Lanczos never served as a traditional planner, never prepared a computing plan, never oversaw the computing staff. Instead, he acted like a visiting scholar, an expert on the methods of calculation who could teach new techniques to Gertrude Blanch, Ida Rhodes, and the other members of the planning committee. Starting in the winter of 1944, he offered seminars on numerical methods, advertising them through the Applied Mathematics Panel and nearby New York University. “His lectures attracted a wide audience, not only from the Project, but from mathematicians at local universities,” recalled Ida Rhodes.91 These lectures brought a small glimmer of respect from the Applied Mathematics Panel. By March, they were starting to address Lowan in more informal terms and to refer to the project as “Lowan’s Group.”92 More important, they were pointing to the Mathematical Tables Project as a successful computing organization. They encouraged prospective computers to visit the organization and copy its operating procedures. Among the visitors that winter was a group of scientists that was preparing to build a computing laboratory for the Manhattan Project in Los Alamos, New Mexico.
Computing laboratories were familiar institutions to the atomic scientists, as most of the major university physics departments had some kind of computing staff. Yet these academic laboratories were far smaller than the scale demanded by the effort to build the bomb. The computing office at the University of Chicago, one of the larger contractors to the Manhattan Project, consisted of just one faculty wife and a few graduate students.93 The senior leaders of Los Alamos wanted to model their organization on the largest computing offices of the Applied Mathematics Panel, the Thomas J. Watson Bureau and the Mathematical Tables Project.
The Watson Laboratory received the first visitors from the Los Alamos staff, a couple named Mary and Stanley Frankel. Stanley Frankel had managed the computing bureau at the University of California that had handled the calculations for isotope separation, the problem of extracting the type of uranium that could be used in a chain reaction.94 It was a small group with none of the equipment that could be found at Columbia. The Frankels spent about three days at the bureau, working with director Jan Schilt and a young graduate student named Everett Yowell (1920–).95 Yowell had the rare distinction of being a second-generation computer. His father, also named Everett Yowell (1870–1959), had computed for the Naval Observatory from 1901 to 1906. The elder Yowell was part of the generation that had known Simon Newcomb, Myrrick Doolittle, and the computers of 1918 Aberdeen.96 After his service as a computer, the senior Yowell had become a mathematics instructor at the U.S. Naval Academy and then had returned to the family home in Ohio to become the head of the Cincinnati Observatory. The younger Everett Yowell spent his youth playing in the halls and chambers of the observatory. His father taught him how to use a telescope, how to record the position of an object, how to reduce astronomical data. His texts were the classic books: Crelle’s Tables, Newcomb’s Positional Astronomy. At the age of twelve, Yowell assisted his father on an expedition to study a solar eclipse. He entered college with a firm understanding of traditional astronomy and arrived at Columbia knowing the methods of hand computers.97
During his first year at the school, Yowell had little contact with the Watson Bureau. “I was sort of drafted as an operator during the summer of ’42,” he recalled. The facility was beginning to do calculations for war research and had lost much of its skilled staff. Eckert was in Washington, and many of the younger workers had left for the military. Yowell learned the techniques of punched card computation by studying Eckert’s Orange Book and by experimenting with the machines. Over the course of a year, he became an expert on wiring plugboards, the mechanisms that controlled the tabulators. Plugboards were flat panels, about the size of a large notebook, that were filled with holes that represented the different operations of the tabulator. By connecting the holes with short cables, Yowell could direct the flow of data through the machines and implement the methods of the Orange Book.98
The Mathematical Tables Project received its Los Alamos visitor a few weeks later, a researcher named Donald Flanders. Knowing the limitations of punched card tabulators, Flanders was organizing a hand computing group that would be known within the laboratory as T-5. The T-5 group was a typical wartime computing office with about twenty computers.99 It earned a certain distinction because of its association with the physicist Richard Feynman (1918–1988). Feynman was a junior staff member at Los Alamos, and he worked with Stan Frankel to prepare computing plans for T-5. One of his plans recalled the work of de Prony or the early computing floor of the Mathematical Tables Project. Feynman divided the computation into specific tasks, such as additions, square roots, and divisions, and then assigned each task to a specific computer. Like de Prony’s computers, one T-5 computer did nothing but add. A second took square roots, using a mechanical calculator. A third only multiplied.
