It is clear that what he had in mind building was something very like the theoretical model in his Computability paper, the model we now call a Turing machine. It worked on one bit at a time, but used a huge amount of memory to do anything of consequence. Since he had proved that anything that was computable could be theoretically computed on such a simple device, why not build one? The CPU would only have required a handful of vacuum tubes. But such a machine is horrendously difficult to program, and even at electronic speeds, it would have been painfully slow for many simple things like floating-point arithmetic.2
One of Turing’s difficulties (or Womersley’s, as his director) was that he didn’t mind talking to the press (either the general press or journals of particular groups, such as the Institution of Radio Engineers), but when he did talk, he raised hopes that did not seem realistically capable of fulfillment, and he was often met with skepticism. And he himself met with skepticism, owing to his odd manner and excessively casual (or, you might say, sloppy) mode of dress, which hadn’t changed much since his older brother had despaired of getting him into his sailor suit with his shoes on the proper feet in 1916. When Turing himself gave a few lectures on the proposed NPL computer, Wilkes attended only one. He felt that Turing’s ideas were irrelevant, because they “were widely at variance with what the mainstream of computer development was going to be.” Womersley sent Turing to the United States to attend a computing symposium in January 1947, and then to visit Princeton and have a look at the project von Neumann had begun there as an academic alternative to the EDVAC. The trip did not change Turing’s mind about his own computer, but momentum was carrying the project away from his ideas. Delays were mounting along with the disagreements. In the meantime, Wilkes, who didn’t have to apply for funding, put his computer together very quickly. In February 1946, Turing had requested that Tommy Flowers, at Dollis Hill, build the prototype of the computer he was designing. Flowers promised the machine by August, but postwar repairs and improvements to the telephone system superseded the project, and by February 1947 the ACE was going nowhere because Turing could not persuade Womersley to commit himself to Turing’s ideas—for example, an engineering department was set up, but made no progress. Possibly, Womersley was the sort of administrator who thinks contradictory ideas constitute a backup plan, but in the end they constituted no plan at all because what had come to be called “von Neumann architecture”—the principles of computer design set out in the “First Draft”—were simply taking over by coming to seem tried and tested.3 Turing quit. In the autumn of 1947, he returned to Cambridge.
1. One reason that Zuse’s autobiography is interesting is that it gives Americans a perspective on life in Nazi Germany that we rarely get. Zuse seems perennially surprised by the power of the Nazis and the events he lives through. My interpretation is that this is a feature of his dedication to and focus on his machine—that thinking about it and building it simply occupied his mind almost completely and drove almost every other consideration, including mortality, out of his consciousness.
2. A Turing machine has been constructed. It can be seen on YouTube: http://www.youtube.com/watch?v=E3keLeMwfHY.
3. One computer that conformed to Turing’s ideas was built by two engineers from Tommy Flowers’s Colossus engineering team. It was called MOSAIC and was used during the cold war to calculate aircraft and anti-aircraft trajectories.
Chapter Eight
By the time Atanasoff was home from Bikini Atoll and finished with his aborted computer project for the navy, John Mauchly and Pres Eckert were deep into their own patent conflicts with the University of Pennsylvania. Originally, Penn had shown little or no interest in Mauchly’s project. As of late 1944, they had accorded Mauchly and Eckert patent rights if something were to come out of ENIAC—the university itself retained only a license to build and use a computer of their own. In late 1945, though, the university was rethinking this policy, and then, in 1946, the army lifted security restrictions on the machine. Penn informed Mauchly and Eckert that, having been constructed with public funds, ENIAC and its parts could not produce private patents. Mauchly and Eckert had put together their patent application in the fall. When ENIAC was unveiled, Herman Goldstine claimed in the publicity material to be one of the three inventors, along with Mauchly and Eckert. He also put his name on the patent application, but Mauchly and Eckert removed it. Goldstine was not pleased.
