When Computers Were Human

by David Alan Grier

  In the history of human computers, this is the one moment that might have provoked a labor action, a strike by those who worked with numbers, but the computers were unwilling to back Arnold Lowan. The United Public Workers made one final attempt to start a protest by charging that Curtiss’s plan “would certainly lead to the Bureau’s having two relatively inefficient computing groups,” but they found that there was no support for their position among either the computers or the scientists.88 By midsummer, they had abandoned their efforts on behalf of the Mathematical Tables Project. That July, the new Computation Laboratory opened its doors in downtown Washington, and simultaneously, Arnold Lowan changed the name of his office to “Computation Laboratory, New York.”

  During the ten-year history of the Mathematical Tables Project, John von Neumann was a distant figure whose letters offered nothing but vague encouragement and deferred hope. He never visited the project offices, though he made time to visit the Admiralty Computing Service in England, and he never offered an unconditional blessing to Arnold Lowan. Only in the final days of the Mathematical Tables Project, just as his letter was sealing the project’s fate, did von Neuman ask Arnold Lowan for computational assistance. He asked whether the computers would test a new technique called “linear programming.” Von Neumann was not really interested in the results of the test computation, just as he was not especially interested in the future of the Mathematical Tables Project. He requested the calculation because he wanted to use the human computers as surrogates for computing machines, as a means of projecting the operation of a programmable electronic computer.

  John von Neumann may have initiated the request, but he did not invent the method of linear programming. The technique was developed by a student of Jerzy Neyman named George Dantzig (1914–). Dantzig had spent much of the war analyzing operational problems for the Army Air Corps. A typical problem looked for the best way of storing spare parts for aircraft. The idea was to find the number of storehouses that would provide the best access to parts yet at the same time minimize the expense of keeping a large inventory.89 Such problems had posed serious difficulties for the air corps, according to one of Dantzig’s contemporaries, and had “required the labors of hundreds of highly trained staff officers.”90

  Dantzig had been able to test his method of linear programming only on small, simple problems, as the work had the same kind of demands as least squares or simultaneous equations. A small problem, such as the task of locating two or three storehouses, was simple and straightforward. As problems grew, the amount of calculation expanded rapidly. If the problem was doubled to six storehouses, then the calculations would require eight times the effort. Dantzig would have been restricted to these simple problems had he not been given the opportunity to present his method to von Neumann. He visited von Neumann at the Institute for Advanced Study and encountered an impatient mathematician. “In under one minute, I slapped the geometric and the algebraic version of the problem on the [black]board,” he recalled. Von Neumann quickly grasped the nature of the problem, took the chalk, and “then proceeded for the next hour and a half to lecture me on the mathematical theory of linear programs.”91 As it happened, Dantzig’s work was related to research that von Neumann had already completed. Von Neumann wanted to see Dantzig’s method tested on a large problem, one that had a classic standing in the economics literature. This problem attempted to identify the cheapest possible diet from among seventy-seven different foods. The list of foods began with wheat flour and ended with strawberry preserves. The diet had to provide 3,000 calories and minimum amounts of eight different nutrients.92 The calculations for this problem included 29,856 additions, 15,315 multiplications, and 1,243 divisions.93 Von Neumann had hoped that the Aberdeen Proving Ground might agree to do this calculation on the ENIAC, but the ballistics researchers refused the request, so he turned to Arnold Lowan and the Mathematical Tables Project.94

  The calculations began in mid-April, just as Arnold Lowan was preparing for his meeting with John Curtiss. The work required twenty-five computers and was overseen by a junior member of the planning committee. It was a somber time in the office, but the work progressed steadily. Dantzig monitored the efforts of the computers and kept von Neumann informed of their progress. Von Neumann would take the letters from Dantzig, turn them over, and calculate the amount of time that the ENIAC would spend on the same work. Dantzig’s last letter reported that twenty-five computers had completed the work in twenty-one days. In half a page of pencil scratching and little diagrams, von Neumann concluded that the ENIAC could do the same work in about nine hours.95 This was the only number that he would ever use from the Mathematical Tables Project. He never saw the final calculations. “The setting up of a computational procedure for this type of problem is the main objective,” acknowledged the final report on the work, “rather than the solution of this specific diet problem.”96

  The computers began to disperse almost as soon as Dantzig’s calculations came to an end. Some went to Washington; a few, including Ida Rhodes, started to learn about the new electronic computers;97 most started looking for jobs in New York City. By January 1949, Lowan had fewer than fifteen computers working for him, and they were all assigned to old problems that had been started under the authority of the WPA. Struggling to keep his facility operating, he still hoped that Philip Morse would be the savior of his New York office this final time, but he was coming to accept that there would be no future for his organization. “The fact that you have not replied to my [last letter],” he wrote to Morse, “would seem to substantiate my feeling that you are seemingly unable to do anything which would change the trend of events.”98 The return mail brought the news he had dreaded. “I very much fear you are right,” Morse had written. “It seems that the situation you are up against is well nigh unbreakable. I don’t like it but there it is.”99 Having served, in its last moment of public glory, as a proxy for the digital electronic computer, the Mathematical Tables Project closed its doors for the final time on Friday, September 30, 1949.

