by George Dyson
In 1946, Bernal sent Booth on a mission to survey the state of computing in the United States. Warren Weaver, now back at the Rockefeller Foundation, agreed to sponsor Booth’s initial visit, and to follow up with a Rockefeller fellowship at the laboratory of Booth’s choice. Booth made the rounds: visiting George Stibitz at Bell Labs, Howard Aiken at Harvard, Jay Forrester at MIT’s Servomechanism Laboratory, the EDVAC group at the Moore School, von Neumann and Bigelow at the Institute in Princeton, and finally Eckert and Mauchly at the Electronic Control Company (who “were totally hostile,” he says). He even visited an abacus-based computing center at the Bank of Hong Kong in San Francisco, and gave talks to audiences ranging from ladies’ clubs to Linus Pauling’s research group.
He also “made some side trips,” including staying “for a couple of nights” with Irving Langmuir of General Electric, whose interests ranged from weather modification to protein structure, but who was most excited, at that moment, by his invention of the garbage disposal, which he termed the “electric pig.” The device was demonstrated to Booth. “I had just eaten a banana,” says Booth, “and I just threw the banana skin in. There was a horrible scrunch and the thing jammed solid. So it finished up with this Nobel Prize winner crawling about on his tummy on the floor, taking the bottom off this thing and getting the banana skin out.”3
“Then I went back to New York, and I met with Weaver,” remembers Booth. “And he said, ‘What do you want to do?’ I was quite clear by then, and I said, ‘Well the only group that is worth talking to is the lot at Princeton.’ They were the only people who had really got themselves out of the business of just waving their hands in the air and doing nothing.”4
Familiar with both the Analytical Engine of Charles Babbage and with Turing’s Universal Machine, Booth saw the IAS project as the practical implementation of these ideas. Booth, who knew Turing from Cambridge, was later asked by the National Research Council to review some of the circuits Turing had designed for the Automatic Computing Engine (ACE) being built at the National Physical Laboratory in London. “They were incredibly complicated,” he says. “And for each of these circuits, I drew an equivalent one the way I would have designed it. And they were about a quarter as expensive in components. I think some of Turing’s would not have worked.”5
Bigelow’s approach was minimalist. “Julian’s fixed principle was that you had to build things with no capacitors,” Booth explains. “If you have capacitors, then you have a limiting speed. If you don’t have capacitors, then if you can put enough juice into it you can get any speed you like.”6 The speed of the IAS computer could also be slowed down and, while debugging programs, even “stepped,” one instruction at a time. Much of the useful computation was done at 8 kilocycles, or about half throttle. There was no fixed “clock speed.” As soon as one instruction was executed, the computer proceeded to the next.
In early 1947, Booth sailed for New York on the Queen Mary, accompanied by his assistant, Kathleen Britten, later the author of an early textbook on computer programming, who had been spearheading the X-ray calculator work. Britten’s passage (and salary) was being paid by John Wilson of the BRPRA, who booked her a cabin in first class. Booth, whose passage (and fellowship at the Institute) was being paid by the Rockefeller Foundation, was booked into steerage. Bernal registered a complaint. “Eventually Johnny [Wilson] put out the money for me to go first class as well,” adds Booth.
