difficulties, after January 1950, once Truman gave the go-ahead, von Neumann was hard at work on the hydrogen bomb, work that accelerated through 1950, when Edward Teller’s first ideas were proven wrong, and into 1951, when Teller and Stanislaw Ulam came up with an idea that worked. Through both these phases of H-bomb development, the IAS computer did produce necessary calculations, especially after James Pomerene was installed to replace Bigelow. One can only wonder how the construction of the computer would have gone if John Vincent Atanasoff had been allowed to bring his exceptional improvisational talents to it—but perhaps from their conversations, von Neumann understood that in addition to being difficult to work with, Atanasoff had an even greater claim to the computer concepts von Neumann wanted to utilize than Mauchly and Eckert did, and, having experienced what he considered to be Mauchly and Eckert’s greed, he did not want to risk that possibility again.
In 1948, a member of Mauchly and Eckert’s business team, George Eltgroth, a patent attorney, was approached by a racetrack owner about using computers to break the monopoly of the American Totalizer Company over bookmaking at American racetracks. Eltgroth saw his chance and went to American Totalizer itself. He found a willing partner in Henry Straus, vice president of the tote company—Straus oversaw the investment of $550,000 into UNIVAC—a $62,000 loan and $488,000 for 40 percent of the company stock. But Mauchly’s payroll continued to expand—by 1949, there were 134 employees—while the contracts kept contracting. At one time, Mauchly had orders for six UNIVACs, but he had received only $150,000 apiece for the machines, and UNIVAC was still not completed. And then, in November 1949, Henry Straus was killed in a plane crash, and American Totalizer asked for their investment back—now worth $432,000. Eckert and Mauchly then approached IBM. Thomas J. Watson, Sr., later said that he wasn’t impressed by Mauchly, but it also turned out that, according to IBM lawyers, antitrust laws forbade IBM from acquiring UNIVAC.
In early 1950, Mauchly and Eckert’s company was denied security clearance and therefore banned from accepting top-secret military contracts—a significant portion of those available to private industry. The reasons for the denial of clearance were a mix of anti-Communist paranoia (a member of the engineering team had supported Henry Wallace; Mauchly himself had signed a petition in 1946 supporting civilian control of nuclear energy) and general suspicion—army intelligence asked the FBI to investigate the drowning of Mary Mauchly, which it did, exonerating Mauchly. A few weeks after the denial of security clearance, Remington Rand bought the Eckert-Mauchly Computer Corporation. They paid off the debt to American Totalizer and gave Eckert and Mauchly $100,000 for the remaining 60 percent of the stock, which included the ENIAC patents. The two principals also got a guaranteed $18,000 per year salary and 5 percent of the yearly profits for eight years, should any profits accrue. Thirteen months later, UNIVAC was finally working.
The first UNIVAC, which had been assembled on the second floor of the Eckert-Mauchly building, an old knitting factory, weighed 29,000 pounds and covered 380 square feet of floor space. It used 5,200 vacuum tubes (less than a third of the number in ENIAC) and consumed 125 kilowatts of electricity (as much as 1,250 100-watt lightbulbs, about 16 percent less than ENIAC). The mercury delay line memory was made up of large horizontal cylinders containing liquid mercury that circulated acoustic vibrations representing stored instructions and other data. The external memory, or ROM, was stored on either magnetic tape or punch cards.
Some difficulties with the manufacture of the first UNIVAC arose almost at once—the Eckert-Mauchly building was not air-conditioned and could get so hot in a Philadelphia summer that tar from the roof would melt onto the computer through the ceiling. In fact, no thought had been given to the computer’s environment—holes were cut in the walls for summer ventilation that then made the vast room impossible to heat in the winter. And, a serious drawback for a commercial venture, the machine could not be delivered—it was too complex and delicate to be quickly disassembled. At any rate, Mauchly (and Remington Rand) wanted to use the first one for demonstrations only in order to gain more contracts.
But eventually, forty-six UNIVAC I computers were manufactured, sold, and delivered to such companies as Metropolitan Life Insurance, Westinghouse, and U.S. Steel, as well as to government agencies: the Army Map Service (one of the original contracts), the Pentagon, and the Census Bureau (though this one stayed at company headquarters and was operated there). Although Mauchly had charged only $159,000 for the computer in the first contracts, the price eventually rose by almost a factor of 10. UNIVAC I gave way to UNIVAC II in 1958.
In 1951, like Mauchly and Eckert, Atanasoff decided to go into private enterprise, but unlike them, he first mastered the basic principles of accounting (which took him three days) and of business law (about a month). He wrote his own articles of incorporation and lured some of his fellow researchers away from the NOL. The plan was to offer testing services, especially to the military—the cold war meant that there were lots of military contracts, and they were lucrative. He set up his offices in Frederick, Maryland, which he chose after studying the weather patterns in the Washington, D.C., area and deciding that, should there be an atomic attack, Frederick would be outside of the radiation plume, and therefore somewhat safer than his first location of choice, Rockville. In Frederick, he had his corporate headquarters built and equipped with what he considered to be the best supplies for protecting and cleaning the building in the event of an attack—a neoprene-coated roof, sheets of plywood to protect the windows, and boxes of Tide detergent for spraying on the building.
