Turing's Cathedral

by George Dyson

  Left to right: Norman Phillips (meteorology), Herman Goldstine (associate director), and Gerald Estrin (engineering), in the MANIAC machine room, 1952. The theoreticians at the Institute for Advanced Study had mixed feelings about the influx of meteorologists and engineers. As Julian Bigelow describes it, those “who had to think about what they were trying to do” did not welcome the arrival of those “who seemed to know what they were trying to do.” (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  The RCA Selectron, or Selective Storage Electrostatic Memory Tube, invented by Jan Rajchman, promised an all-digital 4,096-bit electrostatic storage matrix in a single vacuum tube. The applications included numerical weather prediction, as portrayed in this advertisement in the February 1950 National Geographic, as well as file storage and retrieval at lightning speeds. (RCA/National Geographic)

  Tom Kilburn (left) and Frederic C. Williams (right) at the controls of the Small-Scale Experimental Machine (SSEM) at the University of Manchester, 1948. The first working stored-program electronic digital computer, the Manchester “Baby” ran a 17-line program (a search for Mersenne primes) as a test of its 1,024-bit cathode-ray tube memory on June 21, 1948. (Department of Computer Science, University of Manchester)

  James Pomerene with the Williams electrostatic storage tube. When RCA’s Selectron failed to materialize on schedule, the IAS team, led by James Pomerene, adapted over-the-counter 5-inch cathode-ray oscilloscope tubes into a fully random-access memory based on the Williams-Kilburn ideas. The obstacle to achieving high-speed digital storage was less a memory problem than a switching problem, solved by using the two-axis analog deflection of an electron beam as a 1,024-position switch. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Williams memory tube, exploded schematic view, showing electromagnetic shielding, connection for deflection circuits, and high-gain amplifier integrated into the face of the individual tube. When the electron beam is aimed at one of the 1,024 locations on the inside surface of the tube, and given a “twitch,” a faint electrical signal is produced in the wire screen attached to the outside face of the tube. Amplified thirty thousand times, the character of that signal is “discriminated” to indicate whether the state of charge at that location represents a zero or a one. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Distinction between a dot (0) and a dash (1) has to be determined in 0.7 microseconds, by “inspecting” the character of the faint secondary pulse generated when a given location is “interrogated” by the electron beam. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Adder side of IAS computer, schematic view. Above the Williams memory tubes (labeled from 2–1 on the right to 2–39 on the far left) are the memory registers, the adders, and the digit resolvers (known as “digit dissolvers” when malfunctioning). The opposite side of the machine is similar, with memory tubes 20 through 2–38, and address and instruction registers on the lower level, arithmetic and memory registers above. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  The MANIAC in 1952, arranged like a V-40 engine with overhead valves, was about 8 feet in length, 6 feet high, 2 feet in width; consumed about 19.5 kilowatts of electricity; and ran at about 16 kilocycles at full speed. The Lucite covers over the registers improve the flow of air being exhausted at a rate of 1,800 cubic feet per minute through the overhead ducts. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Julian Bigelow, Herman Goldstine, J. Robert Oppenheimer, and John von Neumann, at the public dedication of the IAS computer, June 10, 1952. “Oppenheimer was never against the machine, and had his picture taken in front of it a few times, but that was his major contribution,” says Bigelow. “I really don’t ever remember seeing him there,” adds Willis Ware. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  IAS engineering team, 1952. Left to right: Gordon Kent, Ephraim Frei, Gerald Estrin, Lewis Strauss, J. Robert Oppenheimer, Richard Melville, Julian Bigelow, Norman Emslie, James Pomerene, Hewitt Crane, John von Neumann, and Herman Goldstine (outside the frame). (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Electronic Computer Project staff, 1952. Known identifications (left to right). Sitting:?, Lambert Rockefellow,?,?, Elizabeth Wooden, Hedvig Selberg (kneeling), Norma Gilbarg,? Standing, middle: Frank Fell,?,?,?, Hewitt Crane, Richard Melville,?, Ephraim Frei, Peter Panagos, Margaret Lambe. Standing, far back:?, Norman Phillips, Gordon Kent,?, Herman Goldstine, James Pomerene, Julian Bigelow, Gerald Estrin,? (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  IAS housing project, 1950. Eleven war surplus wood-frame buildings, purchased at auction in Mineville, upstate New York, were disassembled, transported to Princeton by rail, and reassembled under Julian Bigelow’s supervision in 1946, over the objections of nearby residents to the “deleterious effects upon the fashionable housing area which it will invade.” (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  High-speed wire drive, 1946. Before magnetic tape became available, steel recording wire offered the most immediate path to high-speed input/output, provided that a way could be found to run it at much higher speeds than those used by the audio recording equipment of the time. “Two ordinary bicycle wheels were used for this purpose,” Bigelow reported, “having grooves about ½ inch deep and 1½ inch wide turned in their wooden rims.” (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Oscillogram of 40-bit word produced directly from magnetic recording wire, 1947. The transition from analog to digital was under way. Speeds of up to 100 feet (or 90,000 bits) per second were achieved before it was decided to switch to a 40-track magnetic drum. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  “Electron-Tube Binary Cell Stability.” Sketch prepared by Julian Bigelow for the first “Interim Progress Report on the Physical Realization of an Electronic Computing Instrument,” issued on January 1, 1947. Vacuum tubes were analog devices, and persuading them to behave digitally in large numbers was not an easy problem to solve. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Prototype 11-stage shift register, 1947, built using 6J6 double-triode miniature vacuum tubes. A “positive interlock” approach was taken to all transfers of information within the machine. The three rows of “toggles” allowed all bits to be replicated into an intermediary register before the sending register was cleared. Right shift, left shift, or transfer could be completed in 0.6 microseconds. The neon lamps above the top row of toggles displayed the state of each individual bit. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Fabrication of production model shift registers, summer 1948. Components had to be replicated many times, and local students were hired to do much of the work. According to Julian Bigelow, “a lot of our machines were run by high school girls.” (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Assembling 40-stage shift registers, 1948. All vacuum-tube heater and cathode voltages were supplied at chassis level by sandwiched copper sheet conductors, reducing electronic noise and eliminating visible wiring, except for that directly involved with the logical architecture of the machine. The physical layout was three-dimensional, optimizing air cooling and minimizing connection paths for increased speed. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  “Shifting Register No. 7 Functional Diagram,” March 1948. With no precedents to follow, a wide range of possible ways of interconnecting the elements of the computer were explored. “We enjoyed some interesting speculative discussions with von Neumann at this time about information propagation and switching among hypothetical arrays of cells,” remembers Bigelow, “and I believe that some germs of his later cellular automata studies may
have originated here.” (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Left to right: James Pomerene, Julian Bigelow, and Herman Goldstine, inspecting arithmetic unit, 1952. Of the final total of 3,474 vacuum tubes in the computer, 1,979 were the miniature twin-triode 6J6. Self-diagnostic routines were used to identify suspect tubes before they failed. “The entire computer can be viewed as a big tube test rack,” Bigelow observed. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Willis Ware, explaining computer architecture at the RAND Corporation, in Santa Monica, California, April 1962. Ware, fourth to be hired for the IAS computer project, in March 1946, departed for California upon completion of the MANIAC in 1951. (Courtesy RAND Archives)