  CHAPTER EIGHTEEN

  I Alone Am Left to Tell Thee

  The real power, the power we have to fight for night and day, is not power over things, but over men. …

  George Orwell, 1984 (1949)

  EVEN THOUGH THE FINAL DAYS of the Mathematical Tables Project were filled with drama and emotion, and even though they engaged an unusual cast of characters, they were nonetheless part of a conventional scientific decision. Two scientists, each seeing a different direction for a project, shared a common claim over a single pool of resources. One of the two, John Curtiss, ultimately prevailed and steered the course of the Mathematical Tables Project to his liking. At the end, no one questioned the credentials of the loser, Arnold Lowan, or thought that he was unfit to lead a computing group or believed that he was incapable of developing new methods of scientific calculation. Most of the American computing groups of the Second World War ended in such a manner. As the country returned to the peacetime economy, the government reduced the budget for scientific and engineering research. In response, the leaders of the research laboratories, with or without an argument, cut their staffs, including their human computers.

  As a world war gave place to a cold war, the United States began to rebuild and expand its research laboratories. The navy supported science through its Office of Naval Research. The air force created a private research company in California, the RAND Corporation. A dozen universities created laboratories in order to provide research services to the military. Of these laboratories, the only facility to develop a large computing staff was the one devoted to the mathematical methods of computation, the Institute for Numerical Analysis at the University of California, Los Angeles (UCLA).1 The institute, a division of the National Bureau of Standards, did much productive work during a period of controversy and criticism. This criticism was not concerned with the quality of work done at the Institute for Numerical Analysis or the need for standard methods of computation or ev
en the proper allocation of the institute’s budget. It touched on the right of the institute’s staff to claim the title of scientist and to hold stewardship over the country’s scientific legacy. This criticism came not from within the family but from without, not from the greater community of scientists but from the American public. At the center of this criticism was a quiet but ambitious human computer, the former technical leader of the Mathematical Tables Project, Gertrude Blanch.

  In the spring of 1948, Gertrude Blanch was having no part of Arnold Lowan’s confrontation with John Curtiss. While Lowan was rallying the supporters of the Mathematical Tables Project, Blanch was quietly going about her business. In earlier years, she might have avoided the controversy by disappearing into John von Neumann’s test of linear programming, but she needed no such excuse this time, as she was shortly to depart for California. Curtiss had asked her to be the assistant director for computation at the Institute for Numerical Analysis, a position that would oversee a computing office and engage in some mathematical research. She viewed the new job as an opportunity to move away from Lowan’s shadow and establish her own reputation. Already she was preparing articles for Mathematical Tables and Other Aids to Computation and talks for the Eastern Association for Computing Machinery, a new society for those interested in electronic computers.2

  Blanch spent the month of April cleaning her desk, packing her books, and retrieving those few trinkets that reminded her of all that the Mathematical Tables Project had accomplished. On her last day, the computers took a break from their linear programming calculations and gave her a farewell party. Surrounded by adding machines and piles of paper, they shared one last plate of food with Blanch, offered her a parting gift, and left her a card expressing their best wishes for her future. The computers came from the working-class neighborhoods around New York City: Brooklyn, Yonkers, Jersey City, the Bronx, Harlem, and the Lower East Side.3 A few of them would be offered positions in Washington, D.C., but most would soon be looking for work in New York. None of them would be following Blanch.

  At the start of May, Blanch emptied her apartment and said good-bye to her sister and to her nephews and nieces. She had no family of her own to keep her in New York, no man to tie her heart to Brooklyn, at least none that had been discovered in a routine security check.4 Her train took three days to make the trip to California, time that would allow her to think about the future as the scenery went past the windows. The train lingered by the stockyards of Chicago before passing through the farms of Iowa and climbing the Rocky Mountains. She had been west once before, a visit to relatives in Mexico, but this trip was a fundamental change in her life. She was leaving her friends and her home to pursue one of the new opportunities that had been created for American scientists. There was more than a little risk in the trip, as the United States Congress had become more willing to support scientific research but had not entirely determined how the institutions of democracy should interact with scientists and scientific laboratories. The Research Board for National Security had failed in its first year and had not been replaced. The National Bureau of Standards was acting as the interim coordinator of postwar research.5

  Blanch was more concerned with personal issues than with national policy. She was nearing her fiftieth birthday, and she truly wanted to make a mark on the world. The war had brought many women into science, but only a few of them were finding a place in the time of peace. Ida Rhodes and Irene Stegun, both members of the Mathematical Tables Project planning committee, had jobs at the National Bureau of Standards. Mina Rees, the assistant to Warren Weaver, had found a good position with the Office of Naval Research. Grace Hopper, the Vassar professor who had “considerable experience” with computing machines, had received a commission in the navy and was making her career in the laboratory of Howard Aiken. Blanch believed that the Institute for Numerical Analysis would be her intellectual home. John Curtiss had told her that it would function like an academic research department, “a group of peers working together as a university,”6 though Curtiss had clearly indicated that the institute would be “a computing service containing both standard equipment and high speed equipment” for the aircraft industry of Southern California.7 As she approached Los Angeles, she must have pondered her new role. Would she be able to pursue her own research, or would she spend her days doing calculations for others? Would she have a place of her own, or would she always be subservient to someone else? At the end of her career, would she be able to leave a legacy for others to follow, or would she have been more productive if she had never left that well-appointed photographic equipment office and taken a job in a work relief agency?