Plans for EDVAC, the computer that would replace ENIAC, were even more contentious. In 1941, Mauchly had mentioned in his October letter to Atanasoff that the inventor of the earlier Moore School analyzer, Irven Travis, had left to join the navy. Travis returned in 1946, and he was determined that the university would not relinquish any rights to any future machines. He stated point blank that “all people who wish to continue as employees of the university must turn over their patents to the university.” In an interview in 1977, Travis said, “Well, the record is clear. I’m the one who precipitated the blowup.” Travis felt that the fact that the research was being done under the sponsorship of the military and the university meant that individual researchers did not have property rights in the research. He also felt that the ENIAC patents as they were eventually submitted did not give sufficient credit to other members of the research team, especially Arthur Burks—whom Travis considered “brilliant.” In his interview with Nancy Stern, for the Charles Babbage Institute of the University of Minnesota, he also suggested that the president of Penn, an English professor, had given in to coercion on the part of Mauchly and Eckert, who had threatened not to complete ENIAC if they did not gain patent rights to the machine. The patent dispute quickly escalated, driven, according to Scott McCartney, on the part of Travis and other Moore School engineers by the feeling that Mauchly was an outsider (having not gotten his degrees at the Moore School), while Eckert was volatile and difficult to get along with (by this time nearly twenty-seven years old and having earned only a bachelor’s degree). Travis later said that there were researchers at the Moore School who were doing patentable research on their own and profiting from it and that there was no difficulty with that. The problem had entirely to do with research done on university time and funded by public money. Travis may also have been influenced by the design of ENIAC, which, according to Alice and Arthur Burks in their 1988 history of the Atanasoff computer, owed a great deal to his own modification of the Bush Analyzer. But Mauchly and Eckert felt that Travis’s position was unprecedented at Penn—other faculty members had gotten patents in the past. The university would not yield, and it issued an ultimatum: if Mauchly and Eckert did not give up their claims to patents on EDVAC, they could no longer be employed by the university. Mauchly and Eckert quit that day. One result was that what with the patent brouhaha and von Neumann’s 1945 publication of the principles behind the projected EDVAC, other inventors managed to get ahead of Penn, and when EDVAC was finally completed, it was no longer a first.
Von Neumann, now deeply interested in computers, invited Eckert, Burks, and Goldstine to go to Princeton and join the team at the Institute for Advanced Study, but he did not invite Mauchly (no doubt reflecting the general sense that Mauchly was something of “a space case”). Eckert declined the invitation, saying, according to McCartney, that expeditious and inexpensive production of computers, leading to widespread use, would be more beneficial than “perfect[ing] them in more detail for a long while in universities.” Goldstine and Burks went with von Neumann.
Thomas J. Watson, Sr., builder of Aiken’s Mark I, who had furnished the punch-card system for ENIAC, offered Mauchly and Eckert a lab and financing, though he was not convinced that computers were the wave of the future. Mauchly balked—he thought they had been overcharged for the punch-card machines. It was clear to both Mauchly and Eckert that they worked well together and should stick together. Later, Mauchly’s second wife remarked, “Mauchly was a dreamer. Without Eckert he would never have built a computer. But we often said that without Mauchly, Eckert wouldn’t have thought of
it.” In general, Mauchly never seems to have inspired confidence, wherever he went and whatever the circumstances.
Mauchly and Eckert were tempted to start their own company, and in the late spring of 1946, only a few months after the unveiling of ENIAC, they did. From the beginning, there were uncertainties and conflicts. The main difficulty was that they did not have much money. Pres Eckert’s developer father cosigned a loan for $25,000, but ENIAC had cost $400,000 to build. Investors were skeptical: Howard Aiken, to whom everyone seems to have turned for advice, thought that computers would always be specialty items—sales might run to five or six machines around the nation. Mauchly and Eckert cast about for government contracts and came up with a prospect—a contract with the National Bureau of Standards on behalf of the Census Bureau. Though George Stibitz, of Bell Labs, suggested a more conservative approach—giving Mauchly and Eckert an exploratory grant—the Bureau of Standards was enthusiastic and agreed to award the “Electronic Control Company” $270,000 (about $3.25 million in 2010 dollars), which would be paid out over two years.
