Turing's Cathedral

by George Dyson

  Bigelow, alone, remained at the Institute after all the other engineers dispersed. Although his contributions, in von Neumann’s assessment, had been “very important, considerably more than one would infer from a superficial inspection of the publications,” his lack of academic publications counted against him, even though there were no explicit publication requirements at the IAS.20 The “Interim Progress Report on the Physical Realization of an Electronic Computing Instrument” may have been the most influential document ever published by the IAS, but it didn’t count.21 Without von Neumann, Bigelow no longer fit in, and the School of Mathematics expected him to make a graceful exit to IBM or to return to an institution such as MIT.

  “I think that most of us that were down there—now maybe Bigelow and von Neumann, I won’t speak for them, but certainly for myself, and I suspect Pomerene and most of the other engineering group—we were just doing a job, and it was an interesting kind of a job,” says Willis Ware, remembering the beginnings in the basement of Fuld Hall. “We didn’t have the big foresight and the big omniscience to see all the consequences. Well, they had a little vector started that turned out subsequently to be very important vectors.”22

  How did the von Neumann vector manage to outdistance all the other groups trying to build a practical implementation of Turing’s Universal Machine in 1946? The Eckert-Mauchly group and the von Neumann group were both competing for funding and engineers. “Eckert and Mauchly have a contract with the Govt. Bureau of Standards, at first for 1 year and 50 kilobucks,” von Neumann reported to Klári in November 1946. “They have started, and shanghaied back 2 of our men, whom we had previously shanghaied from them.”23 By 1949 the Eckert-Mauchly UNIVAC computer was ready to go into production, and adoption of their machine by the U.S. government would have put them firmly in the lead.

  “It was decided, after careful consideration … that the Bureau should proceed to contract on three UNIVACs from Eckert and Mauchly, one for the Bureau of the Census and two for the Military Establishment,” reported an undated memorandum from the National Bureau of Standards electronic computing program, evidently written in 1949. “For just about two days the horizon seemed clear,” the report continues.

  No additional obstacles confronted the Bureau. However, this happy state was dispelled quite unexpectedly when the Bureau was informed by Dr. Mina Rees and Colonel Oscar Maier, representing the Office of Naval Research, and the Air Materiel Command, respectively, that the Eckert-Mauchly Computer Corporation had been submitted to a security investigation on which it had not received a “clean bill of health,” and that the Bureau therefore should not use ONR and AMC funds for the procurement of UNIVACs from that company. The Bureau was able to continue negotiations only on the basis of one computer rather than three; the Bureau was constrained not to inform the company about the security investigation.24

  The scales were tipped away from the UNIVAC and toward IMB’s “Defense Calculator,” later known as the IBM 701—the first copy of which was delivered to Los Alamos in 1953. Eckert and Mauchly fell increasingly into debt, until they were forced, in 1950, to sell their company (and patent portfolio) to Remington Rand—whose vice president was General Leslie Groves. “These machines should find a reasonable market,” Goldstine and von Neumann wrote to Groves in 1949, in a nine-page letter that detailed how the ENIAC had been retrofitted to become a stored-program computer, and how Remington Rand could modify their existing punched card equipment to form “an all-purpose machine [whose] memory [could] be used to contain not only numerical data but also logical instructions.”25 After acquiring Eckert and Mauchly’s Electronic Control Company, Remington Rand filed patent-infringement suits against many of their competitors—except IBM, with whom they established a cross-licensing agreement in 1956.

  IBM soon became the dominant force in digital computing and, beginning with von Neumann’s one-day-per-month consulting contract, hired much of the talent that had accumulated at the IAS. James Pomerene joined IBM in 1956, where he led IBM’s early efforts to develop cached, high-speed memory architectures and parallel, multiple-core processors, and was appointed an IBM Fellow in 1976—with complete freedom to pursue any avenue of research, equivalent to a permanent membership at IAS. Herman Goldstine left the IAS in 1958 to supervise IBM’s mathematical research center, housed temporarily at the Lamb Estate in the Hudson Valley while awaiting the completion of the Thomas J. Watson Research Center at Yorktown Heights, where he continued the tradition of scientific computing he had established at IAS, becoming an IBM Fellow in 1969. “At the Lamb Estate we thought of ourselves as princes of the earth because of our computing support,” remembers Ralph Gomory, who joined Goldstine’s group in 1959. “Every day a station wagon left the Lamb Estate and went up to Poughkeepsie. It carried our programs and returned the next day with results.”26

