It Began with Babbage Read online

  Third, Larson also ruled that the ENIAC patent was invalid because of the amount of time that had lapsed between the publication of its main features and the filing of the patent. The latter was done on June 26, 1946. However, in 1945, a document titled “First Draft of a Report on the EDVAC” was written by John von Neumann (whom we have encountered briefly [see Chapter 4, Section III, and this chapter, Section II], but will soon meet at length). This report, as we will see, will initiate another chapter of our story, and make the whole matter of inventionhood in the realm of computing even murkier. What mattered to Larson, however, is that on June 30, 1945, Herman Goldstine disseminated this report on a machine that was intended to be the ENIAC’s successor. The report was legally deemed a publication and an “enabling disclosure” of the ENIAC almost a year before the ENIAC patent application was filed, which also rendered the patent invalid.

  So, legally, the ENIAC patent (and Eckerd and Mauchly’s claim to that patent) was rendered invalid. Legally, Eckert and Mauchly had “derived” its principles from Atanasoff’s work. Yet, outside the law court, the question of who really invented the automatic electronic computer poses much difficulty. We now know, for instance, that the ABC and the ENIAC became operational after the Colossus. So if we allow that becoming operational is a reasonable criterion of inventionhood, the first automatic electronic computer was neither the ABC nor the ENIAC.

  Then there is the fact that (like the Colossus), the ABC was designed as a special-purpose computer, to solve linear algebraic equations. Atanasoff recognized that it could also be applied to the solution of differential equations using numerical integration, but this possibility was never tested; it remained in the realm of possibility. The ENIAC was also special purpose, but much less so than the ABC. Even though it was envisioned as an electronic differential analyzer, thus intended to do the kinds of computations required to produce ballistic firing tables, its computational capabilities were such that its first use was computations needed for the hydrogen bomb project, which began in Los Alamos during the early 1940s.

  Arthur Burks, one of its lead engineers and, later, one of its historians, had no doubts about the generality of the ENIAC. To him it was “the first electronic, digital, general-purpose scientific computer.”75 Writing in 1981, Burks defined a general-purpose computer as a digital computer that manifests two features. First, it affords arithmetic and memory capabilities that enable numbers to be entered automatically through an input unit and stored in an alterable memory, arithmetic operations to be performed on numbers, and the outcome transmitted to the outside world through an output device. Second, it provides a two-level programming facility. At the lower level are facilities for the execution of common sequences of arithmetic, input and output operations; at the higher level there is a capacity to combine these sequences into larger units—entire programs. Given these two features, according to Burks, one has a general-purpose computer that can solve a variety of problems—differential equations and number theoretical as well as data processing—commonly encountered in science, engineering, and accounting.76 The ENIAC, meeting these criteria, according to Arthur and Alice Burks, was a general-purpose device.

  Computer scientist and historian of computing Brian Randell would have none of it. He charged that the Burks’s definition was “rather vague.” For instance, a machine that did not have a multiplier unit (which the ENIAC did) but could perform multiplication by repeated addition, met their definition. But, should such a computer count as general purpose?77 At the programming level, Randell suggested that any computer claiming to be general purpose must have the “crucial” facility to select among items held in its read/write memory based on previously computed results—in present-entered language, a branching capability. The ENIAC did not meet this condition.78

  The problem is that both Burks and Randell used somewhat arbitrary criteria to define a general-purpose computer. Furthermore, both, writing in 1981, suffered from the pitfalls of present-centered (whiggish) history (see Prologue, Section VII). Their judgment of what was a general-purpose computer was colored by their perspectives circa 1981; they both imposed their latter-day perceptions on earlier situations.

  It is interesting to compare these opinions voiced a quarter century after the near completion of the ENIAC with a contemporary account by Herman and Adele Goldstine. Writing in 1946, they described the ENIAC as a general-purpose electronic computer that, although developed primarily for the purpose of calculating firing tables, could, in fact, produce solutions to a variety of numeric problems.79

  A slightly later commentator, also (as we will see) a major participant in this story, would write about the ENIAC that its tiny memory along with its manual programming feature severely limited its use for many problems.80

  Clearly, even among the protagonists of this early part of our story, the notion of a general-purpose computer had no well-defined or well-accepted features. As I have noted before, the ENIAC can best be described as a computer that had sufficient generality across a range of mathematical (or numeric) problems (see Chapter 6, Section III, and this chapter, Section I).

