On John von Neumann and His Role in the Development of the Computer
March 1, 2005
William Aspray
I've decided not to give a technical talk tonight. I'm going to talk about context and personality and individuals, because I felt that, principally trained as a historian of technology and having a math degree that's pretty ancient, it would be like taking coal to Newcastle for me to come to you with a talk about mathematical results.
Von Neumann's career divides pretty well into an early European part and an American part afterward, with a dividing line in the early 1930s. My focus will mainly be on the American period, but let me at least make a few comments about his European career.
Von Neumann's talents as a mathematician were recognized at an early age. Even as he was going through the gymnasium in Hungary, he was taking extra training at the university. At the age of 17, he published his first mathematical paper. In a short span of time he was actively involved in the mathematical community in Europe and around the world---even before he had reached the age of 20.
There were a few mathematicians who were particularly influential for his career in the European days. Early on, while he was still in Hungary, L�szl� Racz was involved. In the German-speaking world, Erhard Schmidt and Hermann Weyl were particularly important. But the most important of all during his European days was David Hilbert. Von Neumann spent time with Hilbert, and it was Hilbert who interested him in two of the major problems that von Neumann worked on and became known for in his early career: his work on the mathematical foundations of quantum mechanics and his work in mathematical logic. He was one of the several young and very talented mathematicians around Hilbert who worked on foundational problems in that period.
Von Neumann was anxious, by the late twenties, to leave Europe. The principal reason was the lack of opportunities for talented mathematicians. He saw a number of other young mathematicians stuck in Germany with a five- or ten-year wait until a professorship would open up, and he was looking for a way to advance more rapidly.
Advancement came largely through the offices of Oswald Veblen, who, in some ways, was every bit as important to von Neumann's career as Hilbert had been. Veblen met von Neumann in 1928, at an international conference in Bologna, and was very impressed with his mathematical capabilities.
Veblen had been actively involved in building up Princeton's mathematics department. Princeton, which had been the College of New Jersey until 1896, was a place where southern gentlemen would go to get finishing training. In the early twentieth century, the transformation of Princeton into a major research university started under the presidency of Woodrow Wilson. The mathematics department was becoming one of the strongest research departments within the university, and Veblen was one of the leaders in that evolution.
There were a number of important hires, including Solomon Lefschetz, the algebraic topologist, and James Alexander; Veblen himself was one of the foremost geometers in the United States. One of the things Veblen had done was to raise money from a man named Jones to use in part for a professorship in mathematical physics. The first appointment to that position was Hermann Weyl. Weyl took the position for one year and then resigned in order to return to G�ttingen.
There was a question about what was going to be done with this chair, and Veblen suggested that it be given to the young mathematician John von Neumann. The faculty wouldn't agree. They decided that this was a position in mathematical physics, and so maybe it should be shared between a mathematician and a physicist. And there were questions about whether it should go to a person as young as von Neumann. So they divided the money---half the year to the mathematician, von Neumann, and half the year to von Neumann's boyhood friend, Eugene Wigner.
For three years, from 1930 to 1933, von Neumann and Wigner split this appointment in mathematical physics. Von Neumann spent the other six months of the year in Berlin, continuing to do research there.
The next thing that happened in von Neumann's career was also partly due to the activities of Veblen. Two families, the Bam-bergers and the Fulds, had made fortunes in department stores in Newark, New Jersey, and wanted to set aside some money for a research institute in medicine. They started out by hiring Abraham Flexner to help figure out where to place it and how to set it up. Flexner soon decided that it wasn't such a good idea to have a medical research center; what should be created instead was a research center in a scientific field. Flexner toured the country to talk to a variety of people about the kind of research center that should be established.
After consulting with a number of people on his tour, he decided to set up the research center in Princeton, and to have it focus primarily on mathematics and mathematical physics. It subsequently became known as the Institute for Advanced Study.