In July and August 1946, the Office of Naval Research sponsored forty-eight lectures at the Moore School on the subject “The Theory and Techniques for the Design of Digital Computers.” Mauchly and Eckert had already signed contracts to give their lectures (eight1 and eleven, respectively) before leaving Penn, so they had to take time from their new company to participate. Three other lectures were delivered by Arthur W. Burks. The course was taking place while Atanasoff was at Bikini Atoll, and his naval computer project was terminated shortly after he got back. Clearly, the navy felt that his efforts would be redundant. It also seems clear that von Neumann (who was scheduled to give a lecture, though no record of the lecture exists) and Goldstine continued to use their influence to nudge computer theory and building techniques into the public domain, and that their efforts continued to bear fruit.
When the lecture series was finished, Mauchly and his wife, Mary, went to the Jersey shore for a vacation. They arrived late on the evening of September 8 and immediately went for a late-night dip. The surf was more treacherous than they had expected, and Mary seems to have been caught in a riptide. She could not save herself and he could not save her. Her body washed up on shore two hours later. Their son, who had accompanied John to Ames in 1941, was now eleven, and their daughter was now seven.
But the new corporation had to move forward. Mauchly and Eckert signed their contract with the Bureau of Standards at the end of September and wrote up a business plan. They got two other contracts—one from the Air Controller’s Office and another from the Army Map Service. Even so, the patents remained a contentious issue, in part, according to filmmaker Kirwan Cox, because “Mauchly was assigned the patent work, and he did it, but he did it slowly, which was a problem.” Cox also points out that the reason that EDVAC was an advance in ENIAC was that its inventors (including von Neumann) “seemed to realize they could get further by going back to ABC concepts. EDVAC was closer to the ABC than to ENIAC. ENIAC was a hybrid machine—partially ABC, partially Bush analyzer, and partially ganged calculators. As they were building it, the inventors realized that they could improve it—and did so by going back to ABC, to a binary counting system and regenerative memory.” According to John Gustafson, they also copied the use of capacitors arranged on a rotating drum.
In England, another attempt was being made to build on what had been learned through Colossus without acknowledging that Colossus had ever existed. The third important figure in the Bletchley Park computer story was Max Newman, Alan Turing’s old professor from Cambridge, from whom he had taken a course in the foundations of mathematics in 1935. It was as a result of Newman’s explications of Hilbert’s questions that Turing had begun to think of the search for mathematical truth as a question of “provability” and even as a “mechanical process” (Newman’s words), thereby conceiving his “On Computable Numbers” paper of 1936.
Max Newman was the only son of Herman Neumann, who had been born in 1864 in Bromberg, Germany (now Bydgoszcz, Poland), a town that passed back and forth between Poland and Prussia from 1346 until the end of the Second World War. Originally a fishing town, and then a trading town, Bromberg/Bydgoszcz came to have a large Jewish population. Like Max von Neumann of Pecs, Hungary, and Ivan Atanasov of Boyadzhik, Bulgaria, Herman Neumann emigrated to the west, in 1879, not to New York or to Budapest, but to London. There he trained as a bookkeeper, and, like John Atanasoff, at thirty-two he married a schoolteacher, twenty-six-year-old Sarah Pike. Max Neumann was born in 1897. Young Max, like young John Vincent Atanasoff, was publicly educated. World War I brought pain and disruption to the Neumann family—Herman was interned at the beginning, then released, but the experience was so grueling that he and Sarah changed the spelling of their name to “Newman.” Nevertheless, Herman returned to Germany after the war and was still there, separated from his family, when he died in 1926. In the meantime, Max attained a scholarship to St. John’s College, Cambridge, in 1915, and though his studies were interrupted by the war (he served as an army paymaster and a schoolmaster), he returned to Cambridge in 1919 and graduated as a mathematician in 1921. He became a fellow of St. John’s in 1923, specializing in topology, which was a more or less unexplored field in England at the time.
