by Kurt W Beyer
Aiken forcefully stated that the focus of the Symposium should be the transmission of ideas concerning the design of calculating machines for scientific applications. He outlined the growing demand for more powerful calculating methods in the mathematical and physical sciences, including the fact that many recent scientific and engineering developments were based on nonlinear effects. Since the only methods available to solve differential equations involved labor-intensive expansions of infinite series and numerical integration, it was imperative that scientists and engineers embrace the new field of calculating machinery in the name of progress.21
Aiken did admit, however, that there might be some room for applications in quantitative social sciences such as economics. He invited his Harvard colleague Wassily Leontief to comment on the applications of Mark I in the field of economics. In his paper titled “Computational Problems Arising in Connection with Economic Analysis of Inter-industrial Relationships,” Leontief described “input-output” economics, an early example of econometrics in which segments of the economy are mathematically represented by large systems of simultaneous linear differential equations. These equations described the not so obvious connections between industries as diverse as steel, textiles, and paper.22
Mathematical models, brought to life by large calculating machines, would provide economists the ability to answer a variety of once inconceivable questions: When a steel mill closes in Pennsylvania, how is the output of textiles affected? When the price of steel goes up, how does the price of automobiles change? With the help of Mark I, Leontief demonstrated that a general wage rise of ten percent in the United States would increase the cost of living by four percent. Leontief’s work was the first application of a calculating machine in the social sciences, and would eventually lead to a Nobel Prize in economics.23
The first technical papers presented on the afternoon of the first day were again focused on the Harvard Computation Laboratory. Richard Bloch described the hardware and operating principles behind Mark I, and Robert Campbell’s discussion focused on Mark II. Grace Hopper was conspicuously absent from the list of speakers. Hopper suggested that she was passed over because the Symposium’s focus was on hardware and engineering, not programming. “There was nothing on software, yet—nothing on programming. You’ll find it was all hardware and engineering, and specific [engineering] problems—not on the programming techniques yet.”24
For the most part, Hopper’s observation was accurate; however, Bloch did manage to inform his audience about some of the Mark I coding practices and techniques developed by the duo. He explained the practice of coding in rudimentary terms, and then highlighted certain changes made in programming and hardware that augmented Mark I’s processing speed. Finally, Bloch described the introduction of branching to the main sequence mechanism. Branching permitted Mark I to make decisions based on intermediate output and to follow any of several calculation paths without the need for human intervention. Branching was the first example of if-then conditional commands, and made Mark I far more flexible and time efficient.25
Aside from Bloch and Leontief , engineering problems dominated the rest of the Symposium, and computer memory was at the forefront of the technical talks. Though a lack of machine memory made Mark I tedious to code and operate, the machine’s slow processing speed allowed both instructions and data to be mechanically fed into the machine in a synchronous fashion. Conversely, electronic machines (such as the ENIAC), which made calculations thousands of times as fast as Mark I or Mark II, could not be coded in such a fashion, and therefore had to be physically reconfigured for each problem. The reconfiguring process was so time consuming that the ENIAC had little utility other than running the same problem hundreds of times, such as in the creation of a specific ballistics table with incremental variable changes.
In order to overcome the ENIAC’s limitations, Presper Eckert, John Mauchly, and John von Neumann concluded that one way to take full advantage of the computational speed of an electronic calculating machine was to hold both the instruction code and data in some form of internal memory. Information would be accessed at electronic speeds, and both data and instructions could be modified as a given problem progressed. The stored-program concept, as it was later named, was first described in the famous unpublished but highly circulated document von Neumann wrote in June 1945 titled “First Draft of a Report on EDVAC.”26
For the attendees of the conference who were contemplating building an electronic computer with a stored-program capability, it was imperative to solve the engineering issues surrounding memory. During the four-day Symposium there were a variety of memory proposals discussed within the corridors of the Computation Laboratory. Harvard’s Benjamin Moore presented on magnetic coated disks and drums, MIT’s Jay Forrester spoke on high-speed electrostatic storage, and T. Kite Sharpless of the University of Pennsylvania lectured on mercury delay lines. Other options discussed included optical and photographic storage techniques, and a variety of high-speed transfer mechanisms between external and internal memory.27
The only apparent tarnish on a perfectly organized and executed Symposium was the noticeable absence of two of the early community’s most notable members. Professor Norbert Wiener, the distinguished mathematician and founder of the field of cybernetics, decided not to attend for moral reasons. Even though Wiener had served during World War I as a human computer for Army Ordnance at Aberdeen Proving Ground and worked on antiaircraft defense systems during World War II, he refused to attend the Symposium on the grounds that the Navy had been a joint sponsor. Six months earlier, The Atlantic published a letter stating Wiener’s new position: he “would no longer cooperate in scientific research that might result in improved guided missiles for bombing or poisoning of defenseless peoples or otherwise do damage in the hands of irresponsible militarists.”28