There were a variety of reasons for this choice. One was that mathematicians weren't very expensive to equip-at that time they needed only chalk and a blackboard. Another reason was the wide agreement in the mathematical community about who were the real stars in the field and what were the major problems. There would not be as much controversy in choosing the kind of people they wanted for the institute---the very top research people in the field.
A third reason for setting up the institute in Princeton was that Princeton had built the first purposely designed mathematics building in the United States. A mathematician from the previous generation, Henry Fine, had died after being struck by a car. In his will, he had left money to establish a building for the mathematics department. It was a grand building with stained glass windows. Every office had oriental rugs, leather sofas, and beautiful wood trim. (This building still exists by the way, occupied now by the Asian studies department.)
There was already a strong mathematics department at Princeton. Flexner had been greatly impressed by Veblen, both as a good institution builder and as a good scientist. So it was decided that the new institute would be brought to Princeton.
Who was going to populate this new research institute? They wanted to make four or five hires at the beginning. They chose Flexner to be the director and administrator. For the professors, Veblen was chosen as one, James Alexander as another, and a man by the name of Einstein was chosen, too. They wanted Hermann Weyl to return as a faculty member at the new institute, but he was somewhat reluctant to come back to the United States. Eventually, a year or two later, he did.
But they wanted to make another appointment. They decided that they didn't want to raid Princeton for any of the other senior faculty members because they had already taken two of their strongest senior people. They decided after some time to hire this promising young researcher, von Neumann. Von Neumann continued then in Princeton virtually for the rest of his career, except for some government service at the very end of his life.
While at Princeton, from 1930 to the early 1940s, von Neumann continued to be a very active researcher in pure mathematics. His research included rings of operators, compact Lie groups, ergodic theory, and some early and pioneering work on the theory of games and economic behavior. All of this work stopped very abruptly in 1941 when he became involved in war work.
Von Neumann was well prepared to do the war work, and, again, this was because of Veblen. During World War I, Veblen had been one of three mathematicians who had set up a mathematics research institute at Aberdeen Proving Ground to help with ballistics research. The civilian operation was disbanded at the end of the war, although there was continuing staff work on ballistics.
In the mid-thirties, becoming concerned about the growing political problems in Europe, Veblen decided it was time to reinitiate the activities at Aberdeen. He set up a new group of mathematical advisers and convinced von Neumann to be one of those advisers. In fact, it was Veblen who convinced von Neumann (in 1937) to become a naturalized U.S. citizen so that he could work at Aberdeen Proving Ground. For the next four years, von Neumann served as an adviser at Aberdeen, immersing himself in some of this research.
Unlike many mathematicians, who settled in one particular place for the entire duration of the war to help the Allied effort, von Neumann was an itinerant. He spent much of the war on railroads or in hotels. If you look at his correspondence during this period of time, you'll see that it's written on train stationery or hotel stationery from various places across the country. He worked for both the Army and Navy on ordnance problems. He directed a small research group doing applied mathematics in Princeton, and later on he became one of the active members of the Los Alamos group working on the design of the atomic bomb, on the triggering mechanism.
In 1943, von Neumann was asked to go to England for about six months, to help with some problems on underwater mines. During this time he was stationed mostly in London, but sometimes he had to go out to the Nautical Almanac office in Greenwich.
One day in 1943, he took a train out to Greenwich with John Todd. During the time on the train, von Neumann wrote his first computing program. It was a program for calculation on a National accounting machine-not for any of the kinds of modern machines you would expect today. He was captured by this idea of computation. After the trip to Greenwich with Todd, he wrote back to a colleague in the U.S.: "I've developed an obscene interest in computation, and I'll be returning to the United States a better and impurer man."
Soon after von Neumann returned from his six months in England, Robert Oppenheimer came to see him and asked him to start participating in the Los Alamos project. There were extraordinarily difficult computational challenges in designing the detonation device for the atomic bomb. Von Neumann took a quick study on IBM punch cards from the people at Los Alamos---Nick Metropolis and Richard Feynman were members of that group---and then he started looking around the United States for other places where he might have access to high-speed computing equipment for the Los Alamos design work.