When Booth and Britten arrived in Princeton at the end of February 1947, the question facing all Institute visitors was, where could they live? “I think there will be no problem in taking care of Miss Britten as well as of Booth,” Goldstine wrote to von Neumann. “Miss [Bernetta] Miller is trying to hire Miss [Hetty] Goldman’s housekeeper to move into one of the apartments and fix meals. In this case presumably both Booth and Miss Britten could move in with Miss Goldman’s housekeeper and observe all the necessary proprieties.” Booth and Britten—who were married in 1950, “after Kathleen had received her Ph.D. in trans-sonic aerodynamics”—were among the first group of Institute visitors to set up housekeeping in the new housing project across Olden Lane from Fuld Hall.7
In March of 1946, Veblen had proposed constructing emergency housing for the computer project personnel. “Professors Panofsky and Morse were not in favor of this proposal,” the minutes report. “They did not think it wise that the Institute divert its funds for a project of this sort for the almost exclusive benefit of the computing group.”8 In June of 1946, as engineers began showing up for work, the situation became acute. New hires were forced to commute from as far as Philadelphia and New York. “The best solution would be to rent a block of apartments from the New York Life Insurance Company,” Aydelotte reported to an emergency meeting of the trustees. “However, the New York Life has been hesitant about leasing these apartments to the Institute; and Dr. Aydelotte suspects that they prefer not to rent to Jews, although they have said nothing to this effect. The Committee decided to submit the New York Life a tentative list of tenants, on which Jews would be included, but no Hindus or Chinese.”9
This approach failed, and in desperation the surrounding community was canvassed for help. The Lawrenceville School, whose headmaster “liked the idea that these men would mingle with the boys, make speeches in chapel, etc.,” agreed to take in two or three Institute scholars, while a chicken farm near Rosedale offered four apartments, with central heating, “especially suitable for couples with children.” Marston Morse proposed that “as a last resort, cots be put in some of the rooms at Fuld Hall.” At the end of 1946, a family of four was still camped out in Fuld Hall.10
The breakthrough came in August 1946, when a cluster of wood-frame apartment houses, built to house a wartime influx of workers at the Republic Steel Company’s iron mines in Mineville, upstate New York, were put up for sale. “I sent an engineer, Mr. Bigelow, of the computer project, to Mineville the same day,” Aydelotte reported to the trustees. “He found representatives of two other universities on the spot eager to secure the houses available. Thanks to the enterprise of Mr. Bigelow, we were able to buy eleven buildings, containing thirty-eight apartments of two and three bedrooms each. These apartments are substantially built, they have hardwood floors, are insulated with rock wool against the Adirondack winters, and are fitted with storm windows, fly screens, clothes lines and garbage pails.”11
There was only one problem: Mineville and Princeton were three hundred miles apart. Under Bigelow’s supervision, the buildings were dismantled into sections, transported by rail to Princeton, and reassembled, with poured-concrete foundations, on Institute property between the Springdale golf course and Fuld Hall. The entire project was completed by January 1947, at a cost of $30,000 for the houses and $212,693.06 for the site preparation and the move—despite the complaints of nearby Princetonians who sought to halt the project “because of its deleterious effects upon the fashionable housing area which it will invade.”12
The Mineville houses were built in the same wartime style as the government housing at Los Alamos, and some who had stayed in the Los Alamos housing project during the war, under Oppenheimer’s directorship, found themselves staying in the Institute housing project after the war, under Oppenheimer’s directorship once again. By February 1947 the first seventeen families, including the Bigelows, were occupying the new apartments, and more were moving in. “Since we have been here,” Bigelow reported to Aydelotte, “we have come to know many of our neighbors quite well, not only those working in mathematics and physics, with whom we have much in common, but what is often more stimulating, we have met people working in other fields with experience and outlook different from our own.”13
In April 1947, the Institute awarded Bigelow an honorarium of $1,000 for his efforts, and Bernetta Miller reported, in September, that “there are now 30 odd children and more coming,” while “tenants are grateful for the prospect of grass now beginning to show.” Informal gatherings among the Mineville houses soon became a fixture of Inst
itute life. “We had a tendency to congregate during the evening, and we got to know one another extremely well,” remembers Morris Rubinoff, one of the second contingent of engineers who arrived in June of 1948.14