  Blanch found a new apartment across the street from the UCLA campus, a short walk from her office. The Institute for Numerical Analysis was housed in an old rehabilitation hospital, a temporary wooden structure that had been constructed during the war. The building, which was only a single story tall, was located in an undeveloped section of the campus. When she first entered the institute building, in May 1948, Blanch encountered a scene reminiscent of the first days of the Mathematical Tables Project. The building was empty except for a corner office that was occupied by the institute’s administrator, Albert Cahn. The computing office was an open room furnished with large tables. The computers would be able to open the windows to catch the breezes from the ocean or work outside on a central veranda that looked across the scrub growth toward the west.

  The institute administrator, Albert Cahn, was no Arnold Lowan and had no authority over the scientific work of the organization. He handled correspondence for the institute, managed the organization’s money, and promoted the new computing service.8 In later years, the staff of the institute would tell the story that Cahn had been hired for the new institute because he had been found “sleeping in a common room at Princeton [University] without a job.”9 He had spent the war at the Chicago office of the Manhattan Project. Holding only a master’s degree in physics, he had handled the tasks that were too simple, too routine, or too boring for the senior scientists with doctoral degrees. After the war, he had gone in search of an academic position, hoping that the cachet of the atomic bomb would help him find a good job. Receiving no offers, or at least none that interested him, he drifted back to the universities of the East Coast and found his way to Princeton.

  Only two other offices were occupied when Blanch arrived. John Todd and Olga Taussky had come from England to be the first visiting mathematicians at the institute. It was “a very welcome and perhaps deserved change,” remarked Taussky. The war had been invigorating and exciting, but it had taken its toll. She had found the demands of maintaining a home to be “rather strong on top of the mathematical activities.”10 Neither of the two was pleased by the prospects for British mathematics. The Admiralty Computing Service had little to do after the war, and many of its computers were preparing to join a new British national mathematics laboratory. Taussky and Todd were invited to be part of this group, but they “didn’t think much of [the] leadership.”11 Discovering that it would be difficult to start a new research program within the University of London, they accepted an invitation to come to the institute. Much of the institute’s work would be conducted by visitors, such as Todd and Taussky. It would have only eight permanent mathematical staff. The rest would come and go as the institute considered different problems and ideas.12

  By the end of the summer, Blanch had assembled a computing office that resembled the wartime Mathematical Tables Project. The group was “founded partly on the older hand-machine techniques,” observed the institute director, “and partly on theories radically new in numerical analysis.”13 It had sixteen computers, seven men and nine women, who had been recruited from the students of UCLA and the residents of West Los Angeles.14 They had electric desk calculators, machines that could multiply and divide as well as add and subtract. They also had an accounting machine that they could use as a difference engine. Their first project was a new variation on the classical calcula
tions of comet orbits, which had been requested by the U.S. Army. The army was building upon the spoils it had seized at the German rocket research center of Peenemünde and was developing a new generation of high-performance rockets. The work was still in its preliminary stages, but army officers were already thinking about the difficulties of boosting a payload beyond the atmosphere. To prepare for the time when they could put a satellite in outer space, they requested tables of “rocket and comet orbits” from Blanch’s computers.15

  43. John Todd, Olga Taussky, and John Curtiss on excursion from Institute for Numerical Analysis

  Most members of the institute staff believed that the human computers would be temporary workers, quickly replaced by a new “automatic computing machine,” as electronic computers were then called. Initially, the National Bureau of Standards intended to purchase a computing machine for the Institute for Numerical Analysis. Three American companies were offering to build such machines: Raytheon Corporation of Massachusetts, Electronic Research Associates of Minnesota, and UNIVAC Corporation of Pennsylvania, which had been formed by the ENIAC designers, J. Presper Eckert and John Mauchly. John Curtiss was not satisfied with any of these proposals and convinced the senior staff at the bureau that the Institute for Numerical Analysis should build its own machine. He argued that a machine could be built quickly and that it would develop improved computing technologies.16

  Under the best of circumstances, a speedy construction would take two or even three years after the first designers arrived at UCLA. To provide an interim machine-computing service, the institute acquired a device called the Card-Programmed Calculator. The Card-Programmed Calculator stood halfway between the IBM punched card tabulators and the new electronic computers. It had been created by engineers at an IBM customer, Northrop Aircraft Company. The engineers had recognized that they could create a fairly powerful computing device by connecting a new IBM card tabulator to one of the company’s accounting machines. IBM had not intended for the machines to be joined in this fashion, but they quickly realized the value of the combination and adopted it as an official product. They eventually leased about 700 of these machines, far more than any of their early electronic computers.17