With his usual confidence and frugality, Atanasoff used his own savings as capital for his business, along with investments of those who would be working with him. According to Tammara Burton, the company, which operated on military contracts, was always solvent and never had to borrow money. Atanasoff now focused on his company and deliberately ignored what was going on in the world of computers. The testing Atanasoff’s company performed ranged from determining how a projectile might approach and strike an airplane in flight to figuring out how best to drop leaflets on a populated area as a form of psychological warfare (the army ga
ve him this contract during the Korean War). Though the company was successful, entrepreneurial life was taxing in some ways—Atanasoff later recalled, “I have a great deal of affection for the men who are associated with me and we generally understood each other pretty well, but nevertheless they regarded me as a kind of a harsh director, always attempting to advance the work at all times of the day and night … I found this discipline severe.”
In February 1951 the first Ferranti-manufactured Mark I, the computer developed at the University of Manchester, was delivered to the new university computer lab. According to Andrew Hodges (and this is important for the development of the computer as we know it), “In many ways, [because of Turing’s lack of interest in the project], the Computing Laboratory remained as secret as Hut 8,” restricting the public relations potential, and therefore sales, of the Manchester computer. EDVAC and UNIVAC dominated the news.
In March of the same year, Alan Turing was elected to the Royal Society, but then, in January 1952, Turing met a young man named Arnold Murray. Turing was now almost forty, Murray was nineteen. Turing cultivated the acquaintance, and Murray bragged about it to a friend. The unfortunate result was that the friend broke into Turing’s house outside of Manchester and stole some of Turing’s possessions. Murray managed to get some of the things back from the friend, but by this time, Turing had already reported the burglary. His report alerted the police, who, upon uncovering an illegal homosexual relationship between Turing and Murray, arrested Alan Turing under the draconian Labouchere Amendment to the Criminal Law Amendment Act 1885 (Section 11), which stated that “any male person who, in public or private, commits any act of gross indecency with another male person shall be guilty of a misdemeanour, and being convicted thereof shall be liable at the discretion of the court to be imprisoned for any term not exceeding two years, with or without hard labour,” the same law that had been used to prosecute Oscar Wilde.
In his usual unashamed fashion, Turing detailed the nature of his relationship to Murray (he had never been ashamed of his homosexuality, nor had he ever shown caution in expressing himself on any subject). In early April 1952, Turing was convicted of “gross indecency” and given a choice between a year in prison and a year of drug therapy designed to inhibit his sexual desires—a course of estrogen shots (chemical castration). Although such a conviction meant, in the cold war atmosphere of the 1950s, that Turing could no longer work for the British government. His friends felt that he was unrepentant about what had happened—under security surveillance (which he knew about), Turing went to Norway, where he had heard that there were venues for all-male dancing. The letters he wrote to his friends were often bemused and, apparently, lighthearted, though not uniformly so. In a 2009 article in the Daily Mail discussing what sort of posthumous honors Turing might receive for his intelligence work during World War II, Geoffrey Wansell points out that the estrogen “transformed his body. The man who had run a marathon in 2 hours and 46 minutes—when the world record was 2 hours and 25 minutes—was reduced to a shadow of his former self. ‘They’ve given me breasts,’ he was reported to have said to a friend, describing the shameful process as ‘horrible’ and ‘humiliating.’ ”
Through 1952 and 1953, Turing engaged in more travel and more work on his theories of brain as machine/machine as brain. And then, on June 8, 1954, Alan Turing was found by his housekeeper, dead of cyanide poisoning in his house in Manchester, a half-eaten apple by his bedside (he customarily ate an apple before going to bed). There was no suicide note.
Turing’s mother never believed that he had committed suicide—she thought that he had died accidentally, as a result of a careless chemistry experiment. Others pointed out that as a convicted homosexual who liked to travel abroad and make contact with young men, he was seen by the British security services as not only a risk, but a growing risk, since the cold war was escalating quickly. Turing was highly knowledgeable about Colossus and all sorts of other state secrets, and now he was a convicted but unrepentant homosexual who was associated with King’s College, which, along with Trinity College, was considered to be a hotbed of Soviet spies (Guy Burgess and Donald Maclean, who had defected to the Soviet Union in 1951, had been at Trinity College in the thirties and were also homosexuals). Some people continue even in 2010 to feel that he was assassinated, with a “suicide” staged by British security. Or perhaps they had simply invited him to commit suicide. Friends remembered Turing wondering aloud about methods for committing suicide—they thought at the time that he was merely engaging in one of his frequent thought experiments. Others have suggested that, thanks to his gross indecency conviction and to his unorthodox ideas, Turing was at the end of his career and knew it. In any event, he died in obscurity, thirty years before either his role in World War II cryptanalysis or his role in the invention of the computer would emerge.
