XVII.

Persistence of Memory

The Computer

 

            Sebastian von Hoerner was a physicist who for many years worked at the National Radio Astronomy Observatory in Green Bank, West Virginia, and who was famous for his pioneering work in telescope design, work that led to completely new ways of thinking about radio telescopes. Sebastian was not only a world-class physicist but great humanist. He had come of age in Germany on the eve of World War II and, after completing his compulsory military service, found himself drafted into Hitler’s army. He lost an eye in the trenches outside Leningrad; he managed to survive the Dresden firebombing; following the War his father was beaten to death in a Soviet concentration camp and his mother released only six years after her incarceration. Despite the setbacks and tragedies, after a ten-year gap Sebastian picked up his education and retained a life-affirming spirit that was unsurpassed in its generosity toward his fellow human beings. He was the most inspiring man I have ever known.

            Sebastian was also a great raconteur. Many were the evenings we gathered around the fireplace at his house on the observatory grounds to listen as he told stories of his youth, the War and the reconstruction of Germany. Around 1946, Sebastian became a graduate student at Göttingen where he worked with the leading German scientists of the day in the group led by Heisenberg. One project was to study the formation of star clusters, a calculation Sebastian performed with the help of a computer that had been built there. This machine--which if memory serves Sebastian referred to as the Mark 1--had several thousand vacuum tubes in it. * Today’s students no longer know what vacuum tubes were, but let’s call them the transistors of that epoch, bulky glass tubes seven or eight centimeters long. The trouble was, vacuum tubes had a lifetime of about a thousand hours, and since there were several thousand tubes in the Mark 1, it meant that one was failing every couple hours. A technician walked around with a tray slung around his neck--much like the tobacco girls in the old movie houses--replacing the tubes as they failed. Once, a tube was failing intermittently, which made it difficult to track down. This went on for a week, making everyone irritable because no one could trust his calculations. Finally the treacherous tube was discovered and the researchers hung it in a noose before all the others in order to teach them a lesson. The output of the Mark 1 was on paper tape. As the computer had no memory to speak of, the output tape was fed back into the input to iterate calculations. Those were the days.

            I recount Sebastian’s story because even twenty-five years ago it struck me as a history entirely parallel to the one then invariably told about the birth of the computer, a birth that took place in the United States, specifically at the University of Pennsylvania, more specifically with the creation of a machine called ENIAC. What is quite extraordinary is that even now--and I mean two months ago in 2002--after the early history of the computer has become widely known, a professor from University of Pennsylvania can appear on the nationally syndicated radio show Been There Done That, and with the connivance of the host repeatedly refer to ENIAC as “the first large-scale electronic computer” and the room where it was built as “the birthplace of the electronic age.” Conspicuously absent from the program were the names of ENIAC’s creators. I can hardly believe this was unintentional, for if some of the earlier controversies involved legitimately independent developments, the history of the digital computer entailed--well, give it your own name.

 

