by George Dyson
One thousand pulses, about a microsecond apart, could be stored in the millisecond it took an acoustic signal to travel the length of a 5-foot “tank.” By regenerating the pulse train, and listening to the data stream as it passed by, it was possible to read and write data with millisecond access times. “After the central control organ listens to all 32 words of one column it passes to the next column,” von Neumann explained to Warren Weaver in 1945, referring to 30-bit code segments for the first time as “words.” Acoustic delay-line memory was used in many first-generation stored-program computers, although, as British topologist Max Newman complained, “its programming was like catching mice just as they were entering a hole in the wall.”36
At the time von Neumann began collaborating with the ENIAC group, planning for ENIAC’s delay-line successor was already under way. The Electronic Discrete Variable Automatic Computer, or EDVAC, “would be quite flexible in its control facilities, would have about 50 times as large a memory, i.e., be able to store about 1000 ten decimal digit numbers, and contain only about 1⁄10 as many tubes,” Goldstine and von Neumann reported.37 The machine would be programmed by loading coded sequences into high-speed memory rather than by setting cables and switches by hand.
“The idea of the stored program, as we know it now, and which is a clear-cut way of achieving a universal computer, wasn’t invented overnight,” explains Rajchman. “Rather it evolved gradually. First came manually changeable plug-ins, relays, and finally the modifying contacts themselves became electronic switches. Next came the idea of storing the state of those switches in an electronic memory. Finally this resulted in the idea of the modern stored program in which ‘instructions’ and ‘data’ are stored in a common memory.”38
Even before the advent of the ENIAC, the elements of stored-program computing were falling into place. In July of 1944, on behalf of the implosion effort at Los Alamos, von Neumann and Stan Frankel were briefed on the series of Bell Telephone relay computers being built by Samuel B. Williams and George R. Stibitz in New York. The new machines were controlled by punched paper tape, and von Neumann reported to Oppenheimer on August 1 that the “problem tape carries numerical data, and operational instructions.” As he described it to Oppenheimer, “an instruction on the control-tape therefore looks like this: ‘Take the contents of register a, also the contents of register b, add (or subtract, or multiply, etc.), and put the result into register c.’ ” Not only were data and instructions intermingled, but also the computer could, in principle, modify its own instructions. “The machine can use in (a) a tape which comes from its own reperforator—i.e., which it has punched itself.”39
Eckert and Mauchly were thinking along similar lines. “All through 1944, and in 1945 as well, we were leading a ‘double life,’ ” Mauchly recalls. “For much of two shifts, 8 AM to Midnight, both ENIAC construction and testing needed supervision. Then as hourly workers went home and project engineers ‘thinned out,’ Eckert and I were left time to consider that ‘next machine.’ Naturally, ‘architecture’ or ‘logical organization’ was the first thing to attend to. Eckert and I spent a great deal of thought on that, combining a serial delay line storage with the idea of a single storage for data and program.”40
In the closing months of World War II, von Neumann circulated between Princeton, Los Alamos, Washington, Philadelphia, and Aberdeen, conveying a stream of new ideas. “None of us was important enough to have persuaded people to accept this kind of thing,” says Goldstine. “In the first place, von Neumann had a real built-in need at Los Alamos.… They had an enormous IBM punched card installation out there doing implosion calculations. I just don’t believe any of us could have gone and persuaded somebody like Fermi of the importance of numerical calculation the way von Neumann could.”41
In early 1945, during the final push to finish and test the atomic bomb, von Neumann’s notes on the EDVAC project were typed up under Goldstine’s supervision and distilled into a 105-page report. The “First Draft of a Report on the EDVAC,” reproduced by mimeograph and released into limited distribution by the Moore School on June 30, 1945, outlined the design of a high-speed stored-program electronic digital computer, including the requisite formulation and interpretation of coded instructions—“which must be given to the device in absolutely exhaustive detail.”42