Another noted mathematician, John von Neumann, had no such reservations, but decided instead to take a well-needed vacation during the Symposium. The famous mathematician was quite familiar with the Harvard Mark I, as stated earlier, having run the crucial Los Alamos nuclear implosion problem on the machine during the war. Furthermore, both von Neumann and Aiken were recently named to an elite Committee on High Speed Calculating Machines established by the National Research Council, and they would see each other during a committee meeting after the Symposium.29
Believing that he and Aiken represented the leadership of the budding computing community, von Neumann unabashedly offered his colleague some words of advice as they worked to construct the boundaries of the new profession:
We are concerned with developing a relatively new subject—at least new as far as its impact on a major sector of the mathematical community is concerned—and we should, therefore, be very jealous of our standards of publications. It is very probable that we have not yet found our proper level, nor the majority of the mathematical talent that will be ultimately interested in this subject (of this, I am quite particularly strongly convinced), nor the literary form and style and vehicles for the expression of our ideas. At this moment, we are going through a period of expansion and the unavoidable accompanying confusion. We should not permit a low literary standard being established in our field, and it seems to me that it would be a serious mistake to go too easy with neophytes.30
Von Neumann went on to recommend that the two of them should work to establish publishing standards and regulatory mechanisms as customary to more established fields in the exact sciences. Since he would not be in attendance, his final recommendation was that “the speakers should be encouraged to submit manuscripts but the full editorial and reviewing authority should be retained by you.”31
AFTER HOURS
For all of Aiken’s attempts to control the agenda of the conference, what he could not direct was what occurred between the Symposium’s formalities. “I don’t think that any [of us] ever stopped talking the whole time,” Hopper recalled, “and everybody stayed up all night talking about things—it was just
a steady run of conversation.” The dearth of computing publications was a poor indication of the amount of knowledge already generated by the evolving community of practice. Instead of being physically located in books and journal articles, knowledge resided in the minds of the pioneers. The symposium’s most unexpected outcome was that the majority of learning occurred in between and after the formalities of the day, aided by the lubricating flow of alcohol.
The informal learning, according to Hopper, diverged considerably from Aiken’s planned program of events. Much of the discussion in the halls and during meals centered on the potential uses for the machine. Aiken’s assertion that computers were primarily scientific instruments was obvious, but the potential was there for so much more, including automated command and control, aeronautical design, medicine, insurance, and a variety of statistical applications in the social sciences. Another concern centered on the need to attract more young people into the field in order to fuel the envisioned growth. This included not only machine designers, but also mathematically trained coders and operators. “The informal conversations are all lost,” Hopper lamented, “but, I dare say, that it gave quite an impetus to the industry because everybody began getting ideas from everybody else; they’d been living in isolation, and this brought them all together.”32
RAMIFICATIONS OF THE SYMPOSIUM
As the letters poured into the Harvard Computation Laboratory in the following months, it became evident that the conference had been an immense success. “I can’t pinpoint any particular thing that occurred as a result of it,” Hopper later recalled, “but I think a great deal that happened in the next few years were a result of it, because we began to get cross fertilization and communication.”33
The most obvious effect was the creation of a forum for dispersed people to come together and exchange ideas concerning the new technology. “People that hadn’t seen each other in years, because one had been at Oak Ridge, and one had been at Harvard, and one had been out on the Coast, got together,” Hopper recalled. For many invitees it was also the first time that they were able to interact with people from abroad, particularly England.34
The need for personal communication was more pressing because of limited publications. Apart from Hopper’s Manual of Operation, her three articles on Mark I, and von Neumann’s unpublished “First Draft of a Report on EDVAC,” there was little else the computing community could turn to in January 1947 other than a number of marginally informative newspaper articles. The scarcity of publication was exacerbated by enforced secrecy surrounding certain projects, such as MIT’s Whirlwind, or fear that disclosure would threaten patent positions, as was the case with Eckert and Mauchly.35
Besides the cross-fertilization of ideas, the Symposium encouraged the development of institutional connections. Business leaders were exposed to university mathematicians and scientists who had a better understanding of how the new technology could be applied. Military and government leaders controlled funding that university scientists and mathematicians needed for the next generation of machines. Business leaders identified potential government and university customers. All groups realized that the field’s future depended on continued cooperation and support.36
For Aiken and the original crew (Hopper, Campbell, and Bloch), the Symposium was a celebration of their hard-earned accomplishments and long hours of labor. Harvard officialdom, fellow computing pioneers, and leaders in government and the military were all in attendance to dedicate the new Harvard Computation Laboratory, to see Mark I and Mark II, and to pay homage, as the Harvard Alumni Bulletin proclaimed, to “what Professor Howard Hathaway Aiken and his staff had accomplished.”37
For Aiken, the event marked the end of his continuous battle for legitimacy at the University. He had brazenly bypassed an unsupportive Harvard administration and turned to industry and the military in order to construct a calculating machine, and that machine had now shown its worth as a scientific and mathematical tool. Rear Admiral Joy’s opening remarks praised Aiken’s drive and perseverance and educated the audience on the Harvard crew’s work in support of the war effort. Richard Babbage spoke of his great-grandfather’s vision of automated calculation, proclaiming that Aiken had fulfilled his ancestor’s dream.
