Just as efforts in Britain concentrated on increased computational power for the war effort, so too did Turing’s equivalent in Germany. Konrad Zuse had also built a computational machine utilising relays and vacuum tubes. Unlike Turing’s efforts, Zuse’s came to naught as a result of Allied bombing in 1944. But the wartime efforts in Britain were mirrored and eventually surpassed in America. One reason might have been Winston Churchill’s mistaken view on the need for secrecy after the war, which led to much of Bletchley Park’s code-breaking equipment being, as one novelist later put it, broken up into pieces “no bigger than the size of a man’s hand”.⁷⁹ Progress in Britain was retarded as a result, but in America the scientific community had been accelerating efforts in the field of computation even before Churchill’s mistaken policy. No doubt the diversion of two Colossus machines by the secret service in partial repayment for US armaments during the war also helped. The use of punched cards and the organisation of computational astronomical work by Comrie in Britain had been observed by the eminent Harvard astronomer Professor H. Brown and later by Wallace Eckert. Eckert was to join the staff at Columbia University where a strong relationship already existed with IBM through donations from Thomas Watson. This relationship would allow Columbia to develop its computing expertise in conjunction with IBM after the war, but during the immediate pre-war period the main thrust of the company’s academic efforts focused on Harvard and the construction of what became known as Harvard Mark I. Also known as the IBM Automatic Sequence Controlled Calculator, the machine began life in 1936 and was finally completed in 1943.
The team at Harvard was headed by Howard Aiken, who had gravitated towards numerical machines as an alternative to the precision-engineered but limited analogue incumbents. In his quest to solve these problems, Aiken came across the work of Babbage and determined to follow it to its logical conclusion. He was able to persuade IBM to adopt the project with initial funding of $15,000 ($330,000), soon raised to $100,000 ($2m). Just as importantly, IBM assigned experienced senior research staff to the project, including, as head, Clair Lake who had been responsible for IBM’s first printing tabulator in 1919. The ASSC or Harvard I was completed in 1943 and announced with a fanfare on 7 August 1944. The size of the machine and its automatic nature captured the public imagination despite the fact that in terms of operational speed it represented an incremental advance rather than a quantum leap. This was a consequence of the components used in its construction, most notably the use of electrical relays. The quantum leap was to come from the abandonment of slow mechanical relays and their replacement with vacuum tubes.
The other item of note is that Aiken studiously ignored the massive contribution made by IBM, not just in terms of funding, but more importantly their scientists who had brought the initial concept to a working physical machine. Apparently this stemmed from a disagreement between Watson and Aiken over the ‘look’ of the machine, with Watson desiring something more marketing friendly than an untidy electrical engine. Watson won in the sense that a shiny cover was put on the machine, but in turn Aiken refused to attribute due credit to IBM’s contribution. In a technological sense the Harvard Mark I did represent a milestone, but more as a symbol of the ending of one period than the beginning of a new one.
Next stop the vacuum tube
The beginning of the new period was marked by the transition to the use of vacuum tubes rather than relays and ran contemporaneously with the development of Harvard Mark I. It was a route being pursued simultaneously by a number of different groups. In 1937 in Iowa a mathematician named John Atanasoff, assisted by one of his graduate students named Clifford Berry, began work on constructing a new computer. Atanasoff drew on his experience in building radios to make use of vacuum tubes as components to allow electronic rather than mechanical calculation. He also used his experience of number systems to arrive at the conclusion that only a binary approach would work. Putting these two streams of thought together, he and Berry built a small prototype which was completed in early 1939. The pressures of the American war effort caused the work to be abandoned in 1942, when Atanasoff took up a senior position with the US Naval Ordnance Laboratory in Washington.
The uncompleted Atanasoff Berry Computer (ABC) had drawn some attention from the scientific community, in particular from John Mauchly, then a member of the faculty of the Moore School of Engineering at the University of Pennsylvania. The Moore School worked very closely with the US military in the mathematically intensive field of ballistics. Mauchly had become acquainted with Atanasoff through a conversation following a lecture Mauchly gave to a meeting of the American Association for the Advancement of Science, where the topic quickly became their mutual interest in the use of vacuum tubes in the construction of computers. Mauchly subsequently travelled to Iowa to meet with Atanasoff and discuss the properties of his ABC machine. This meeting, and the correspondence it produced, many years later formed the basis of a lawsuit about important patents in the development of computers.
Mauchly’s scientific interest had begun with the study of molecular action and like many scientists of the day he was frustrated by the lack of available computational power. He sought to address this by building prototype computational machines using neon and vacuum tubes and binary counters. He subsequently switched to the study of weather patterns, but found the limited research budget at Ursinus College restricted his study. In 1941, he left to join the Moore School. By this time President Franklin D. Roosevelt had declared a state of national emergency and a National Defense Research Committee (NDRC) had been formed to oversee the efforts of the academic and business communities. One of the most pressing mathematical problems of the time was the need to improve the computation of ballistics data. This was to involve cooperative efforts between the Aberdeen Ballistics Research Laboratory, MIT and the Moore School. At the time ballistics data was calculated using an analogue differential analyser, which used numerical integration to calculate the ballistic tables which gave gunners necessary information on trajectory etc., depending upon the firing conditions.