            Before cloak and dagger, background. The internet, for perhaps obvious reasons, is replete with histories of computers, some of which actually contain references, and several good books exist on the subject. Mercifully, therefore, I am allowed to be brief and we blissfully forego the long history of purely mechanical calculators, including those of Pascal, Leibniz and Babbage. But to pause for the slightest moment, contrary to popular belief Blaise Pascal was not the first to construct a mechanical calculator; that honor evidently belongs to astronomer Wilhelm Shickard of Tübingen. Charles Babbage--let’s pause for a second instant--attempted to build a vast “analytical engine” in the middle of the nineteenth century, and although he failed to complete the project due to technological and financial limitations, the philosopher-visionary originated many modern computer concepts that his successors consciously took over, among them programming. No, we plunge directly into the electro-mechanical age. When that age started is not exactly clear, but certainly by the 1930s many scientists were beginning to think about the possibility of electrical or electronic computing machines. One of the most important developments took place between 1939 and 1944 when Howard Aiken at Harvard developed an “automatic sequence controlled calculator,” known best as the Harvard Mark 1. Like modern computers the giant Mark 1--which was over fifteen meters long--boiled down complicated mathematical computations into the basic arithmetical operations of addition, subtraction, multiplication and division. What’s more, the Mark 1 could be actually be programmed, though most of us probably wouldn’t recognize the process, which involved punched paper tape and switches, as “programming.” But because Aiken’s brainchild was largely mechanical, requiring large banks of switches and relays, it has rightly not been considered an electronic computer. It does have an important place in history since for many years the Mark 1 was believed to have been the world’s first program-controlled computer; it remained in operation to the late date of 1959. Most famous to the public of the devices from that era, however, were undoubtedly the Colossus machines, built at Bletchley Park under the direction of M.H.A. Newman and T.H. Flowers, and used by the British to crack the German Enigma code during World War II. The Colossi were electronic, programmable and binary, but they were also single-purpose machines and so they have also not been regarded as true electronic computers.*

            In the decades following the War, it gradually became apparent that much work on computers had been carried out in Germany, work that is still generally unfamiliar to Americans. One pioneer who has received some belated recognition is Konrad Zuse (1910 -? ), an aircraft engineer with no formal training in electronics who quit his job in 1935 and set up a workshop in his parents’ living room to devote himself entirely to the computer. With financial help from parents and relatives and intellectual assistance from friends, Zuse had produced the memory unit of his Z1 by 1936 and the entire machine by 1938. The Z1 as it turned out did not function reliably, but doing his best to ignore WWII, Zuse continued his work and completed a Z3 by 1941. Although Zuse’s machines were electromechanical and utilized sliding levers to perform the calculations, the Z3 was programmable, had a memory (of 176 bytes) and relied on binary arithmetic (an idea Zuse credits to two Frenchmen, Raymond Valtat, who patented a binary calculator as early as 1932, and Louis Couffignal, who in 1938 published an important dissertation on binary mechanical calculators). With these features the Z3 was almost certainly the first programmable, digital computer in the world. Zuse, who was also an exceptionally gifted painter, founded Zuse KG and continued to design higher Z’s well into the 1960s when the company was sold to Siemens. The historic Z3, unfortunately, was destroyed by an Allied air raid in 1944.

            Actually, Zuse might be given a run for his money by the American George Stibitz (1904-?), an engineer at Bell Labs. At about the same time Zuse began his work, Stibitz was inspired not to make a better moustrap but a better gadget for keeping track of the bets placed on a horse race. Basing his design on binary arithmetic, Stibitz constructed his “totalisator,” as betting machines were called, with telephone relays, which naturally assume an off or on position. Thrown together one weekend in the summer of 1937, Stibitz’s calculator seems to have been the first to use binary arithmetic. Soon, his project became official and by early 1940 Stibitz had created the “Complex Number Computer,” designed to perform arithmetic operations on complex numbers. The memory consisted of about four hundred relays and data was entered by teletype. Stibitz’s original paper on the machine does not survive and so some of its features remain murky, but the Complex Computer was publicly unveiled in 1940 at a demonstration in New York city, where it was operated remotely by a teletype in New Hampshire. The birth of the internet. Stibitz’s subsequent machines, some of which remained in service until about 1960, became ever more sophisticated, incorporating such features as automatic error correction, but because the early ones were mechanical, single-purpose devices, they tend to be referred to as calculators, not computers.