  Flow diagram, December 9, 1947, for a Monte Carlo problem being coded, with Klári von Neumann’s assistance, as part of a hand-computed rehearsal conducted at Los Alamos in advance of running the problem on the ENIAC, now moved from the Moore School in Philadelphia to the Aberdeen Proving Ground. (Von Neumann Papers, Library of Congress)

  Above: Stan Ulam winning at a T-Division poker game at Los Alamos (date unknown). Clockwise from lower left:?, Carson Mark, Bernd Matthias, Stan Ulam, Foster Evans, George Cowan, Nicholas Metropolis. (Claire and Françoise Ulam; photographer and date unknown)

  Klára (Klári) von Neumann, as pictured on her French driver’s license, issued July 15, 1939. (Marina von Neumann Whitman)

  John von Neumann with Cadillac V-8 coupe, en route to Florida, January 1939. (Marina von Neumann Whitman)

  Klári von Neumann, Florida, January 1939. Johnny and Klári were married in Budapest in November 1938 and, leaving a Europe that she described as “a powder keg with the lighted fuse burning rapidly and dangerously short,” immediately sailed for the United States. After attending a meeting of the American Mathematical Society in Virginia, they drove through the Florida Everglades to Key West. (Marina von Neumann Whitman)

  Stanislaw and Françoise Ulam, 1940s. Polish mathematician Stan Ulam arrived in Princeton at von Neumann’s invitation in December 1935, with a $300 lectureship from the Institute for Advanced Study, enough to get his foot in the door of the United States. He then secured a fellowship at Harvard, retrieved his younger brother, Adam, from Poland, and met Françoise, a graduate student at Holyoke, in the fall of 1939. (Claire and Françoise Ulam)

  Nicholas Metropolis, Los Alamos badge photograph, ca. 1943. Los Alamos needed its own copy of the IAS computer as soon as possible, and von Neumann recommended Nick Metropolis for the job of building it in July 1948. “We seem to be getting slowly organized here for [the] building program,” Metropolis wrote to Klári von Neumann on February 15, 1949. “Even you will like Los Alamos, with a machine to tinker with.” (Los Alamos National Laboratory Archives)

  Paul Stein (left) and Nicholas Metropolis (right) observing a game of “anti-clerical” chess being played (on a 6-by-6 board, without bishops) by the MANIAC-1 at Los Alamos, 1956. Perforated tape input/output is visible at left, and the machine’s modular Williams tube memory is in the racks overhead. (Los Alamos National Laboratory Archives)

  John von Neumann (top, wearing business suit and facing backward) and Klári von Neumann (fourth from bottom), visiting the Grand Canyon, sometime in the late 1940s. (Marina von Neumann Whitman)