As the demands for increased accuracy and the volume of information grew, so the limitations of existing methods became more binding. The analogue machine was ultimately limited by its engineering accuracy, and the process remained a people-intensive one requiring scores of ‘human computers’. There was a crying need for a machine that could automate production of this information. The environment was therefore receptive to Mauchly’s suggested new computing machine. At the Moore School, Mauchly had teamed up with a young faculty member named J. Presper Eckert to discuss his proposals. The two scientists made a good team, and Eckert brought with him his knowledge of logic and mathematics. After much discussion, Mauchly made a proposal in a memo entitled ‘The Use of High-Speed Vacuum Tube Devices for Calculating’. Although the memo was lost for a period, it formed the basis of a new project sanctioned by the US Army in 1943. The project was to build an all-purpose electronic computer.
While the army was receptive, the same could not be said of other parties. The Massachusetts Institute of Technology, where specialisation had tended to be on analogue systems, voiced outright opposition to the project. At the NDRC, opposition was encountered for much the same reasons. The NDRC was headed by Vannevar Bush, an eminent analogue scientist, and included on its committee two members of the MIT faculty and George Stibitz of Bell Laboratories, who had been working on electromechanical devices. The project was approved by the army in May 1943 with initial funding of $67,500 ($1m). The project name for the machine to be created was ENIAC, short for the Electronic Numerical Integrator and Computer. The construction of ENIAC was a monumental task, involving a large team of dedicated scientists operating under the direction of Eckert. The machine was built from a series of interacting modules dedicated to achieving specific tasks. It was vital that there was a set of common standards to which all groups worked. By May 1944 the machine was operating and able to solve second order differential equations.⁸⁰
ENIAC and EDVAC
The process of building ENIAC, however, had not been an altogether smooth one. Given the egos involved, and the intense time pressure under which the ENIAC project was being run, it is not surprising that personality conflicts soon developed. ENIAC was a great success. The machine was not without design flaws, most notably the fact that it could not store programs. This meant instructions had to be re-input each time a task was set up. Mauchly and Eckert took a pragmatic approach, deciding that solving this flaw should wait for a new project rather than slow down the completion of ENIAC. They also avoided the trap into which Babbage had fallen, recognising that continual delay and cost overruns quickly tire the patience of the funding body.
Support was also forthcoming from another direction. John von Neumann, a brilliant mathematician from the Institute of Advanced Study at Princeton, had been working for some time on the top-secret Manhattan Project at Los Alamos. Since late 1943, he had been engaged in trying to solve the mathematics of implosion in order to develop a controlled detonation of an atomic bomb. Resolving the problem required the solution of an enormous system of partial differential equations, something that if it was to be conducted by conventional means would require huge resources of time and manpower. When Herman Goldstine of the ENIAC engaged Von Neumann in conversation about the project at the Moore School, he was unaware of the Manhattan Project and Von Neumann was unaware of ENIAC. Given the task in which he was engaged, Von Neumann was enthralled by the specific purposes to which the new machine could be put and the potential for further general applications. He was to become involved in the proposal for a new version of the machine that overcame its initial shortcomings. In the meantime, the pressing needs of the Los Alamos Project led to ENIAC being used to assist with the determination of the feasibility of the atomic bomb trigger. Although it was a cumbersome process, inhibited by the shortcomings of the machine and the need for over one million IBM cards,⁸¹ ENIAC provided invaluable assistance.
Von Neumann’s involvement was important in a number of regards. As a highly respected figure in the academic community, his seal of approval for ENIAC carried a great deal of weight and helped win over the sceptics. Secondly, in developing a new upgraded version, Mauchly and Eckert were able to draw on Von Neumann’s logical and mathematical prowess. Against that, the collaboration with Von Neumann also introduced further tension to the group. The tension was raised in June 1945 when Von Neumann circulated a paper under his own name on the proposed successor to ENIAC, to be known as the Electronic Discrete Variable Automatic Calculator, or EDVAC. Although ENIAC had ultimately cost nearly $500,000 ($7m) to develop, unlike Babbage who a hundred years before had lost the confidence of the Duke of Wellington by attempting to move to a new machine before the old one was even working, the scientists at the Moore School were able to retain the confidence of their sponsoring body, the US Army. As a consequence, in late 1944 a further grant was issued by the army of $105,600 ($1.5m) to build the new EDVAC machine.
8.4 (a) and (b) – ‘A marvel of electric ingenuity’: the unveiling of the ENIAC computer
Source: New York Times, 15 February 1946.