  Last, we pause on the judgment rendered by Larson in the court case that the ENIAC was “derived” from the ABC and so cannot count as the “first automatic electronic computer.” We must not forget that the ABC was never a fully, correctly operating machine. On the contrary, it died a quiet death when Atanasoff and Berry both joined war projects. A machine’s claim to priority (not the ideas underpinning the machine) must lie not just in the design, but also in its implementation. It is the operational machine that becomes a functioning artifact, not its underlying principles or its partial operationalism.

  This never happened with the ABC. The ENIAC, on the other hand, was actually operational even before its formal dedication in February 1946. Its first computation was on a problem that pertained to the fledgling hydrogen bomb project in Los Alamos, and this computation was performed in December 1945.81 The machine was running “satisfactorily” before its dedication.82 In November 1946, it was transferred from the Moore School to the Aberdeen Proving Ground, although it was not started up until February 1947. Thereafter, it operated continuously “until 11.45 pm on 2 October 1955.”83 The contrast between the fate of the ABC and that of the ENIAC was stark. In speaking of the first operational electronic computer, there was what logicians might call a category mistake in comparing the one with the other.

  NOTES

  1. B. Randell. (1980). The Colossus. In N. Metropolis, J. Howlett, & G.- C. Rota. (Eds.), A history of computing in the twentieth century (pp. 47–92). New York: Academic Press (see especially p. 74).

  2. A. W. Burks. (1980). From ENIAC to the stored program computer: Two revolutions in computers. In Metropolis, Howlett, & Rota (pp. 311–344).

  3. A. W. Burks & A. R. Burks. (1981). The ENIAC: First general-purpose electronic computer. Annals of the History of Computing, 3, 310–399 (see especially p. 311).

  4. H. H. Goldstine. (1972). The computer from Pascal to von Neumann (p. 156). Princeton, NJ: Princeton University Press.

  5. R. Moreau. (1984). The computer comes of age (p. 33). Cambridge, MA: MIT Press.

  6. Randell, op cit., pp. 74–75.

  7. Goldstine, op cit., p. 153.

  8. Burks & Burks, op cit., p. 337.

  9. Ibid., p. 311.

  10. A. W. Burks. (1947). Electronic Computing Circuits for the ENIAC. Proceedings of the Institute of Radio Engineers, 35, 756–767.

  11. Ibid.

  12. Ibid., p. 767.

  13. Ontogeny is “the life history of an individual, both embryonic and postnatal.” S. J. Gould. (1977). Ontogeny and phylogeny (p. 483). Cambridge, MA: Belknap Press of Harvard University Press.

  14. Goldstine, op cit., p. 128.

  15. Ibid., pp. 131–133.

  16. Burks & Burks, op cit., p. 311.

  17. W. Thomson. (1878). Harmonic analyzer. Proceedings of the Royal Society of London, 27, 371–373.

 
18. G. P. Zachary. (1977). Endless frontier: Vannevar Bush, engineer of the American century (p. 49). New York: Free Press.

  19. V. Bush. (1931). The differential analyzer, a new machine for solving differential equations. Journal of the Franklin Institute, 212, 447–488.

  20. Burks & Burks, op cit., p. 314.

  21. Zachary, op cit., p. 51.

  22. Ibid.

  23. Burks & Burks, op cit., p. 314.

  24. Zachary, op cit., p. 73.

  25. Goldstine, op cit., p. 130.

  26. Ibid., p. 133.

  27. Burks, 1980, op cit., p. 314.

  28. Ibid.

  29. Goldstine, op cit., p. 149.

  30. Burks, 1980, op cit., p. 314.

  31. J. V. Atanasoff. (1940). Computing machine for the solution of large systems of linear algebraic equations. Unpublished memorandum. Printed in B. Randell. (Ed.). (1975). The origins of digital computers (2nd ed., pp. 305–325). New York: Springer-Verlag (see especially p. 305).