“Julian and Mary were the heart and soul of the housing project,” remembers Freeman Dyson, who also arrived in 1948. “If you had any personal problems, you went to Mary. She would give you what you needed: moral support, good advice, and just the warmth of her character. And if you had practical problems: a car that needed fixing, or a rat in the basement, or a problem to get the coal furnace either to work at all or not to work too much, Julian would always be able to fix it. Those were wonderful years.”15
With the housing crisis solved, and the computer building completed, the engineers could start building the computer—along with its power supplies and cooling equipment—rather than working on one component at a time. The computer building was adjacent to the housing project, and, says Rubinoff, “you could get to work and get home to lunch and back to work in less time than you can do it anywhere else.” It was Los Alamos all over again. “They would work till eight or nine o’clock, go out to dinner for two hours and then go back to work again,” remembers Thelma Estrin, an electrical engineer who arrived, with her husband, engineer Gerald Estrin, in June 1950, in the middle of the final push to finish the machine. “Sometimes they would work all night.”16
“I had just completed my doctorate,” adds Gerald. “I had never heard of a computer; I didn’t know anything about it,” but while searching for a job, he was told that “there’s an interesting project going on at the Institute of Advanced Study.” Von Neumann invited the Estrins for a visit, and hired them on the spot. “Von Neumann liked to be in the middle of stuff,” says Gerald. “We stepped off on the grounds and fell in love with the place. There were little signs on the grass, tiny ones, that said, ‘Please Don’t Walk on the Grass.’ ” The Estrins spent the next three years at the Institute, before moving to Israel to build a copy of the IAS machine. “They were very concentrated years. Being in such a small group, I really learned about every part of the computer, helping in every way that I could.”17
In an age when most microprocessors require only a single voltage—somewhere between 1 and 5 volts—it is difficult to comprehend how many different voltages it took to run a vacuum tube computer. There were seven main branches leading from the 120-volt, three-phase power lines entering the building from outside. First there were three branch circuits that supplied three-phase power to the vacuum tube heaters: about 6.5 kw to the arithmetic unit, and 1.5 kw to the memory. Second, DC power was supplied to the core of the computer through four separate rectifiers, subdivided further into twenty-six different voltages, ranging between –300 and +380 volts. Finally, the Williams tube deflection circuits required a regulated 1075, 1220, and 1306 volts. A useful current in one circuit could easily induce noise somewhere else, not to mention noise introduced by transients in the incoming power lines. Initially, there was so much trouble with noise in the DC power supplies that a 300-volt, 180-ampere-hour battery house was built outside the computer building, to supply clean DC power until the memory could be better shielded and more stable power supplies designed.
None of these voltages meant anything without reference to a common ground. This resulted in a landmark known, briefly, as the Rosenberg ground. “We had developed the machine in two sections,” James Pomerene explains. “And at some point, due to things which I don’t remember anymore, when we came to put them together, what I called ‘ground’ in my design was at a different voltage level than what Rosenberg called ‘ground.’ And so for a while we had batteries that made the adjustment between my ground and the Rosenberg ground.”18
The computer consisted of four “organs”: input/output, arithmetic, memory, and control. The choice of memory drove the design, yet was the last element to be resolved. “Once the form of the high speed memory has been decided most of the other components of an electronic computer become semi-invariant,” Booth and Britten observed in their report on their stay at IAS.19 The anticipated all-digital Selectron memory tube from RCA, if not yet in existence, was specified accurately enough that the rest of the computer could be designed around the assumption that a plug-and-play memory tube would be forthcoming in time.
Midway in complexity between the bit-level components such as 6J6 toggles and these system-level organs were 40-fold registers that stored, transferred, and shifted data, in parallel, 40 bits at a time. A register known as the accumulator provided access from the arithmetic unit to the memory, and the memory register provided an exit from the memory going the other way—analogous to separate intake and exhaust valves serving the individual cylinders in the engine of a car.
All these registers were “double-rank,” containing two parallel rows of 6J6 toggles, with entry in and out of the toggles controlled by two additional rows of gates. This redundancy prevented bits being lost in transit: all data were replicated at the destination before being cleared from the source, in the same way the transmission of a data packet across the Internet is not considered complete until the packet signals that it has arrived intact.
