When Computers Were Human (53 page)

In this final flurry of activity, the Applied Mathematics Panel had arranged for the navy to fund the Mathematical Tables Project as a special-purpose computing laboratory. This arrangement would reunite the two divisions, joining the New York Hydrographic Office computers, who were already under navy authority, to those who had been directed by the panel. The navy, viewing the project as the seed of a larger computing laboratory, agreed to provide Arnold Lowan with IBM punched card tabulators. Naval officers would take no role in the daily operation of the center and stated that they would support “special computations for various navy bureaus, [and] also the work on basic tables.”
⁴⁵
The consolidation required one final move. The computers of the Mathematical Tables Project had to pull their mathematical books off the shelves, pack their notes, box their adding machines, and join the New York Hydrographic Office staff in the Hudson Terminal Building. The combined facility was the best office that had ever been given to the group. It had separate space for the punched card equipment, a room for the planning committee, and a view of New York Harbor.
⁴⁶

As the computers prepared to move to their new offices, Arnold Lowan, accompanied by Milton Abramowitz, traveled north to attend a conference on computation. The conference was sponsored by the Subcommittee on the Bibliography of Mathematical Tables and Other Aids to Computation and was the evidence of R. C. Archibald's faith and efforts. After a dozen years of uncertain fortune, the committee was able to organize a weekend discussion of computation for eighty-four researchers, “those chiefly active in connection with mechanical computation
on both sides of the Atlantic,” according to Archibald.
⁴⁷
Much of this new authority came from the success of the committee's journal,
Mathematical Tables and Other Aids to Computation
. In a little more than three years, the journal's subscription list had grown to nearly three hundred. The periodical could be found with every contractor for the Applied Mathematics Panel and was generally accepted as the scholarly record of human computers.
⁴⁸

41. Hudson Terminal Building, last home of the Mathematical Tables Project

The meeting, held at the Massachusetts Institute of Technology, was a
mixture of the old and the new, Depression-era methods and wartime accomplishments. L. J. Comrie mingled with John von Neumann. Arnold Lowan watched Howard Aiken operate his Mark I. From the same stage, speakers talked about difference engines and differential analyzers, the punched card machines of the
Nautical Almanac
and the relay computers of Bell Laboratories. “The conference was most notably successful,” Archibald reported, “and one heard on every side expressions of the hope that such a conference might become an annual event.”
⁴⁹
Yet this meeting was the only time that the war computers gathered as equals, the one moment when mathematicians and human computers, punched card clerks and differential analyzer operators, electrical and mechanical engineers came together and talked about their experiences with equations and numbers. A second meeting, held just three months later, gave clear signs that machine designers were starting to outpace human computers. This meeting was a press conference, held at the University of Pennsylvania, that announced the age of electronic computation. In the building that held the old differential analyzer, university officials unveiled its replacement, the ENIAC. The machine had been proposed in 1943, but it had not been finished by the end of the war. It had done its first complete calculations in November, just at the time of the MIT conference.
⁵⁰

Like the meeting that had been organized by MTAC, the ENIAC announcement pointed toward the future while not letting go of the past. The machine was fast because it was entirely electronic. The calculations did not have to pause for a gear to turn or a telephone relay to click. It was more precise than older machines because it computed digitally. Each number was represented as a series of electrical pulses rather than as the turn of a wheel or the level of a voltage. The output was more useful than that of the differential analyzer because it came as numbers, not as an ambiguous graph. Yet for all of its benefits, the ENIAC was not quite a modern computer. It was really a collection of electronic calculators. Most of these devices were nothing more than adding machines, a few did multiplication, and one took square roots. The engineers prepared for a calculation by connecting these units with large, black cables. Following the computing plan, they arranged the cables so that they took a number from a punched card, passed it to an adder, sent it to the square root unit, returned it to the adder, and finally left it on the multiplying unit. The staff called this cabling process “programming,” but only the word, not the actions, would be the legacy of the machine.
⁵¹

The press conference was a time of great pride and accomplishment, but it was also marked with a bit of impatience. As the ENIAC became operational, the University of Pennsylvania researchers had come to appreciate the limitations of their machine. “We designed [the ENIAC],” recounted Adele Goldstine's husband, Herman, “and immediately lost interest
in it.”
⁵²
In building this machine, they had recognized that they could design a more flexible machine that was controlled by a special list of electronic instructions rather than by the bulky cables. These lists would inherit the name “program.” The design team had conceived a way to modify the ENIAC so that they could control it with a primitive program, but they were more interested in building a new machine that would be entirely programmable.
⁵³

The new, programmable machines were at least two years away, and in the interim, the University of Pennsylvania had to live with the ENIAC. The press conference generated a tremendous interest in computing devices. In response to requests to study and use their machine, the school organized a seven-week course on computing machinery, a conference that would be called the Moore School Lectures. When the course convened in July 1946, the list of lecturers constituted “a Who's Who of computing of the day.” Most of the speakers came from the ENIAC project, though George Stibitz, Howard Aiken, and John von Neumann also conducted sessions. The talks had little in common with the discussions at Archibald's conference three months before. The Moore lectures dealt with circuit design and the preparation of problems for machine computation. The students were not human computers but “a select group of seasoned professional engineers and mathematicians.”
⁵⁴
The discussions made little reference to human computing groups, and few in attendance had any experience with organized calculation. No one from the Mathematical Tables Project was invited to attend. L. J. Comrie was not chosen as a representative of Great Britain. There were no computers from any Nautical Almanac Office, the Manhattan Project, or any of the projects of the Applied Mathematics Panel.
⁵⁵