 

            To return to the old story, which persists on certain radio shows, the true dawn of modern history took place in 1945 with the completion of ENIAC, the Electronic Numerical Integrator and Computer. ENIAC, built at the Moore School of Electrical Engineering at the University of Pennsylvania, was the invention of physicist John Mauchly (1907-1980) and engineer J. Presper Eckert (1919-1995). ENIAC was a monster: it weighed nearly thirty tons, contained 18,000 vacuum tubes and had the brain power of a digital watch. Its first major calculation concerned a model of the hydrogen bomb that turned out to be incorrect. The machine could not store a program--no software--and had to be reprogrammed for each task, a job that required resetting nearly 5,000 switches. ENIAC wasn’t even as reliable as the Harvard Mark 1--as in Sebastian’s story, technicians were constantly prowling around replacing vacuum tubes--and the inventors themselves acknowledged that with all the errors in the switch settings, the machine only got the right answer about twenty percent of the time. Despite its drawbacks, ENIAC was by far the largest and fastest machine of its day, about one thousand times faster than the Harvard Mark 1. It was also, unlike Colossus for example, a general-purpose machine, and for that reason ENIAC is generally referred to as the “first general-purpose electronic computer,” a formulation Been There Done That was careful to follow on the air if not on the website. For the construction of ENIAC, Mauchly (pronounced “Mawkly”) and Eckert received both the patents and the glory and for decades were known as the fathers of the electronic computer. Shortly after ENIAC was publicly unveiled in 1946, they founded their own company, which became UNIVAC, a division of Sperry Rand.

            What the standard formulation disguises by omission is the extraordinarily bitter controversy that arose over ENIAC’s origin. In most people’s minds the controversy has been laid to rest but the bitterness and misrepresentation still hangs in the air. To follow the affair we must first visit John Vincent Atanasoff (1903-1995), a physicist of Bulgarian descent who taught at Iowa State College, now University, in Ames. Subjected to difficult mathematical calculations in the course of writing his Ph.D thesis, Atanasoff gradually got interested in developing machines that would relieve the tedium. By 1936 he had helped build a mechanical, analog device, then turned his attention to the possibility of electronic calculators. For months Atanasoff was stymied as to how to proceed. Finally in an act typical of scientists who need to clear their minds, JV jumped into his car and took a celebrated drive that ended up 200 miles later across the state line in Illinois at a roadside tavern. While sipping bourbon at the dive, Atanasoff’s thoughts came together: he would build a programmable, fully electronic computer that performed in binary and which would have a memory bank composed of capacitors. The memory units would be periodically “jogged” or regenerated to ensure no loss of information during calculations.

            With a grant from Iowa State, Atanasoff hired a gifted electrical engineering graduate student Clifford Berry and construction of the Atanasoff Berry Computer, the ABC, commenced. It was to be a special purpose machine, for solving systems of simultaneous equations and by 1941 work had progressed enough so that the two men applied for patents. Unfortunately, World War II intervened, Atanasoff dropped development of the computer to take up defense work at the Naval Ordinance Lab in Maryland and the university officials dropped the ball on the patent front. After the War, Atanasoff stayed on with the Naval Ordinance Lab and in 1948, unbeknownst to him, the physics department at Iowa State destroyed his computer--which could not fit through a door--in order to make room for a lab.

            To what extent the ABC actually functioned has been a matter of debate. Detractors claim it was never fully operational. That is true in the sense that there is no report of it ever solving 29 simultaneous equations, its design goal. There are eyewitness reports that it did solve smaller sets of equations* and, as the history of the Bell telephone so lucidly demonstrates, functionality is not a prerequisite for patentability. In any case, ABC’s basic components were in place by 1941.

            And that is when John Mauchly, two years before he began work on ENIAC and before he had done serious work on computers, drove to Iowa to pay Atanasoff a visit. It is a matter of sworn testimony that Mauchly stayed at Atanasoff’s house for four days and five nights, thoroughly examined the machine and the construction manual that Atanasoff had written. Moreover, on Atanasoff’s recommendation, Mauchly received a part-time job at the Naval Ordinance Lab and during 1943 met there repeatedly with Atanasoff to discuss computers. Only in 1944 did he reveal that he and Eckert were building a big machine for the Army, ENIAC.