The functional elements of the computer were separated into a hierarchical memory, a control organ, a central arithmetic unit, and input/output channels, making distinctions still known as the “von Neumann architecture” today. A fast internal memory, coupled to a larger secondary memory, and linked in turn to an unlimited supply of punched cards or paper tape, gave the unbounded storage that Turing had prescribed. The impediment of a single channel between memory and processor is memorialized as the “von Neumann bottleneck,” although its namesake attempted, unsuccessfully, to nip this in the bud. “The whole system will be well balanced, so, that if it is properly and intelligently used, there will be no bottlenecks,” he explained to Max Newman, “even not at the outputs and inputs of the human intellects with which it has to be matched.”43
When a subject captured von Neumann’s attention, he reconstituted it on his own terms from the bottom up. Digital computing required no such process of reduction; it was all axioms from the start. In 1945 the ENIAC and EDVAC were still classified military projects. Von Neumann could speak freely in logical abstractions, but not in specific electronic circuits. So he did. He was also, as Julian Bigelow put it, “clever enough to know that his forte was not in experimental work or in making things function in the real world.”44
During the war, both open publication and individual credit had been suspended—for both computers and bombs. After the war, it was decided that bombs would be kept secret and computers would be made public, with a scramble for credit as a result. The EDVAC report engendered widespread controversy, despite the small number of copies that were distributed before the mimeograph stencils gave out. Von Neumann was listed as the sole author, without any acknowledgment of the contributions made by other members of the EDVAC group. Eckert and Mauchly, who had been pledged to silence about the ENIAC and EDVAC, felt slighted by a publication that was based on their own unpublished work. “Johnny was rephrasing our logic, but it was still the SAME logic,” says Mauchly.45 Adding injury to insult, the EDVAC report would be deemed to constitute a legal publication invalidating any patents not filed within a year.
“It wasn’t even a draft when he wrote it,” explains Eckert. “He wrote these letters to Goldstine, and when we asked what he was doing this for at the time, Goldstine said, ‘He’s just trying to get these things clear in his own mind and he’s done it by writing me letters so that we can write back if he hasn’t understood it properly.’ ” The report, compiled by Goldstine and accompanied by rough hand-stenciled sketches, contained blank spaces where references were to be inserted. The word EDVAC never appears in the text of the report. “He grasped what we were doing quite quickly,” adds Eckert. “I didn’t know he was going to go out and more or less claim it as his own.”46
“I certainly intend to do my part to keep as much of this field ‘in the public domain’ (from the patent point of view) as I can,” von Neumann explained to Stan Frankel, arguing for an open systems approach at IAS.47 “The primary purpose of this report was to contribute to clarifying and coordinating the thinking of the group working on the EDVAC,” he testified in 1947, when questions as to the disposition of patent rights first arose. The secondary purpose was to publish the preliminary results as soon as possible, “in order to further the development of the art of building high speed computers,” he explained, concluding that “my personal opinion was at all times, and is now, that this was perfectly proper and in the best interests of the United States.”48
With the war over, individual interests eclipsed the interests of the United States. The Moore School was too academic for Eckert and Mauchly, and not academic enough for von Neumann. Eckert and Mauchly left to f