  Left to right: Françoise, Claire, and Stanislaw Ulam with John von Neumann, who coined the phrase “Los Ulamos,” in reference to the hospitality of the Ulam household at Los Alamos and, later, Santa Fe. (Stanislaw Ulam papers, American Philosophical Society; courtesy of Françoise Ulam)

  Actors in the hydrogen bomb drama, ca. 1950. Joseph Stalin (with “Made in U.S.S.R.” bomb), J. Robert Oppenheimer (as an angel), Stanislaw Ulam (with spitoon), Edward Teller (center), George Gamow (with cat). (Montage by George Gamow, courtesy of Claire and Françoise Ulam)

  Von Neumann (left) at Redstone Arsenal, 1955, to observe a missile test with (left to right) Brigadier General Holger N. Toftoy (commander of Redstone Arsenal), German American rocket pioneer Werner von Braun, Brigadier General J. P. Daley, and Colonel Miles B. Chatfield. (Von Neumann Papers, Library of Congress)

  Computational grid over Northern Europe, used by Lewis Fry Richardson in his numerical model developed during World War I and published in his Weather Prediction by Numerical Process of 1922. (Lewis Fry Richardson, 1922)

  “Electrical Model illustrating a Mind having a Will but capable of only Two Ideas,” proposed by Lewis Fry Richardson in a 1930 study that raised the possibility, later taken up by Alan Turing, that random electronic indeterminacy could be amplified into creative thinking and even free will. (Lewis Fry Richardson, “The Analogy Between Mental Images and Sparks,” Psychological Review 37, no. 3 [May 1930]: 222)

  The ENIAC meteorological expedition, Aberdeen Proving Ground, March 1950. Left to right: Harry Wexler, John von Neumann, M. H. Frankel, Jerome Namias, John Freeman, Ragnar Fjørtoft, Francis Reichelderfer, Jule Charney. (MIT Museum)

  Five main problems (left) addressed by the IAS Electronic Computer Project, 1946–1958, with time scale in seconds (center) and representative phenomena (right) for comparison. The human attention span falls exactly in the middle of this range of twenty-six orders of magnitude in time. (Courtesy of the author)

  John von Neumann in Florida, January 1939. Fascinated by biology, von Neumann began to formulate a comprehensive Theory of Self-Reproducing Automata general enough to encompass both living organisms and machines. (Marina von Neumann Whitman)

  Nils Aall Barricelli, as pictured on his application to the U.S. Educational Foundation in Norway for a travel grant, under the Fulbright Act, to visit the Institute for Advanced Study, “to perform numerical experiments by the use of large calculating machines, in order to clarify the first stages in the evolution of species,” December 8, 1951. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  General Arithmetic Operating Log, November 23, 1954. After “Barricelli on” at 12:45 a.m., the computer will “not duplicate” the numerical evolution experiment; the log notes “Barricelli off” at 1:58 a.m. Most codes were represented in hexadecimal notation; Barricelli worked directly at the binary level, as evidenced here. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Barricelli’s Universe, 1953. Five out of every one hundred generations of numerical symbioorganisms were sampled and the data transferred to punched cards assembled into an array and contact-printed onto blueprint paper, leaving the imprint visible here. The rules governing this particular universe were the “Blue Modified Norm”—parasites disqualified but mutations allowed. The results favored “smaller numbers and probably more rapid uniformity” than the “Blue Norm” (without mutations), where an “initially large flora of new organisms, later probably one species, expands to the whole gene universe,” Barricelli reported in August 1953. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Alan Turing (standing) with Brian Pollard (left) and Keith Lonsdale (right) seated at the console of the Ferranti Mark 1 computer at the University of Manchester in 1951. The Ferranti Mark 1, with 256 40-bit words (1 kilobyte) of cathode-ray tube memory, and a 16,000-word magnetic drum, was the first commercially available implementation of Turing’s Universal Machine. At Turing’s insistence, a random number generator was included, so that the computer could learn by trial and error or perform a search by means of a random walk. (Department of Computer Science, University of Manchester)

  Left to right: James Pomerene, Julian Bigelow, John von Neumann, and Herman Goldstine, at the Institute for Advanced Study, date unknown. Von Neumann, who succumbed to cancer in 1957, “died so prematurely, seeing the promised land but hardly entering it,” remembered Stan Ulam in 1976. (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)

  Final entry in the MANIAC machine log, 12:00 midnight, July 15, 1958, signed by Julian H. Bigelow (JHB). (Shelby White and Leon Levy Archives
Center, Institute for Advanced Study)

  Relics discovered in the basement of the West Building, Institute for Advanced Study, November 2000. Bottom: Source code for “Barricelli’s Drum Code.” Center: Output card from one of the periodic samplings of “numerical symbioorganisms” as they evolved. Above: Note to Mr. Barricelli, concluding, “There must be something about this code that you haven’t explained yet.” (Shelby White and Leon Levy Archives Center, Institute for Advanced Study)