            Later, Mauchly would dismiss the ABC as “a little piece of junk” and deny that he learned anything of value from Atanasoff. The courts disagreed. Atanasoff, to answer the inevitable question, never sued Mauchly. The ABC’s existence came to light only in the 1960s when Honeywell Corporation, in attempting to break Sperry Rand’s hold on computer patents, heard about the ABC and realized if it could prove that Mauchly and Eckert derived ENIAC’s concepts from existing prior art, Sperry’s patents would be invalidated. In 1973, after six years of litigation, US District Judge Earl Larson handed down a landmark decision in which he said, “Mauchly derived from ABC ‘the invention of the automatic electronic digital computer’ claimed in the ENIAC patent.” The mantle of “father of the electronic computer” shifted from Mauchly and Eckert’s shoulders to Atanasoff’s.

            You can be sure Larson’s decision did not sit well with Mauchly and Eckert. They remained consumately bitter about it until the end of their lives. To a large extent, though, Mauchly was hoisted on his own petard. During the protracted depositions he continually changed his testimony, initially swearing that he had spent no more than “one and one-half hours” with the ABC, that he only saw it in the shadows, that he did not believe the cover was off and that he had spent only “a couple of days” in Ames. As the case wore on, he gradually admitted that he had spent five days in Ames and had examined the computer on at least four of those days, that he had studied the thirty-five page manuscript given to him by Atanasoff and that he had seen the machine operate.

            Despite all this, Mauchly never conceded that he learned anything from Atanasoff. (There is no evidence, by the way, that Mauchly ever told Eckert about his initial visit with Atanasoff.) Mauchly and Eckert would repeatedly point out that ABC’s memory was in capacitors, ENIAC’s in vacuum tubes; ABC was a binary machine, ENIAC a base-ten machine. Not to mention the vast difference in speed. While there can be no argument that the Penn team had built the best machine in existence, as someone once said there are no patents for degrees of perfection. It seems inescapable that Mauchly got his basic ideas on regenerative memory, one of the patent claims, from Atanasoff, as well as the idea of logical switching with vacuum tubes. The Penn website itself concedes “there is actually little doubt that Mauchly was inspired by Atanasoff’s work,” and Arthur Burks, one of the ENIAC design team, went so far as to coauthor a book with his wife arguing that ENIAC derived its basic concepts from the ABC. ABC’s very existence made the claim that ENIAC was the first electronic digital computer untenable. For this reason the standard formulation became “first general-purpose electronic digital computer.”

 

            Nowadays, most computer scientists, I believe, would credit Atanasoff with the invention of the first electronic digital computer, but naysayers survive. Not long ago journalist Scott McCartney published a spectacularly partisan book, Eniac, intending to re-correct the historical record in favor of Mauchly and Eckert. The subtitle, The Triumphs and Tragedies of the World’s First Computer, which ignores the standard formulation, leaves no doubt as to the agenda. If we compare this work to journalist Clark Mollenhoff’s orthogonal and also not unbiased Atanasoff, as well as the Burkses seminal book, we see that the latter present the documents and large swatches of direct testimony. McCartney, on the other hand, mistakes hearsay for evidence and tosses out blanket declarations such as, “There was no way that Mauchley’s visit to Iowa could undo all the work that had become ENIAC.” The tactic simply deprives the reader of the right to make an informed judgment. His book undoubtedly sold better. There is also not a little irony present throughout in the claims of Eckert that the famous mathematician John von Neumann stole his celebrated ideas on computer architecture from him (which may be true), while denying the possibility that the ENIAC team learned anything from Atanasoff. The one point of interest raised by McCartney is that during the war and after Atanasoff kept more abreast about the progress of computers in general and ENIAC’s in particular than he ever let on. “Records show” that Atanasoff (when, McCartney?) even headed a significant, classified computer development project for the Naval Ordinance Lab, which ultimately failed. Although this may explain why Atanasoff never sued Mauchly, McCartney never establishes its relevance to the question of who invented the computer.