Instead of creating computing sheets, Feynman used standard index cards to hold the results of the computations. These cards passed from computer to computer as the calculation progressed. “We went through our cycle this way until we got all the bugs out,” recalled Feynman, and it “turned out that the speed at which we were able to do it was a hell of a lot faster than the other way, where every single person did all the steps. We got speed with this system that was the predicted speed for the IBM machine.”100 This claim, the notion that the T-5 computers could equal the speed of a punched card office, was tested late in the war when Feynman organized a contest between the human computers and the Los Alamos IBM facility. He arranged for both groups to do a calculation for the plutonium bomb, the “Fat Man.” For two days, the human computers kept pace with the machines. “But on the third day,” reported an observer, “the punched-card machine operation began to move decisively ahead, as the people performing the hand computing could not sustain their initial fast pace, while the machines did not tire and continued at their steady pace.”101
The competition between the T-5 computers and the punched card equipment is generally reported as a scientific version of the tortoise and hare fable, a story that predicted the triumph of computing machinery and a sign that human computers would soon be replaced by the electronic computer. The result can also be interpreted the other way, as suggesting that, through much of the war, human computers were closely matched to their mechanical counterparts. Since human computers did not demand the kind of preparation required by punched card machines, they outperformed the tabulators on many military calculations. The Los Alamos scientists relied on human computers to check large calculations. The plutonium bomb calculations were compared to a similar set of numbers that had been prepared on Howard Aiken’s Mark I at Harvard.102 As the war entered its last year, human computers might still be considered the equals of automatic computing machinery.
CHAPTER SEVENTEEN
The Victor’s Share
We cannot retrace our steps.
Going forward may be the same as going backward.
We cannot retrace our steps retrace our steps.
Gertrude Stein, The Mother of Us All (1947)
SOMETIME IN 1944, computers became “girls.” The University of Pennsylvania hired “girl computers”; Warren Weaver started calling Applied Mathematics Panel computers “girls”; Oswald Veblen, who had once led a team of computing men, used the term “girls”; George Stibitz began ranking calculating projects in “girl-years” of effort.1 One member of the Applied Mathematics Panel defined the unit “kilogirl,” a term that presumably referred to a thousand hours of computing labor, though in at least one letter it suggested an Amazonian team of computers.2 L. J. Comrie, in an article entitled “Careers for Girls,” stated that girls “can be made proficient and give good service [as scientific computers] in the year before they (or many of them) graduate to married life and become experts with the housekeeping accounts.”3 Even at this date, computing was not the sole domain of women. It was really the job of the dispossessed, the opportunity granted to those who lacked the financial or soc
ietal standing to pursue a scientific career. Women probably constituted the largest number of computers, but they were joined by African Americans, Jews, the Irish, the handicapped, and the merely poor. The Mathematical Tables Project employed several polio victims as computers, while the Langley research center kept an office of twelve African American computers carefully segregated from the rest of the staff.4
For all of the Applied Mathematics Panel computers, female and male, the end of the war was first glimpsed on September 25, 1944, when Warren Weaver received a letter reminding him that the National Defense Research Committee was “a war time agency which will go out of existence at the end of the war.” The letter informed the panel that they needed to prepare their plans for demobilizing the organization.5 There was scant military news to support the idea that the war was nearly over and that research was no longer needed. Germany, though in retreat, still commanded a strong military that could inflict substantial injury on Allied forces. Japan promised a long and bloody fight for the control of its home islands. Still, the senior leadership of American science felt that the last months of the war “would involve almost no research and very little development work.”6 They viewed the National Defense Research Committee as “an emergency organization” that “deserved to die a dignified death.”7
“For planning purposes,” Weaver told the members of the Applied Mathematics Panel, “we are assuming that Germany will fall by November 15, 1944.”8 Shortly after the end of the European war, the panel would start terminating research programs and liquidating contracts. The panel would have to provide guidelines on how project records should be preserved, which materials could be published, and when computing groups could be disbanded.9 As the panel developed its demobilization plans, Warren Weaver recorded that the members were “becoming deeply interested in the post-war possibility of the important new computing techniques, which are being developed during this war.” They discussed computing machines over an eight-month period, sometimes reviewing their progress as part of a formal meeting, at other times sharing speculations about Stibitz or Aiken or the Aberdeen Proving Ground over lunch. “Quite outside of the mere furnishing of labor-saving assistance,” Weaver explained, these machines “may, in fact, have theoretical consequences of very great significance.”10
In its discussions, the panel was beginning to recognize that there might be a substantial demand for computing services after the war. Weaver observed that the navy was using Aiken’s Harvard machine for three shifts a day, seven days a week. From this, he concluded that “two machines could be kept busy on [naval] Ordnance work alone” and that the navy should consider building “a central machine … under a broad agency in Washington.”11 He argued that the navy should make this last machine available to both military and civilian researchers, as it might “take care of a variety of problems.”12
All of the major computing laboratories had a full load of requests that winter. The New York Hydrographic Office was completing LORAN tables for the Pacific Ocean. The group’s leader, Milton Abramowitz, had lost almost half of his staff, yet he had reduced the computing time from six weeks per table to four. He had received some help from the Mathematical Tables Project staff, to be sure, but he had achieved much of this improvement by developing a new method of preparing the tables.13