  32. Ibid., p. 306.

  33. Ibid.

  34. J. V. Atanasoff. (1984). Advent of electronic digital computing. Annals of the History of Computing, 6, 229–282.

  35. Atanasoff, 1940, op cit., pp. 307–308.

  36. Ibid., p. 309.

  37. Atanasoff, 1984, op cit., p. 242; Burks & Burks, op cit., p. 317.

  38. Burks & Burks, op cit., p. 329.

  39. Ibid.

  40. Ibid., p. 330.

  41. Atanasoff, 1984, op cit., p. 255.

  42. C. R. Mollenhoff. (1988). Atanasoff: Forgotten father of the computer (p. 255). Ames, IA: Iowa State University Press.

  43. Atanasoff, 1984, op cit., pp. 254–255; Mollenhoff, op cit., pp. 55–58.

  44. Mollenhoff, op cit., p. 57.

  45. Quoted by Burks & Burks, op cit., p. 332.

  46. Ibid.

  47. J. V. Mauchly. (1942). The use of high speed vacuum tube devices for calculating. Unpublished memorandum. Printed in Randell (pp. 329–332), 1975, op cit.

  48. Goldstine, op cit., pp. 155–156.

  49. Ibid., p. 154.

  50. Ibid., p. 155.

  51. R. K. Richards. (1955). Arithmetic operations in digital computers (p. 98). Princeton, NJ: Van Nostrand.

  52. Burks, 1947, op cit., p. 760.

  53. Quoted from J. P. Eckert, J. W. Mauchly, H. H. Goldstine, & J. G. Brainerd. (1945). Description of the ENIAC and comments on electronic digital computing machines. Contract W670 ORD 4926. Philadelphia, PA: Moore School of Electrical Engineering. Oxford English Dictionary [On-line]. Available: http://oed.com.

  54. Goldstine, op cit., p. 241.

  55. Oxford English Dictionary, op cit.

  56. This example is taken from Goldstine, op cit., p. 160.

  57. P. B. Medawar & J. S. Medawar. (1983). Aristotle to zoo: A philosophical dictionary of biology (pp. 225–226). Cambridge, MA: Harvard University Press. For an authoritative text on this law see Gould, op cit.

  58. J. P. Steadman. (1979). The evolution of designs. Cambridge, UK: Cambridge University Press; G. Basalla. (1988). The evolution of technology. Cambridge, UK: Cambridge University Press; H. Petroski. (1988). The evolution of useful things. New York: Alfred A. Knopf; S. Dasgupta. (1996). Technology and creativity (Chapter 8). New York: Oxford University Press.

  59. Dasgupta, op cit., p. 146.

  60. This diagram is taken from Dasgupta, op cit., p. 147.

  61. Mollenhoff, op cit., p. 59.

  62. Mauchly, 1942, op cit., p. 329.

  63. Burks & Burks, op cit., pp. 334–335.

  64. Ibid.

  65. Ibid., p. 341.

  66. Ibid., p. 364.

  67. Ibid., p. 363.

  68. Ibid.

  69. Ibid., pp. 371–372.

  70. N. Stern. (1980). John William Mauchly: 1907–1980 (obituary). Annals of the History of Computing, 2, 100–103.

  71. Goldstine, op cit., p. 326.

  72. The records of the trial documented as The ENIAC Trial Records, U.S. District Court, District of Minnesota, Fourth Division: Honeywell, Inc. v. Sperry Rand Corp. et al, No. 4-67, Civ. 138. Decided October 19, 1973: Judge Earl Larson. Judge Larson’s decision was published in the U.S. Patent Quarterly, 180, 673–773. The court records are available at the Charles Babbage Institute of the History of Computing, Minneapolis, Minnesota, and also online. Available: www.cbi.umn.edu

  73. Burks & Burks, op cit., p. 312.

  74. Larson, as cited in the records of the ENIAC Trial, www.cbi.umn.edu, op cit.

  75. Burks & Burks, op cit. (pp. 311–312).

  76. Ibid., p. 385.

  77. Randell, 1980, op cit., pp. 74–75; see Comment by B. Randell in Burks & Burks, op cit., p. 397.

  78. Burks & Burks, Comment by B. Randell, op cit., pp. 396–397.

  79. H. Goldstine & A. Goldstine. (1946). The Electronic Numerical Integrator and Computer (ENIAC). Mathematical Tables and Other Aids to Computation, 2, 97–110. Reprinted in Randell, 1975, op cit. (pp. 333–347). (See especially p. 333.)