He talked to Warren Weaver, who was a friend, and who was helping to organize scientific personnel to aid the war effort. Weaver told von Neumann about a number of projects that were going on in the United States to develop computing machines. He mentioned Howard Aiken's work at Harvard and some work at Bell Labs. What Weaver did not mention to von Neumann was a project at the Moore School of Electrical Engineering, at the University of Pennsylvania, to build an electronic calculator (which would become known as the ENIAC).
Von Neumann continued for another nine or ten months working on his problems, still stymied by the computational aspects of the work at Los Alamos. Finally one night in 1944, as he was returning by train from Aberdeen where he had attended a scientific advisory committee meeting, he had a fateful meeting.
Standing on the train platform, waiting to return to Princeton, was a young PhD mathematician by the name of Herman Goldstine. He knew who von Neumann was and approached the already well-known von Neumann and introduced himself as a young mathematician who was the Army's liaison to the University of Pennsylvania. Goldstine reports in his book that their conversation was a congenial one, until Goldstine happened to let it slip that he had a machine that could do 300 multiplications per second. All of a sudden, the conversation turned into what seemed to Goldstine to be more like a doctoral dissertation defense.
Within two weeks, von Neumann had gained clearance to visit the ENIAC project in Philadelphia. By the time of his first visit, the people there had already done most of the design work on the ENIAC, and they were starting to build a new machine, the EDVAC, to rectify some of the limitations of the ENIAC, including its programming. They would program the ENIAC by plugging cables in, and then turning a set of switches---it could take one to two days to program a problem, even after they knew what the algorithms were. The designers wanted to have some kind of capability for quicker programming.
Von Neumann was still traveling during this time between Princeton, Los Alamos, and Aberdeen, and he couldn't spend very much time in Philadelphia. He got there about once a month during the last few months of the war, but he had several intense conversations with J. Presper Eckert and John Mauchly, the designers of the ENIAC.
While traveling, von Neumann wrote most of a report known as Draft Report on EDVAC, which described the stored-program concept for the first time. The description was functional rather than written in an engineering way. To avoid the engineering details, von Neumann used abstract neural net terminology taken from a paper by Warner McCulloch and Walter Pitts that had been published two years earlier. Eckert and Mauchly, who were interested in the EDVAC primarily as an engineering artifact, undervalued von Neumann's contribution.
As he traveled, von Neumann was in regular correspondence with Goldstine. He mailed a copy of the Draft Report to Goldstine, who decided to distribute it widely. It had only von Neumann's name on it, and von Neumann got the credit, which caused terrible rifts later on, with Eckert and Mauchly, over the credit for the stored-program concept.
Even before the war was over, von Neumann decided that he wanted to build a high-speed computing machine for himself at Princeton. The decision meant a challenge to carry it through since the Institute for Advanced Study was a difficult place to build a computing machine.
Think about the way the Institute was organized. It was planned as a place for only a few really deep thinkers---not as a place for technicians, laboratory staff, or laboratory facilities. But a computer requires programmers, technicians to maintain the machine, and many engineers to build it in the first place. The Institute didn't have laboratory equipment, and many of the parts had to be designed. As you can imagine, there was a lot of skepticism on the part of von Neumann's faculty colleagues at the Institute. Veblen was very supportive, and physicist John Wheeler and astronomer Martin Schwarzschild at Princeton University were very supportive. But von Neumann didn't have strong support from others at the Institute.
So what did von Neumann do? He used a carrot-stick approach to get approval for the computer project. He made a pitch to the trustees of the Institute, something like this: "Here are ten problems in science that we could never get at before, because we didn't have computational tools. Imagine how famous you're going to be when we solve all these problems with this new computational device." That was the carrot side of it. The stick side was to secure promises of job offers from the chairs of the math departments at Columbia, MIT, and Chicago. He told the Institute trustees that he would leave if he didn't get to build a computer. The trustees gave their approval.