  Jack Rosenberg left the IAS for a position at General Electric in Syracuse in 1951, moved to Los Angeles in 1954, and, after von Neumann’s death ended his plans to work for the proposed new computing laboratory at UCLA, joined the Los Angeles Scientific Center of IBM. “A long time IBM engineer showed me some circuit diagrams of IBM’s first electronic computer, the 701,” he remembers. “It was a copy of the Von Neumann computer, which I had developed in 1947–51.” Rosenberg was offered an IBM Fellowship in 1969, which he declined, explaining that “the company was too large and corrupt.”27 He still lives in Pacific Palisades, listening to music through a set of the same phase-synchronized “coherent sound” loudspeakers that he had installed in Einstein’s house in 1949. Einstein, in appreciation, granted Rosenberg a wide-ranging and candid interview, which Rosenberg recorded on high-fidelity equipment, but will not release. “Einstein said it must never be made public,” he explains.28

  Gerald and Thelma Estrin moved to Israel in 1954 to supervise the construction of the WEIZAC at the Weizmann Institute of Science in Rehovot, returned to the IAS in 1955, and moved to UCLA in 1956, where they helped establish the new Department of Computer Science, nurturing a new generation of entrepreneurial computer scientists, including two of their own daughters, Deborah and Judy Estrin, and Paul Baran of RAND. “It was a marvelous accident, wonderful for everything that happened afterwards, that things didn’t get classified,” say the Estrins, looking back at the Electronic Computer Project at IAS.29

  Andrew and Kathleen Booth returned to England, where they remained instrumental in the continued development of digital computing and X-ray crystallography, before moving to Canada in 1962. “Kathleen and I were amused at the concern for our moral wellbeing!” Andrew responded, when shown a copy of the February 1947 discussion between Goldstine and von Neumann concerning their housing arrangements at the IAS.30

  Joseph and Margaret Smagorinsky helped found the Princeton Geophysical Fluid Dynamics Laboratory, where climate modeling continued from where the IAS meteorology project had left off. Jule Charney and Norman Phillips settled at MIT, forming the nucleus of a computational meteorology group that resolved some of the differences between John von Neumann’s reasons for believing that weather could be made predictable and Norbert Wiener’s reasons for believing that it could not. Hedi Selberg transferred her expertise to the Princeton Plasma Physics Laboratory, and Ralph Slutz became director of computing at the National Center for Atmospheric Research, in Boulder, Colorado. Richard Melville and Hewitt Crane went to the Stanford Research Institute, developing, among other things, the ERMA system for electronic clearing of machine-readable checks between banks. Dick Snyder returned to RCA, working on magnetic-core memory but unable to persuade RCA, as Zworykin had managed to with television, to take the lead. Morris Rubinoff returned to the University of Pennsylvania, and for an interval to Philco, where he supervised the design of the Philco 2000, the first fully transistorized computer, with asynchronous arithmetic, a feature that had been developed at IAS. Arthur Burks settled at the University of Michigan, where he founded the Logic of Computers Group in 1949, edited von Neumann’s Theory of Self-Repr
oducing Automata (1966), and, with Alice Burks, published the definitive Who Invented the Computer? in 2003.

  Robert Oppenheimer was stripped of his security clearance in 1954, one day before it would have expired on its own, in a deliberate act of public humiliation that brought postwar dreams of civilian control of nuclear weapons to an end. “The military wanted the whole deal: the laboratories, the computers, the whole future, in nuclear weapons, from A to Z,” explains Harris Mayer. “When we set up the AEC, the military was shut out of what was their major firepower, and they never forgot that, and they wanted it back. They got back the command of the nuclear weapons and the computers, and a part of this, a minor part, actually, was to discredit Oppenheimer.”31 His AEC safe and the guards who watched over it in Fuld Hall were removed. The Institute faculty put aside their differences to support Oppenheimer against those who wanted him ousted from the IAS, where he remained as director until 1966, when, stricken with throat cancer, he resigned and moved out of Olden Manor to become our neighbor for his final year. The former ruler of Los Alamos mesa and Olden Farm was now a ghostly pale, thin figure pacing the yard on the other side of our hedge.