orm the Electronic Control Company and build commercial computers—first BINAC and then UNIVAC, a brand synonymous with computing for a time. Von Neumann decided to go build his own computer, as a scientific instrument, somewhere else. Spare time on the ENIAC and even the EDVAC would not be enough. “It was, therefore, the most natural thing that von Neumann felt that he would like to have at his own disposal such a machine,” says Willis Ware. “If he really wanted a computer, the thing to do was to build it,” adds Arthur Burks.49
Von Neumann’s initial thinking was to transplant the entire core of the ENIAC group. “At the end of the war, he had a whole new set of ideas that were not incorporated in the ENIAC,” Ware explains. “I can imagine Johnny thinking to himself, ‘Well, here’s myself and Herman and Eckert and Mauchly, and Burks. What a team to go do this thing that I want to do!’ ”50
Eckert declined von Neumann’s invitation to lead the IAS engineering team, going into business with Mauchly for himself, while von Neumann entered into a series of lucrative personal consulting contracts with IBM. “Von Neumann agrees to assign to IBM, with the exception of the inventions specified below, the entire rights to any and all improvements and inventions made by him,” reads the draft of a retainer agreement with IBM, dated May 1, 1945. As Eckert later complained, “he sold all our ideas through the back door to IBM.”51
Von Neumann’s customary good nature appeared to be breaking down. “Eckert and Mauchly are a commercial group with a commercial patent policy,” he explained to Stanley Frankel, who remained on good terms with Eckert and Mauchly ever since the weeks of all-night troubleshooting in late 1945 and early 1946. “We cannot work with them directly or indirectly, in the same open manner in which we would work with an academic group,” he warned. “If you wish to maintain the same type of close contact with the Eckert-Mauchly group—which is for you and you alone to decide—then you should not put yourself into an incompatible position by communicating with us too.”52
At the Moore School, the EDVAC was orphaned, superseded by the Eckert-Mauchly enterprise on the one hand and the Institute for Advanced Study project on the other. By the time the EDVAC was finished, in 1951, its mercury delay-line memory and serial architecture had been overtaken by advances sparked by its own draft report. To achieve a “practicable” and fully random-access memory, it was explained near the end of the EDVAC report, the Farnsworth-Zworykin iconoscope, rather than the Eckert delay line, might be the “more natural” approach. “This device in its developed form remembers the state of 400 × 500 = 200,000 separate points,” it was noted. “These memories are placed on it by a light beam, and subsequently sensed by an electron beam, but it is easy to see that small changes would make it possible to do the placing of the memories by an electron beam also.” The iconoscope “acts in this case like 200,000 independent memory units,” with the switching between individual capacitors being performed “by a single electron beam—the switching action proper being the steering (deflecting) of this beam so as to hit the desired point on the plate.”53
Sixty years later, most primary computer memory is embodied, in silicon, as dynamically refreshed arrays of capacitors—the current implementation of Farnsworth’s, Zworykin’s, and Rajchman’s original translation between coded sequences in time and arrays of charge in space. Ten million capacitors now cost less than one cent. Memory locations are addressed directly by digital switching rather than indirectly by the deflection of an electron beam, but the underlying principle and logical architecture remain unchanged. Our ever-expanding digital universe is directly descended from the image tube that imploded in the back seat of Zworykin’s car.
Where to build the new computer? The Institute did not even have a workbench where you could plug in a soldering gun. “One could hardly imagine a more improbable environment,” Julian Bigelow adds. “How does all of this fit in with the Princetitute?” asked Norbert Wiener, in March of 1945. “You are going to run into a situation where you will need a lab at your fingertips, and labs don’t grow in ivory towers.”54 Wiener helped arrange for an invitation, at department-head level, to MIT, with the assurance that all the resources of MIT would be at von Neumann’s disposal to build the envisioned computing machine.