He convinced RCA to build the memory device for him. Memory devices were the hardest unit to build for computers at that time. RCA had recently moved its research laboratories to the edge of Princeton, and it was an opportunity for them to get involved. One of their principal engineers, Vladimir Zworykin, was quite interested.
Von Neumann convinced the Army and Navy to provide the money to run this project. He didn't want this to be a production computer that just did a lot of military jobs, though. He wanted it instead to be a machine that was designed to do research and would be used only for that purpose. So he cut a deal: He would write meticulous progress reports that the Army and Navy could distribute to several other research centers, which would build production computers for the Army and Navy. This way the Institute computer could be reserved for scientific research. Later on, the project received funding from the Atomic Energy Commission and the Air Force for operation of the computer.
In addition to the work on the computer project, von Neumann was actively involved as a consultant and spokesperson in the military-industrial complex. His consulting work included continued consultation for Aberdeen Proving Ground. He also worked for the Atomic Energy Commission-first on the scientific advisory committee and, later, as one of the five commissioners who governed the organization. He worked for Los Alamos and for the RAND Corporation, as well as for the National Security Agency and its predecessor organization, which was headed by his friend William Friedman. He advised the Air Force and was a member of the original committee that worked on the intercontinental ballistic missile. There were also several industrial consulting relations, including one with IBM on the design of equipment.
Von Neumann left Princeton for Washington in 1954 to serve on the AEC. He didn't plan on returning to Princeton. He had signed a contract to join the faculty at UCLA as a scientist-at-large, where he could focus on scientific computing. But he became quite ill while he was in Washington, and died of cancer in 1957.
Before closing, let me mention the five major areas of von Neumann's research during this period:
1. He was actively involved in the design and construction of computers. A series of draft reports done in the 1940s were used by many different organizations that were building computers.
2. He did work in the area of scientific computing. His major test of the computer as a scientific instrument was to use it for numerical weather forecasting. Peter Lax will say something about these related issues in his talk, so I won't say much other than that he had a two-dimensional model to run on the ENIAC in 1950. A three-dimensional model that was running in 1952 on the Institute for Advanced Study computer was successfully used for both military and commercial forecasting. It became the basic model used by the National Weather Service and so was seen as a great success.
3. His third research area was numerical analysis: numerical stability for differential equations, error analysis for linear equations, Monte Carlo methods, generation of pseudo-random numbers, linear programming, and artificial viscosity in shock calculations.
4. He was interested in what we would now call theoretical computer science and in particular in automata theory. He tried to make sense of how Turing's work and McCullough's and Pitt's work fit together with Wiener's work on cybernetics. He wrote a series of papers on probabilistic automata with reliability, on complexity issues that came up in automata theory, and the first papers on cellular automata.
5. Finally, let me end by saying that he was a real legitimizer and promoter of the computer and scientific computing in government and military circles. He was one of the very few senior, distinguished, internationally recognized scientists in this new field of computation, and he was listened to. He was able to get research funds in a way that no one else could at this time. He made this a new field of study.
Discussion
JOE GRCAR: You mentioned Einstein. What was von Neumann's relationship with Einstein at Princeton? Could you say briefly whether they were friends, whether they worked together?
BILL ASPRAY: I don't think they worked together. There's very little evidence in the archival record about what their relationship might have been. There are occasional discussions, say, in proceedings of scientific meetings about their interchanges on concepts, but I don't see any substantive relationship.
CLEVE MOLER: We've heard stories about von Neumann's mental computing capabilities. Were they true?
BILL ASPRAY: I've heard lots of stories-I believe they're probably true to some degree. He was tremendously skilled at doing computations in his head and doing them rapidly. I remember a story George Dantzig told me. Dantzig had gone to Princeton to see von Neumann at one point to talk to him about the idea of linear programming. Dantzig went to the board and started writing. Von Neumann would interrupt and say, "No, here is the solution as I understand it. Go on." Dantzig said that he thought he would have a three-hour lecture to give, and in fifteen minutes he'd exhausted all his material. That's the type of story that I've heard from a number of different people about the lightning speed of von Neumann's mind.