  Lewis Strauss, Oppenheimer’s nemesis, remained a trustee of the Institute until 1968, when, stung by lingering disapproval over his role the Oppenheimer affair, he finally quit. He remained a friend of the FBI, and a memo from the FBI’s special agent in charge, New York Office, to the director records that, after returning from Geneva on August 21, 1955, “Admiral and Mrs Lewis STRAUSS and Admiral STRAUSS’ aide … were met by a liaison agent of this office who facilitated their entry through Customs and extended the usual courtesies. Admiral STRAUSS made a number of favorable comments concerning the Director and the Bureau and one comment was ‘Mr. HOOVER is always there when you need him.’ ”32

  Abraham Flexner, who died in 1959, had little more to do with the Institute after his departure in 1939. According to his daughter, Jean Lewinson, “when he was through with it it was a complete separation.” In 1955, when her father was eighty-eight, she reported that “this summer in Ontario Dr. Flexner swam in water so cold that Mr. Lewinson won’t go near it. He doesn’t saw wood now but he does fish and take walks.”33

  Vladimir Zworykin died in 1982, working on biomedical applications of electronics and discouraged that television, the invention he had the highest hopes for, had been so misused. After their abortive venture with the Selectron, RCA never again took a decisive lead in digital computing, devoting their resources to commercial television and to their broadcasting spinoff, NBC.

  Lewis Fry Richardson lived until 1953, just long enough to see his dreams of numerical weather prediction and fears of unlimited weaponry fulfilled. Although he hailed the ENIAC forecasts of 1950, reported to him by Jule Charney, as “an enormous scientific advance,” he had long retired from meteorological work, having switched instead to applying himself to a study, initially published in 1944, of “The Distribution of Wars in Time.” The evidence was discouraging. “The agreement with Poisson’s law of improbable events draws our attention to the existence of a persistent background of probability,” he concluded. “If the beginnings of wars had been the only facts involved, we might have called it a background of pugnacity. But, as the ends of wars have the same distribution, the background appears to be composed of a restless desire for change.”34

  Norbert Wiener died of cardiac arrest on a visit to Stockholm in 1964. Disillusioned over military ambitions in general and the use of nuclear weapons against civilians in particular, the founder of Cybernetics had begun speaking out against military-sponsored research. “Machines can and do transcend some of the limitations of their designers,” he warned in the pages of Time magazine. “This means that although they are theoretically subject to human criticism, such criticism may be ineffective.” The author of Extrapolation, Interpolation, and Smoothing of Stationary Time Series saw that ever-faster machines would inevitably leave human beings behind. “By the very slowness of our human activities, our effective control of our machines may be nullified,” he added, citing computer-controlled nuclear weapons and computer-controlled manipulation of the stock market as two of the ways that power was being relinquished to the machines.35

  Stan Ulam lived until 1984, dividing his time between Los Alamos, Boulder, and later Santa Fe. He remained as imaginative and mathematically creative as he had been as a child, circulating among his colleagues still working at Los Alamos and keeping alive the conversations that had begun in Lwów at the Scottish Café. “Just as animals play when they are young in preparation for situations arising later in their lives it may be that mathematics to a large extent is a collection of games,” he concluded in 1981, “and may be the only way to change the individual or collective human mind to prepare it for a future that nobody can now imagine.”36 Hordes of self-replicating Turing-complete digital organisms, much as he imagined them in 1952, now populate an unbounded matrix, while the companies and individuals who nurture them are ever more richly rewarded in return.