Shift registers, as Leibniz had demonstrated 260 years earlier, could perform binary arithmetic simply by shifting an entire row of binary digits one position to the right or left. Data were never transferred directly between adjacent toggles; instead, the state of each individual toggle was replicated upward into a temporary register, the lower register was cleared, and then and only then were the data shifted, diagonally, back down into the original register. There was no lower bound to how slowly the computer could be stepped through a sequence of instructions. Unlike the well-behaved physical marbles that Leibniz had imagined shifting from column to column in 1679, electrons were always looking for an escape.
“Information was first locked in the sending toggle; then gating made it common to both sender and receiver, and then when securely in both, the sender could be cleared,” Bigelow explained. “Information was never ‘volatile’ in transit; it was as secure as an acrophobic inchworm on the crest of a sequoia.” Data were handled the way ships are moved through locks in a canal. “We enjoyed some interesting speculative discussions with von Neumann at this time about information propagation and switching among hypothetical arrays of cells,” remembers Bigelow, “and I believe that some germs of his later cellular automata studies may have originated here.”20
“We did not move information from one place to another except in a positive way,” emphasizes James Pomerene. “That is used absolutely universally now. I think we were the first to do it. And I regret that we didn’t patent it.” Patentable inventions were being generated right and left. “The original patent agreement provided that the Institute would have title to patents, but they would pay to the inventor all royalties in excess of their costs,” Pomerene adds. “Pretty nice!” No patents, however, were ever applied for. “We were young, eager engineers. We were most interested in getting the machine going than the filing of patents,” he explains.21
In April 1946, von Neumann drafted a patent policy that “strikes a reasonable middle position between leaving everything to the employee or taking everything for the Institute.” Employees agreed to assign their rights to the Institute, while the Institute agreed that “it will promptly and at no expense to the Employee have prepared, filed, and prosecuted an application for United States Letters Patent (and for patent in countries foreign to the United States, if it so decides) on each invention which the Institute determines is or may be useful to it.” Furthermore, “the Institute agrees to pay to the Employee all royalties, if any, received … on each invention … over and above the total cost to the Institute of procuring the patent or application therefore.”22
This was greeted with enthusiasm by the engineers. The Institute retained a patent attorney, whose assessment was that the project offered a wealth of patentable inventions—and the machine had not even been constructed yet. “The
Institute for Advanced Study, had it asked for it, could have gotten from each engineer a release, and produced a large endowment compared to anything they now have,” says Bigelow. But it would have cost a significant amount of money to secure, let alone defend, those patents—an approach the Institute was unwilling to take. It was Abraham Flexner, after all, who had announced, in 1933, that “the moment that research is utilized as a source of profit, its spirit is debased.”23
In mid-1947 the original patent agreement was weakened, unilaterally, by a decision to turn most patent rights over to the government. “For the probably few exceptional developments that give rise to extremely valuable commercial applications,” Goldstine reassured the engineers on June 6, “the engineering staff will recommend to the Director of the Institute that an application be prosecuted directly by the Institute.” This was an empty promise, for Goldstine had already agreed, at a meeting with the Office of the Chief of Ordnance in April of 1947, “that any papers or reports covering the logical aspects of the Computer would be regarded as scientific publications, and therefore accessible to all interested scientists.” This would undercut patent claims to most of the inventions made so far. “To prevent any commercial interest from attempting to exploit what should belong to the scientific community,” Goldstine advised, “I would appreciate having you send a copy of the report entitled Preliminary Discussion of the Logical Design of an Electronic Computing Instrument, by A. W. Burks, Herman H. Goldstine, and John von Neumann, dated 28 June 1946, to the Patent Office with a request that they treat this as a publication in fact.” In June of 1947, Goldstine, Burks, and von Neumann gave a sworn deposition stating that “it is our intention and desire that any material contained therein which might be of a patentable nature should be placed in the public domain.”24