It was Newman who had introduced Turing to the Entscheidungsproblem, and it was to Newman that Turing gave the first draft of his paper in the spring of 1936. Newman, who had been interested in mathematical machines since working on his own dissertation in 1921, instantly recognized the brilliance of Turing’s ideas and, more important, understood them, and it was Newman who helped get Turing’s paper into the Proceedings of the London Mathematical Society. Newman was well connected in the mathematical world and in the literary world, too—he was married to writer Lyn Irvine, whose first book was published by Leonard and Virginia Woolf’s Hogarth Press in 1931.
Newman and Lyn followed Turing to Princeton in 1937, and at the Institute for Advanced Study Newman worked on a proof for the Poincaré Conjecture (“Every simply connected, closed 3-manifold is homeomorphic to the 3-sphere”).2 When he thought he had it, he gave a five-hour lecture about it to the assembled mathematicians, and no listener found a flaw. Unfortunately, it was Newman himself who found the flaw, shortly after returning to England. The conjecture, one of the most famous in theoretical mathematics, was proposed in 1904 and not proven until a hundred years later (by Grigori Perelman, a reclusive Russian mathematician—his proof was accepted in 2006, and he won a million dollars for it from the Clay Mathematics Institute). But even though Newman’s Poincaré proof failed, he was awarded a fellowship to the Royal Society in 1939.
At the beginning of the war, Max Newman was forty-two, and he and Lyn had two sons. That he felt that he had to send his wife and his half-Jewish children to the United States in 1940 is an index of how uncertain the outcome of the war with Germany seemed at the time. Newman continued at Cambridge and then tried for another fellowship to Princeton, in order to join his family, but finally, in August 1942, he followed many of his friends from Cambridge to Bletchley Park. He was asked to choose between work on Enigma and work on Tunny, and he chose Tunny. Soon after he got there, one of the young mathematicians, William Tutte, came up with an insight into how the Tunny code functioned. Newman began to consider how the repetitive and time-consuming parts of the decoding could be done by machines, and he was put in charge of what came to be known as “the Newmanry.” The first machine they came up with was the Heath Robinson, which the members of the Newmanry improved and tinkered with for many months until it was succeeded by Colossus. Newman, like Turing, came to know Tommy Flowers quite well. It eventually became Newman’s job to oversee the Colossus and coordinate how it worked to break Tunny codes.
As soon as the war was over, Newman accepted a position at the University of Manchester, and in 1946 he got a university grant of £35,000 for computer development. He then went back to Princeton for a year, and there met up with von N
eumann, who, as we’ve seen, was full of his own computer ideas. Newman, privy to the “First Draft” report, very quickly adopted several of von Neumann’s ideas for the Manchester computer. The chief engineers on the project, who came from the Telecommunications Research Establishment (TRE), were F. C. Williams and Thomas Kilburn, whose experience was in radar and electrical circuit design rather than code breaking—they didn’t even know that Colossus had existed.
F. C. Williams was not at first impressed with the computer lab in Manchester: “It was one room in a Victorian building whose architectural features might best be described as ‘late lavatorial.’ The walls were of brown-glazed brick and the door was labelled ‘Magnetism Room.’ ” Williams was ready to go, though—he brought with him an idea he had already been working on at the TRE, using a cathode ray tube as a storage device. What he then invented was called a Williams tube. The stored program was a “pattern of dots” on the face of the tube. Williams tubes were installed in the first Manchester computer, known as the Manchester Baby, as the repository of the computer’s random access memory.
By that September, Turing had given up on the project at the NPL and was back at Cambridge. There, he wrote two papers, played chess, went for walks, and attended a wide variety of lectures. He gave a lecture in January 1948 entitled “The Problems of Robots” to the Moral Sciences Club, an association under the auspices of the philosophy department at Cambridge that for many years had offered a venue for the philosophical jousting of thinkers such as Hegel, Wittgenstein, and Karl Popper. At the end of his two Cambridge terms, he wrote a paper entitled “Intelligent Machinery,” in which he at first likened the human brain to a machine “which can be organized by suitable training” and went on to define and give examples of machines that did various forms of work (a bulldozer, a telephone, ENIAC) and to propose an as yet uninvented machine that could do work and could also develop or, you might say, learn—his model was the cryptanalysis work done by Colossus, though of course he could not mention it.