  Edward Teller outlived von Neumann by forty-six years. He remained unrepentant over the development of the hydrogen bomb, but regretted having allowed his testimony at the Oppenheimer hearings to be used to identify Oppenheimer as a security risk, and questioned whether secrecy was a path to security. “Science thrives on openness,” he reflected in 1981, “but during World War II we were obliged to put secrecy practices into effect. After the war, the question of secrecy was reconsidered … but the practice of classification continued; it was our ‘security,’ whether it worked or failed.… The limitations we impose on ourselves by restricting information are far greater than any advantage others could gain.”37

  Teller had grown up competing with the older and faster Johnny, and when he finally caught up to him, it was a tragic moment for both. “The last few weeks, the last few months of his life, I saw him quite frequently, although I had to cross the continent to come to see him,” Teller remembered. “We used to discuss everything in the world. He was incredibly fast. Beyond him, we have never seen. And then in the hospital, he wanted to continue. But he was no longer ahead of me. For Johnny von Neumann, thinking and mathematics was a vital necessity. And he wanted to see me, again and again, because he wanted to prove to himself, ‘I still can do it.’ But he couldn’t.”38

  Near the end, von Neumann, who could no longer work entirely without notes, asked one of his visitors, identified only as “JmcD,” for “a note regarding what we talked about last Wednesday,” which was recorded as follows:

  We talked somewhat randomly but this was the pattern of it: You said you were in a state of introversion and struggling with a problem of claustrophobia in space and time: in space because your physical body gets in the way; in time because of the slowness of elementary reactions.… These problems you said might be overcome with a mechanical device … that would project a book page on a photosensitized surface on the ceiling, a phosphorescent pencil for writing on it, and a device with options: to move pages forward and backwards one page or several pages, the luminous pointer to be in several colors with a method of erasure. You said such an invention was difficult but not impossible.… The idea is to be able to read and write “pure bred in consciousness without physical interference.”39

  After Johnny’s death, Klári remained in Washington, D.C., sorting out his affairs and arranging for the publication of his collected works. Even after a decision to include only previously published papers, the Collected Works still amounted to some 3,689 pages, spread over six volumes and finally published in 1963. Oppenheimer, still struggling with what role the Institute should play with regard to Einstein’s papers, was at a loss over what to do when von Neumann’s literary godfather appeared. This was “Captain” I. Robert Maxwell, the Czechoslovakian-born publishing magnate and future member of Parliament who offered to undertake the publication of the Collected Works, assuring Klári (and Oppenheimer) that “my role in this project is that of the ‘mechanic’ who has the
facilities and the ‘know-how’ for such a task, and may I say that I am glad to be able to assist with this noble cause.”40

  Maxwell had visited Los Alamos while launching his Pergamon Press, and in addition to befriending the von Neumanns, he became particularly close to the Ulams. “They sent us their children on vacations,” says Françoise, whose daughter, Claire, spent her year abroad in Oxford under the Maxwells’ wing. “I used to joke that he would either become prime minister or end up in jail,” she adds. “He came close to both.”

  In October 1957, Oppenheimer telephoned Maxwell to try to solidify terms for publication of the Collected Works. “If it costs too much it will very much limit usefulness. Do you have any notion?” he asked.

  “My idea is about £10,” Maxwell replied.

  “For the whole thing?”

  “Yes.”

  “Miraculous!” exclaimed Oppenheimer.

  “This set is to be my contribution to this man,” answered Maxwell, who suggested Eugene Wigner as editor. Oppenheimer suggested mathematician Shizuo Kakutani. Klári suggested geophysicist Carl Eckart, who lived in La Jolla and had been one of the reasons that von Neumann’s agreement with the University of California included a clause allowing him to spend as much time at Scripps Institution of Oceanography as he wished.41

  Eckart declined the assignment, which went to Abraham Taub, but met with Klári to discuss the project. They were married in 1958. It was Klári’s fourth marriage.42 The first had been for romance, the second for money, the third for brains, and the fourth for California. Klári now settled down in La Jolla, just above Windansea Beach, whose antiestablishment surfing culture was soon to be immortalized in The Pump House Gang, by Tom Wolfe. Although Carl Eckart and John von Neumann, according to Klári, were “both in similar fields,” they were, “as human beings, farther apart” than Johnny had been from the “non-intellectual banker” she had married long ago in the aftermath of her first, gambling husband in Budapest. “For the first time in my life I have relaxed and stopped chasing rainbows,” she wrote on the last page of an unfinished memoir of her life.