            Scheherazade is perceiving the dawn of day and is about to cease her permitted say. In later years, Eckert claimed that if Edison was the inventor of the light bulb, then he and Mauchly invented the electronic computer, which in light of exhibit XII is an interesting statement, but perhaps the more relevant comparison is again with Alexander Graham Bell. Bell received his patent for a basic process, the transmission of voice by an “undulatory current,” and his claim beat off all challenges. By that standard, although Mauchly and Eckert vastly improved the original system, Atanasoff could claim the basic concepts. At the very least, science thrives on the exchange of ideas and not to credit a colleague with his ideas is, as the British might say, the height of bad form.

            The final irony of the whole affair is that in the modern sense none of these early machines were computers. Not one of them was capable of storing a program. Memory banks were solely for the purpose of storing data and all instructions came from an external device. Once a calculation was off and running it was impossible to alter a program’s execution. Human beings stood by on alert. Although Alan Turing outlined the principles of a stored-program machine as early as 1936, the first computer actually constructed with a stored program, which allowed some measure of self control, was not built until 1948. This was the Selective Sequence Electronic Calculator, completed in 1948 at IBM, which seems to have come in six months ahead of the Manchester Small-Scale Experimental Machine, the “Baby.” Those machines, however, take us to another chapter of history and with the sad tale of the computer we pass from the Technology Domain of the Panopticon to slightly softer endeavors.

 

Reference and Notes for Chapter 17

 

            A good, impartial overview of the evolution of computers is Georges Ifrah, The Universal History of Computing (John Wiley: New York, 2001).

            An even more general overview is George Dyson, Darwin Among the Machines, (Perseus: Reading, Mass., 1997)

            Some of the important, early papers on computing are collected in Brian Randell, ed., The Origins of Digital Computers, third edition (Springer-Verlag: Berlin, 1982). These papers are fairly technical.

            An excellent collection of articles, also fairly technical, on early computers, including all those mentioned here, is Raœl Rojas and Ulf Hashagan, editors, The First Computers--History and Architecture (MIT Press: Cambridge, 2000).

            Konrad Zuse describes his work on the Z’s in his autobiography The Computer--My Life (Springer-Verlag: 1993).

            The definitive work on the Atanasoff-ENIAC controversy, with technical details about the ABC’s architecture and the logical path from ABC to ENIAC is Alice Burks and Arthur Burks, The First Electronic Computer, The Atanasoff Story (University of Michigan Press: Ann Arbor, 1988).

           See also Clark Mollenhoff, Atanasoff, Forgotten Father of the Computer (Iowa State University Press: Ames, 1988). Some technical details are to be found in Allan R. Mackintosh, “Dr. Atanasoff’s Computer,” Scientific American, August 1988, 90-96.

            For the other side, see Scott McCartney, Eniac, The Triumphs and Tragedies of the World’s First Computer (Berkley: New York, 2001).

           

p. 1:   The Been There Done That show concerning ENIAC was broadcast on 8 June 2002 and is archived on the WHYY website.

p. 3:     For Shickard, see introduction to Randell and references therein.

pp. 4-5: For more on Aiken, Zuse and Stibitz, see Ifrah and articles in Randell.

          A good, but somewhat technical, article on Zuse’s machines is Raul Rojas,, “The Architecture of Konrad Zuse’ Early Computing Machines,” in Rojas and Hashagan and also online.

pp. 8-9: For more on the ABC replica, see John Gustavson, “Reconstruction of the Atanasoff-Berry Computer” in Rojas and Hashagan.

          Mauchly’s “piece of junk” remark was made to Molenhoff. See p. 220 of that book. The Burkses and Molenhoff includes large chunks of Mauchly’s testimony. For Mauchly and Eckert’s reaction to the verdict see Molenhoff, chapter 21 and McCartney, chapter 8.

          “There was no way...” McCartney, p. 193