Competing offers were made by Harvard, the University of Chicago, and IBM. “We are all very much interested in your man, von Neumann,” Harvard president James Conant wrote to Frank Aydelotte. “The question is, could we get him.” Von Neumann played the invitations against one another—and against those who resisted the construction of a computer at the Institute—until he got his way. “The question of how to hold on to John von Neumann is growing more urgent from day to day.… It would be a tragedy if we lost him,” James Alexander warned Frank Aydelotte. “I doubt whether he would be willing to stay with us if this meant giving up entirely the work on high speed mathematical machines.”55
Loaning von Neumann to Los Alamos during the wartime emergency was one thing, but losing him to a rival institution would be a serious blow to the IAS. Von Neumann was quick to take advantage of this. “These negotiations,” in Bigelow’s opinion, “were a part of a bargaining way of life which von Neumann, so to speak, managed with his left little finger, while the other fingers on his hands were doing more effective and important work.” Aydelotte countered the rival offers, reassuring Alexander that “you may feel free to say to von Neumann that I have every confidence in finding the funds from one source or another to enable him to carry out his plans.” There was little time. Not only was von Neumann impatient to get started on the computer, but Aydelotte was about to leave for Palestine, as a member of the Joint Anglo-American Commission that concluded, unanimously, in April 1946, that “Palestine must ultimately become a state which guards the rights and interests of Moslems, Jews and Christians alike.”56
At the Institute for Advanced Study in early 1946, even applied mathematics was out of bounds. Mathematicians who had worked on applications during the war were expected to leave them behind. Von Neumann, however, was hooked. “When the war was over, and scientists were migrating back to their respective Universities or research institutions, Johnny returned to the Institute in Princeton,” Klári recalls. “There he clearly stunned, or even horrified, some of his mathematical colleagues of the most erudite abstraction, by openly professing his great interest in other mathematical tools than the blackboard and chalk or pencil and paper. His proposal to build an electronic computing machine under the sacred dome of the Institute, was not received with applause to say the least.” It wasn’t just the pure mathematicians who were disturbed by the prospect of the computer. The humanists had been holding their ground against the mathematicians as best they could, and von Neumann’s project, set to triple the budget of the School of Mathematics, was suspect on that count alone. “Mathematicians in our wing? Over my dead body! and yours?” Aydelotte was cabled by paleographer Elias Lowe.57
Aydelotte, however, was ready to do anything to retain von Neumann, and supported the Institute’s taking an active role in experimental research. The scientists who had been sequestered at Los Alamos during the war, with an unlimited research budget and no teaching obligations, were now returning in large numbers to their positions on the East Coast. A consortium of thirteen institutions petitioned General Leslie Groves, former commander of the Manhattan Project, to establish a new nuclear research laboratory that would be the Los Alamos of the East. Aydelotte supported the proposal and even suggested building the new laboratory in the Institute Woods. “We would have an ideal location for it and I could hardly think of any place in the east that would be more convenient,” Aydelotte, en route to Palestine, cabled von Neumann from on board the Queen Elizabeth. At a meeting of the School of Mathematics called to discuss the proposal, the strongest dissenting voice was Albert Einstein, who, the minutes record, “emphasizes the dangers of secret war work” and “fears the emphasis on such projects will further ideas of ‘preventive’ wars.”58 Aydelotte and von Neumann hoped the computer proj
ect would get the Institute’s foot in the door for lucrative government contract work—just what Einstein feared.
Aydelotte pressed for a proposed budget, and von Neumann answered “about $100,000 per year for three years for the construction of an all-purpose, automatic, electronic computing machine.” He argued that “it is most important that a purely scientific organization should undertake such a project,” since the government laboratories were only building devices for “definite, often very specialized purposes,” and “any industrial company, on the other hand, which undertakes such a venture would be influenced by its own past procedures and routines, and it would therefore not be able to make as fresh a start.”59
Aydelotte first sought funding from philanthropist Samuel Fels, emphasizing the “contributions to mathematics, physics, biology, economics and statistics which might be made by an electronic computer,” and promising that the new device would open up new areas of knowledge “in the same remarkable way that the two hundred inch telescope promises to bring under observation universes which are at the present moment entirely outside the range of any instrument now existing.”60 Despite being offered a private audience with Einstein, Fels declined to lend support.
Aydelotte then approached the Rockefeller Foundation’s Warren Weaver, who was familiar with the other laboratories working on computers, and in a unique position to evaluate the proposal from IAS. “I am somewhat surprised that von Neumann is himself interested in the actual construction and operation of a great new computing engine,” Weaver answered on October 1, 1945. “The device we are all dreaming about is something very much more than a computing device.… It is a device wherewith one carries out, accurately and rapidly, certain electrical and mechanical processes which are isomorphic with certain important mathematical processes.” Weaver needed no persuasion as to the importance of von Neumann’s project, but he explained to Aydelotte that “I start out with the idea that the Institute is not a natural physical setting for such a development. I would, however, love to have you change my mind.”61