Had any human computers been at the Moore School Lectures, they would have heard a somewhat fanciful history of calculating devices that ignored the contributions of workers like themselves. This history, the first of the lectures, also overlooked the punched card machines of Herman Hollerith and described the difference engine of Charles Babbage as “a special purpose [device] developed for the satisfaction of personal curiosity or as an intellectual stunt.”
⁵⁶
Historians of the conference claim that the talk was meant to “entertain and inspire,”
⁵⁷
but a close examination of the text suggests that it was an attempt to build a distinguished lineage for the electronic computing machine, a pedigree that ignored the influence of commerce and the hard labor of human computers. To many at the talk, the human computer was already starting to fade from memory. Most of the wartime computing groups had been shut down, reduced to a small remnant, or replaced by punched card equipment. The major employer of human computers, the Applied Mathematics Panel, had finally ceased operations after one last meeting. Led by Warren
Weaver, the mathematicians had gathered in their old conference room to celebrate their accomplishments. There would be those who would criticize the panel and argue that it missed an opportunity to advance the cause of mathematics, but this was not the time for such discussions.
⁵⁸
The members of the panel spoke their praise to mathematics, to science, and to their accomplishments, though no one felt it necessary to express gratitude toward the workers who had undertaken the calculations.
⁵⁹

Even though Arnold Lowan and Gertrude Blanch were not invited to the Moore School Lectures in the summer of 1946 or to the final meeting of the Applied Mathematics Panel, they were confident about the future. The navy and the Army Air Corps had guaranteed funds to sustain the operations of the Mathematical Tables Project for two more years, and the military services seemed likely to continue providing that money for a subsequent period, at least until they were in possession of a comprehensive facility with electronic computers. Even then, they seemed likely to retain the members of the Mathematical Tables Project as the support staff for the new computing machines.
⁶⁰
For the moment, the military provided the project with new and interesting computations, work beyond the traditional labor of preparing mathematical function and LORAN tables. The services asked for guided missile trajectories, radar wave deflections, and shock wave propagations. In addition to this work, the Mathematical Tables Project was now receiving requests for calculations relating to the development of atomic energy, such as problems that described the interactions of particles or analyzed reactor designs.
⁶¹

The only worrisome development for the project was a change of leadership at the National Bureau of Standards. The bureau had been the fixed point of the Mathematical Tables Project through the fluctuations of the Depression and the war. Lyman Briggs, director of the bureau, had been the champion of the project, sponsoring it for the WPA, serving as the executive manager under the Applied Mathematics Panel, and helping to obtain the postwar funding from the military.
⁶²
The planning committee of the project viewed him as the “Beloved Boss” who had supported Arnold Lowan, corrected his mistakes, both gently and not so gently, and spoken for him in the high councils of science.
⁶³
Briggs had retired in the fall of 1945 and had been succeeded by the physicist Edward Condon (1902–1974). When he first arrived at the Bureau of Standards, Condon had questioned the value of the group. He had changed his mind after a visit to the project's offices, when he met Lowan, talked with the planning committee, and observed the computing staff. Condon concluded that the Mathematical Tables Project might be able to contribute to the National Bureau of Standards, though he also observed that there was no reason for Lowan to report directly to the bureau director. After
returning to New York, he informed Lowan that the project would be guided by his new assistant, John Curtiss.
⁶⁴

Curtiss was familiar with the work of the project and sympathetic to the group. Back in 1940, he had reviewed the first volumes from the Mathematical Tables Project in the
American Mathematical Monthly
and found them quite acceptable. By training, he was a statistician, but he came from a mathematical family and was well connected in the mathematical community. His father had been a professor at Northwestern University and had served as president of the Mathematical Association of America, the professional society devoted to mathematical instruction. Curtiss had studied at the University of Iowa and at Harvard. His rise to prominence had come during the war, when he had taken a naval commission and served as a statistician for the Bureau of Ships. Most of his work concerned quality control, the statistical methods that tested batches of manufactured products to ensure that all of them operated properly or met basic standards. At the end of the war, he was recommended to Condon as a mathematician who understood the nature of organizational politics.
⁶⁵

Perhaps remembering Curtiss's review, one of the few favorable signs in the early days of the Mathematical Tables Project, Arnold Lowan welcomed the arrival of John Curtiss and invited the statistician to visit New York. Curtiss accepted the offer and, like Condon before him, came away from the project offices with a favorable impression. He decided that the group should be a key part of a new mathematical research organization that he hoped to create within the National Bureau of Standards. For the moment, he was using the name “National Applied Mathematics Laboratories” to describe his idea. In his plan, the laboratories would have four distinct units. The first would research the methods of applied mathematics, the second would consider problems of applied statistics, and the third would develop new computing machines. The Mathematical Tables Project would be the fourth and final laboratory within the group. Renamed the Computation Laboratory, it would provide “a general computing service of high quality and large capacity, for use by private industry, Government agencies, educational institutions, etc.”
⁶⁶