The Mathematical Tables Project found that it could add only a few numbers each week to the unfinished WPA volumes. It was handling about two dozen calculations at once. A computer might work on two or three or four assignments, advancing each a little each day. That winter, the big calculation was a rather gruesome optimization problem. It attempted to determine the “combination of [high explosive] and [incendiary bombs] that should be used for specific Japanese targets.” High explosives would shatter Japanese buildings but not set them on fire. Incendiaries would light fires, but the damage would be limited if the buildings remained intact. A well-planned combination of the two kinds of bomb would create firestorms that would destroy Japanese cities.14 The mathematical analysis of this problem had been assigned to Jerzy Neyman, who, again, had failed to meet his deadlines and keep his expenditures on budget. The Mathematical Tables Project computers had to work at a stiff pace in order to complete the work before the start of the Pacific bombing campaign in March 1945.15
With two years of experience, the Applied Mathematics Panel was becoming reluctant to label every computing problem as essential to the war effort. When General Electric requested thirty computers, the panel demurred. General Electric wanted these computers to prepare tables for the antiaircraft systems of the B-29 bombers, the planes that would lead the bombing campaign against Japan and would carry the atomic bomb. The work was important, but Warren Weaver informed General Electric that the Applied Mathematics Panel had no computers to spare and “suggested that [the company] use the computation facilities of some insurance company.”16
General Electric was not alone in its rejection. The Applied Mathematics Panel accepted few new problems that winter. Those that went from the panel to the Mathematical Tables Project or the Thomas J. Watson Astronomical Computing Bureau tended to be expansions of old calculations. The work was becoming routine, and hence it is probably not surprising that Warren Weaver became interested in an unusual problem that had no obvious mathematical solution. In March, he told the panel of an issue that was “being held very secret.” The secret would not last long, he informed them, because “all Police Departments in the US have been informed about it.” Nevertheless, the panel should treat the request as confidential, for any public discussion of the problem might incite panic among American citizens. Large balloons had been discovered across the country. They were constructed out of rice paper and were about thirty feet in diameter. Most had landed in Oregon and California, but several had made it over the Rocky Mountains and landed as far east as Michigan. They apparently came from Japan, but no one knew where they were launched or what they were designed to accomplish. Some members of the National Defense Research Committee believed that the Japanese planned “to use bombs in dry weather to start large scale forest or crop fires.” Others suggested that the balloons might be intended to carry poisonous gas or deadly viruses. Weaver could offer no theory of his own, though he argued that they must be weapons of last resort, as any damage to the United States mainland would be “a poor bargain in view of our demonstrated ability to put bombs on Japan.”17
Weaver asked the panel members to estimate the number of balloons that had reached North America, given “the number of recoveries which have been made and of the tolerable assumptions concerning the probability of recovery.” It seemed likely that most of the balloons were lost at sea and that many of the remainder had landed in remote sections of the coastal mountains or in the deserts of the western states. He also asked the mathematicians to consider “what other questions related to the problem can have rough numerical answers.”18 With hindsight, it is difficult to understand how Weaver’s announcement could have been anything more than a reaction to an unsettling piece of news. Given how little the military knew about the balloons, the Applied Mathematics Panel would have found it difficult to prepare a mathematical analysis of the attacks. Weaver’s response may have also been shaped by a growing anxiety among the members of the panel that “mathematics is not adequately represented” in the country’s plans for postwar science.19 President Roosevelt had recently announced a new organization, the Research Board for National Security, to coordinate scientific research after the war. Only a single mathematician, Oswald Veblen, had been appointed to this committee.
As a whole, the members of the Applied Mathematics Panel had no coherent vision for postwar mathematics. Few of their number showed any interest in building a new institution for postwar mathematical research. John von Neumann argued that mathematics should have no organizational hierarchy, as “decentralization would be the more efficient manner of organization.”20 Richard Courant, from New York University, wanted a small executive body to coo
rdinate research but warned that “the enthusiasm among capable scientists for war research will abate after a short time.” Organized mathematics research would have “no attraction whatsoever for scientists of high quality unless research can be conducted with as few strings attached as possible.”21 Only Warren Weaver seemed interested in promoting a government office for mathematical research, yet he described such an office only in terms of the way that it would operate. The American government should “set up laboratories in a place where scientific people like to be,” he stated. It would also have to “pay high salaries” and would need to “provide better materials and greater leisure than are commonly provided in government service.”22
The death of Franklin Roosevelt in April introduced new uncertainty into the discussions of the Applied Mathematics Panel. “To many,” wrote historian David McCullough, “it was not just that the greatest of men had fallen, but that the least of men—or at any rate the least likely of men—had assumed his place.”23 The mathematicians knew little about his successor, Harry Truman. No one knew how the new president felt about government sponsorship of science or whether he would be able to get the approval of Congress for any ambitious scientific program. The members of the Applied Mathematics Panel were somber when they convened their first meeting after Roosevelt’s death. Warren Weaver was ill, and Thorton Fry held the meeting at Bell Telephone Laboratories. They heard reports from several projects, approved a few minor actions, and accepted a request for calculations from New York University. They assigned the work to the Mathematical Tables Project but asked that Arnold Lowan “secure some time estimate” before starting the work. Just as the meeting broke up, the panel agreed that “meetings in the future [will] be held at intervals of 3 to 4 weeks until further notice.”24
When Computers Were Human Page 36