But not all those in attendance were in awe of Aiken and his accomplishments. Aiken and his crew had demonstrated that computers could produce useful solutions on a consistent basis. Reliability was achieved, however, by way of conservative technical choices during hardware design. The unveiled Mark II was a disappointment, for it did not incorporate technologies that other members of the community took for granted as the state of the art. “I think Mark II was neglected because it was built out of relays and its slow speed,” Hopper recalled. “Nobody paid any attention to her because by the time anybody knew anything about her, she was a dead duck, and everybody was going electronic.”38
For a number of the younger attendees, their experience during the war made them look at computing issues from a different technical perspective than Aiken. As Maurice Wilkes recalled, wartime work in radar, ionospheric research, and television made him very comfortable with electronics. “This was the world in which I and others of my generation had grown up, and we saw the possibilities of achieving, with these means, very high speeds with elegant economy of equipment,” he recalled. “Very few people of Aiken’s generation developed green fingers for electronics, but not all were afraid of high speed electronics to the extent that he was.”39
Paul Morton, representing the University of California and its budding computer project, praised the Harvard crew for developing a reliable machine, but rejected Aiken’s technical choices on similar grounds. “I did give him [Aiken] a great deal of respect for the fact that he didn’t make any compromises with reliability,” Morton stated. “He knew everything really worked, and it did work, even though it was brute force and everybody made fun of it.”40
Just as Aiken was criticized for choosing electromechanical relays over electronic elements, with memory, too, he found himself on the wrong side of the technical divide. His choice of latched relays for Mark II and his penchant for drum memory for the proposed Mark III ran counter to the developing consensus around the Williams tube and mercury delay lines. Frederic Williams of the University of Manchester was developing a series of cathode ray tubes that could hold electric charges that represented data. The data could be accessed rapidly and randomly, a significant advantage when working with a machine that operated at electronic speeds. Of the machines that were developed in the 3 years after the Harvard Symposium, the Manchester Mark I, Whirlwind, MANIAC, SEAC, and SWAC utilized the Williams Tube, which was in fact the first example of random access memory (RAM).
The other preferred storage technology, mercury delay lines, came by way of radar technology during the war. Delay lines stored data by converting it from an electric signal into an acoustic signal moving through a mercury-filled tube. Upon reaching the far end of the tube, the signal could be converted back to electric form, or else it could be amplified and bounced back to the beginning of the tube indefinitely until ready for use. Again, the mercury delay line’s chief advantage comes in storing and accessing information at high speeds, though the information must be accessed sequentially. Mercury delay lines would be used successfully in EDVAC, EDSAC, ACE, BINAC, UNIVAC, and RAYDAC.
For the most part, Hopper defended Aiken’s choice of technology, particularly for Mark II. Designing the second machine commenced in the fall of 1944, and the wartime environment hindered the team from exploring more exotic technological options. Hopper was even impressed by the creativeness of the crew despite less then optimal work conditions. “There weren’t any relays that were fast enough to attain the speeds that Mark II needed,” she recalled, “and those fast latch relays and the other fast relays were actually built for Mark II and Mark II only. And much of that had to be invented under pressure.”41 Bloch also backed Aiken’s decisions in the face of later criticism. “Howard Ai
ken accomplished a great deal at a time when the tools were few and far between,” he strongly asserted.42
For other members of the Harvard crew, however, the lack of innovative technical design at Harvard was more systemic. Fred Miller, the electrical engineer who supervised the construction of Mark II, believed that Aiken’s lone wolf attitude interfered with the process of innovation. “I’m afraid that we have to blame him [Aiken] somewhere. He was so independent and the industry grew so rapidly that one man couldn’t do it all himself .”43 According to Miller, the only people who had influence during the design process other than Aiken were Campbell and Hopper. Aiken’s imposed hierarchical structure within the Computation Laboratory, even after he and his staff had become civilians, hindered the exchange of ideas.44
Despite their privileged position within the Harvard hierarchy, Campbell and Bloch concurred that Aiken was not one to seek out advice from others, especially people beyond his inner circle. “Aiken was the type of individual who could devise many of these ideas on his own,” Bloch recalled, “and who also did not frequently bother to take the time to become completely acquainted with everything going on at the time.”45 If he got a concept stuck in his brain, no amount of discussion would change his mind. For instance, Aiken was convinced that the computational speeds generated by electronic machines such as the ENIAC were superfluous. Bloch, Miller, and Campbell recalled his typical argument that speed was ultimately limited by input and output, and therefore what was the use of all of that computational speed if one was limited by exogenous factors. Aiken’s personal defense of relay technology on the grounds that excess speed was superfluous comes in stark contrast to Hopper’s insistence that reliability and expediency were Aiken’s guiding design principles.46