For Mauchly and Eckert, the public unveiling of ENIAC in February 1946 should have represented a triumph, but in reality, as supportive as the public reception was, it had already been pre-empted in the scientific community by Von Neumann’s EDVAC paper of the previous year. The whole question of precedent and the patenting of the work involved in ENIAC had also become an emotive subject. Mauchly and Eckert wished the work to be patented, as they believed it had commercial potential that could be exploited when the war ended. As sponsoring body, the army was also concerned at the lack of patent protection and the dangers of others working in the field pre-empting them with patents. A deal was reached whereby the patent would be taken by the inventors and the rights shared between them, the government and the university. Within the Moore School, even those who did not object to a patent application showed little interest in the application. This was only to change later as the commercial potential of the computer became more obvious. At the time the future uses of the technology were not clearly seen by scientists or businessmen.
Eventually, in 1947, a patent application relating to ENIAC was submitted. The delay stemmed from several things: desire to include as much in the application as possible, a wish to give the patent the longest possible period to expiry and also the relative inexperience of the parties involved. All of these factors later came back to haunt Mauchly and Eckert. One of the key issues in the commercial development of any technology is the creation of a barrier to entry for competitors. Rarely is the scientific lead of sufficient magnitude to represent a barrier in itself. In the absence of patent protection, the only barriers to entry are economic ones, associated with control of distribution, or economies of scale in production, and/or the protection offered by government on ‘national interest’ grounds.
So far as the EDVAC was concerned, while the application for funding had been successful, the team at the Moore School was in the process of disintegrating. The display of ENIAC had been followed in 1946 by a series of lectures at the university where its properties were described to a large band of visiting scientists. By the end of the course, information on EDVAC was also supplied to several computer luminaries. This spurred development in Britain by scientists who had been involved in the computer code-breaking activities at Bletchley Park, and resulted in the construction of a working digital computer with stored programs at Cambridge University, which was named the Electronic Delay Storage Automatic Computer (EDSAC). Britain and America were the only two countries actively involved at the forefront of computer research immediately after the war.
For Mauchly and Eckert, their days at the Moore School were effectively ended by the question of patents. A new policy on patents was instituted by the recently appointed head of research, Irven Travis. This placed extremely restrictive covenants on academics and cut across the previous agreement on ENIAC and future EDVAC patents. The contracts offered to Mauchly and Eckert would have involved them renouncing their patent claims. This they deemed unacceptable. Given no other choice, they left the Moore School to seek alternative funding. As the two walked out the door, so did the university’s pre-eminence in the field of computing.
The patent issue did not finish there, however. The belief held by Mauchly and Eckert that they had an agreement regarding the EDVAC patents was shattered when Von Neumann tried to file patents based on the paper he had published under his own name. As the body sponsoring the research, the US Army tried to resolve the increasingly acrimonious dispute between the former three colleagues. The eventual resolution was unsatisfactory in as much as it proved impossible to reach a compromise on EDVAC. It was concluded that the paper circulated by Von Neumann represented a publication and since more than 12 months had elapsed since its distribution the information was technically in the public domain and therefore not eligible for patent protection.
Up against the funding wall
The two ex-Moore scientists had offers of employment, including one from IBM, but concern over operational independence pushed them into forming their own operation. Although companies had shown an interest in the two scientists, there was little recognition of the potential market for computers. Thomas Watson allegedly believed in the late 1940s that the total marketplace for computers extended to not much more than 12 machines. The consensus of the time was still that computers were essentially large, expensive specialised machines with little practical application outside a limited sphere dominated by government uses. Having decided to set up business on their own, Mauchly and Eckert established a partnership that soon became a limited liability company. Their first priority was to find capital to finance the building of a computer with similar properties to that set out in the EDVAC proposal. The new version was named the Universal Automatic Computer, or UNIVAC. The founders had to persuade prospective providers of funds that their technolog
y was sound, deliverable and had a ready customer base. They also had to persuade potential clients of their financial solvency before any would be willing to commit resources and tie themselves to such a new project.
This was a classic chicken-and-egg situation: customers wanted to see committed financial resources and the providers of financial resources wanted to see committed customers. Either the providers of capital or the customers had to be willing to take on trust the assurance of the founders that the other would be forthcoming. The process tends to be much smoother if the principals have a prior record of commercial success. If the principals have spent their lives in academic or government work, persuasion can be difficult. Mauchly and Eckert struggled to raise sufficient funds to capitalise their new venture. Initial funding of $20,000 ($215,000) came from Eckert’s father, a further $200,000 ($2.5m) was raised from other contacts in the Philadelphia area, and in March 1946 the Electronic Control Company (ECC) was formed. Following the example of Hollerith 50 years before, the two founders approached one of the largest peacetime users of computational equipment, the Census Bureau. The bureau was prohibited by federal law from funding research directly and decided to use the National Bureau of Standards (NBS) as a conduit through which to fund the UNIVAC machine.
Engines That Move Markets (2nd Ed) Page 41