1. A computer engineer in England suggested using a delay line with the cylinders filled with gin.
Chapter Ten
By the early 1950s, three computers had made their way into the marketplace.
In England, a second Ferranti Mark 1 was ordered for the Atomic Energy Research Establishment, near Oxford, to be delivered in 1952. But after the Labour government headed by Clement Atlee was thrown out in October 1951, the new Tory government, headed by Winston Churchill, canceled all government contracts worth more than a hundred thousand pounds, and so the second Ferranti machine was never completed. Work on the computer was halted, and it was later bought for very little by the University of Toronto. However, seven other Ferranti computers (of a slightly different design) were sold, one to Shell Labs in Amsterdam. But it was not only expense that killed the development of computers in England, it was also vigilant secrecy. According to Kirwan Cox, the Canadian filmmaker, because Churchill had found himself quoted in Mein Kampf about how England had won the First World War, he “became paranoid about information that had enabled the British victory getting out again.” Presumably, the enemy to be wary of was now the Soviet Union.
There was much more money and much more self-promotion in the United States. In March 1951 UNIVAC became available, and in 1952 the IBM 701 was unveiled at the end of April. The 701 was an offshoot of von Neumann’s IAS computer. Like the IAS, it used Williams tubes for memory (72 in one version, 144 in another). It was intended for use as a scientific calculator (and had been known while in development as the “Defense Calculator”). The 701 was joined by the 702, the 650, and the 705. The 701 and the 650 were designed for business use; IBM seemed destined to consolidate a share of what was turning out to be an actual market, but then, in the November presidential election between Dwight D. Eisenhower and Adlai Stevenson, the UNIVAC scored a big public relations victory when it predicted the outcome for CBS based on early returns. The PR coup might have been designed by an advertising agency—at first the UNIVAC’s predictions looked so out of whack that network operators fiddled with them in order to avoid embarrassment, but then the network had to admit even greater embarrassment—the original unfiddled predictions turned out to be very close to the actual results of the election. When CBS revealed what had happened on the air, UNIVAC became the face of the computer in the 1950s public imagination, and the result for Remington Rand was more sales, this time lucrative ones, to companies rather than to the government.
IBM had two commercial advantages, though: one was the punch-card system that many offices already had in place, and the other was the business model, which focused upon leasing and service rather than outright sales. It looked as though IBM was to dominate the business market and foil von Neumann’s plan for the computer to be based upon common intellectual property rather than proprietary patents.
But von Neumann’s dissemination of the ideas behind ENIAC meant that there were people working on designing and building computers all over the United States—challenges to the original ENIAC patents by Control Data, Honeywell, Burroughs, General Electric, RCA, and National Cash Register began almost immediately, and th
ey meant that the ENIAC patents (which made more than a hundred proprietary claims) were slow to be awarded to Remington Rand, who had obtained them when they bought out Mauchly and Eckert.
In October 1953, Pres Eckert published an article on computer memory in the Journal of the Institute of Radio Engineers in which he knowledgeably described the structure and the function of the ABC’s memory system and also expressed admiration for its frugality: “There may have been similar systems prior to Atanasoff’s, but none was as inexpensive to construct.” Eckert’s article served to motivate the patent department at IBM, which, like the smaller companies, had come to believe that Eckert and Mauchly’s ENIAC patents might be broken. Clifford Berry learned that IBM was looking for information about “capacitor drum storage devices,” or, as Atanasoff had called his invention, “regenerative memory.” Berry’s work on the ABC was known at Consolidated Engineering, his place of business in Pasadena, and what the IBM representative learned from a lawyer in the patent office at Berry’s company was the subject of an IBM in-house memorandum of September 30, 1953—Consolidated Engineering planned to visit Iowa State and look into Berry’s claims. IBM decided to collaborate on this investigation. The Consolidated Engineering patent attorney also informed IBM that “he had heard rumors that Burroughs, National Cash, and IBM were planning, as part of a team, to form a patent pool, particularly with a view of fighting the Eckert-Mauchly patents.” Kirwan Cox believes that the sequence of events was slightly different—Berry saw Eckert’s article, read the patent, and told his employer that the patent was based on the prior art of the ABC. Consolidated Engineering was already doing business with IBM, and so contacted IBM about the apparent patent infringement. The younger Thomas J. Watson, much more interested in computers than his father had been, was eager to circumvent the ENIAC patents. In April 1954, a representative from IBM interviewed Clifford Berry in California. On June 14, when he visited Atanasoff in Frederick, Maryland, the IBM representative, a man named A. J. Etienne, even said, “If you will help us, we will break the Mauchly-Eckert computer patent; it was derived from you.”