  80. D. J. Wheeler. (1951). Automatic computing with the EDSAC (p. 5). PhD dissertation, University of Cambridge.

  81. Goldstine, op cit., p. 226.

  82. Ibid., p. 231.

  83. Ibid., p. 235.

  8

  A Paradigm Is Born

  I

  IN THE ENIAC story so far, John von Neumann has had a fleeting presence. We saw that the BRL formed a high-powered scientific advisory committee at the start of World War II, well before the United States entered the war. von Neumann was a member of this committee and it is unlikely that anyone in the committee was as influential in the American scientific world or, for that matter, in the corridors of power in Washington, DC, than him.

  By the beginning of the 1940s, von Neumann had a massive reputation in the mathematical universe. His contributions spanned many regions of pure and applied mathematics, mathematical physics, even formal logic. He was one of the six mathematicians originally appointed as professors at the Institute of Advanced Study, Princeton, when it was founded in 19331—another was Einstein. In 1944, von Neumann and economist Oskar Morgenstern (1902–1977) published a book titled The Theory of Games and Economic Behavior, thus founding and establishing for posterity the scientific discipline known as game theory.

  Herman Goldstine, who came to know von Neumann very well—first through their involvement with the BRL and then, after the war, at the Institute of Advanced Study, where Goldstine went to work with von Neumann on what came to be called the IAS computer project2—wrote vividly about von Neumann’s intellectual persona, of his ever-ready receptiveness to new ideas, his responsiveness to new intellectual challenges, his mental restlessness when between projects, and the single-mindedness with which he pursued an idea that captured his attention.3

  Oddly enough, despite his involvement with the BRL, he was apparently unaware of the ENIAC project until a chance meeting with Goldstine in a railway station in Aberdeen, Maryland. Goldstine recalls how the entire tone and tenor of their first conversation, initially casual and relaxed, changed when von Neumann realized that Goldstine was involved with the development of a high-speed electronic computer. Thereafter, Goldstine writes, he felt as he was being grilled in a doctoral oral examination.4 Thus began their association, a relationship that only ended with von Neumann’s death from cancer in 1957.5 And thus began von Neumann’s engagement with computers.

  Soon after this meeting in August 1944, von Neumann accompanied Goldstine to Philadelphia to witness the ENIAC.6 From that point on, von Neumann was a regular visitor to the Moore School, attending meetings with the ENIAC designers.7

  However, von Neumann’s interest in computers did not originate only out of intellectual curiosity. Like every one else, his engagement with automatic computing stemmed from dissatisfaction with the status quo. Working in the field of applied mathematics called hydrodynamics (or fluid dynamics), he had pondered on the use of computational approaches for solving analytical equations.8 Beginning in 1943, von Neuma
nn was also associated with the Los Alamos Scientific Laboratory (later renamed the Los Alamos National Laboratory), where the Manhattan Project was underway to develop the atom bomb and, in this context, concerned with solutions to nonlinear systems of equations, he was in search of methods for expediting solutions to such problems.9

  II

  Even as the ENIAC was being implemented, its designers were recognizing its weaknesses. Its memory, constructed of vacuum tubes, was far too small in capacity for handling many large problems, but larger memories using tubes were deemed unrealizable. A different electronic technology was needed for the memory. Besides, the use of plugboards, cables, and manual switches to program and set up the machine for each fresh task was too cumbersome, too slow. Goldstine would recall that, by August 1944, he was chafing at how clumsy the mechanism was for programming the ENIAC.10 Clearly, despite what he and Adele Goldstine would write in their report on the ENIAC in 1946 (see Chapter 7, Section IX), well before that report was written, the ENIAC was not deemed general-purpose enough.

  The very act of implementation revealed not only that the designas-theory was inadequate in satisfying the ENIAC’s intended purpose, but also that the purpose itself needed to be extended. Perhaps further thought during the design and planning process revealed these shortcomings. But, human beings—even the most intellectually brilliant and the most creative of them—are ultimately limited in their cognitive capacities. People suffer from what polymath scientist Herbert Simon, conceiver of the idea of the sciences of the artificial (see Prologue, Section III), famously termed bounded rationality.11 This is why the empirical approach is so important in the sciences—natural or artificial. The act of experimentation (in the case of the artificial sciences, this entails implementation) is a way of circumventing bounded rationality. The ENIAC’s implementation yielded information about the ENIAC that thinking alone about the design did not.