by George Dyson
The culprit was Army Air Corps lieutenant Philip Duncan Thompson, one of a small group of meteorologists who had been recruited by von Neumann in 1946. “Von Neumann singled out the problem of numerical weather prediction for special attention,” Thompson later explained, “as the most complex, interactive, and highly nonlinear problem that had ever been conceived of—one that would challenge the capabilities of the fastest computing devices for many years.”2
Thompson, who was born in 1922, dates the beginning of his scientific education to the age of four, when his father, a geneticist at the University of Illinois, sent him to post a letter in a mailbox down the street. “It was dark, and the streetlights were just turning on,” he remembers. “I tried to put the letter in the slot, and it wouldn’t go in. I noticed simultaneously that there was a streetlight that was flickering in a very peculiar, rather scary, way.” He ran home and announced that he had been unable to mail the letter “because the streetlight was making funny lights.” His father returned with him to the mailbox, explained that he had been trying to insert the envelope the wrong way, and “pointed out in no uncertain terms that because two unusual events occurred at the same time and at the same place it did not mean that there was any real connection between them.”3
In the spring of 1942, then in his third year at the University of Illinois, Thompson attended a lecture by Carl-Gustaf Rossby, a Swedish-born, Norwegian-trained meteorologist transplanted to the University of Chicago, where he was training prospective weather officers—eventually 1,700 of them—for the war. In May of 1942, as soon as classes were over, Thompson enlisted in the Army Air Corps in order to join Rossby’s group. After completing his training, he was stationed in Newfoundland, monitoring the North Atlantic weather systems that had led the Scandinavians to develop the theory of frontal waves and otherwise lead the way in understanding what the weather might do next. At the end of the war he was assigned to Long Beach Air Force Base in California, as weather officer liaison to Norwegian meteorologist Jacob Bjerknes at UCLA, where he became close friends with Jule Charney, who had just received his PhD.
In 1945 meteorology had become a science, while forecasting remained an art. Forecasts were generated by drawing up weather maps by hand, comparing the results with map libraries of previous weather conditions and then making predictions that relied partly on the assumption that the weather would do whatever it had done previously and partly on the forecaster’s intuitive feel for the situation and ability to guess. On average, forecasts beyond twenty-four hours were still no better than “persistence”—predicting that the weather tomorrow will be the same as it was today.
World War II, with its growing dependence on aircraft, increased the demand for forecasts, while weather radar and radio-equipped weather balloons increased the supply of observational data needed to produce them. Thompson, trained in mathematical physics, was convinced that given accurate knowledge of the current state of the atmosphere and its outside influences, it should be possible to make predictions, based solely on the laws of physics, about its state at some near-future time. He had a single mechanical calculator, his father’s advice not to infer cause from coincidence, and the knowledge that his predecessor, Lewis Fry Richardson, had made a similar attempt, and completely failed.
Lewis Fry Richardson, a Quaker and ardent pacifist who resigned from the British Meteorological Office when it was taken over by the Air Ministry, had begun developing a numerical atmospheric model while serving as superintendent of the Meteorological and Magnetic Observatory at Eskdalemuir in Dumfriesshire, Scotland, in 1913. The observatory, a branch of the National Physical Laboratory, had been moved to Eskdalemuir from Kew, near London, when electric railways came into use. The damp, secluded outpost, deliberately situated as far as possible from any artificial magnetic fields, suited Richardson, who cultivated an “intentionally guided dreaming,” letting his mind balance between almost awake and almost asleep. “It is the ‘almost’ condition that is advantageous for creative thinking,” he explained.4
At the outbreak of World War I, Richardson found himself “torn between an intense curiosity to see war at close quarters [and] an intense objection to killing people.”5 He applied to join what would become the Friends’ Ambulance Unit when it was formed in 1914, and was finally granted leave from the observatory to do so in May 1916. After basic training in how to keep the ambulances running and the wounded alive, he departed for France in September, where he served at the front lines, attached to the Sixteenth Division of the French infantry, until 1919.
The Society of Friends had grown respectable since the time of Charles II and the imprisonment of William Penn. Bound by a humanitarian mission to assist the wounded, combined with Quaker refusal to submit to military authority, the Friends’ Ambulance Unit served with heroic self-discipline during the Great War. Richardson’s convoy, known as Section Sanitaire Anglaise Treize, or S.S.A. 13, reached a full strength of twenty ambulances and forty-five men. Between February 1914 and January 1919 they transported 74,501 patients over 599,410 kilometers of evacuation runs.6
A poor driver but a gifted mechanic, Richardson endeared himself to the rest of the group. “The other day my electric lighting dynamo went wrong,” noted Olaf Stapledon, the future author of Last and First Men, on December 8, 1916. “The mechanic was away, & I know little about electricity, so I was dished. Fortunately we found that our eccentric meteorologist was also an expert electrician. He and I had a morning on the job unscrewing, tinkering, cleaning and generally titivating, sometimes lying under the car in the mud, sometimes strangling ourselves among machinery inside.”7
A year later, Richardson and Stapledon celebrated the fourth Christmas of the war. “The moon is brilliant, and the earth is a snowy brilliance under the moon. Jupiter, who was last night beside the moon, is now left a little way behind. Venus has just sunk ruddy in the West, after being for a long while a dazzling white splendour in the sky,” Stapledon reported on December 26, 1917. “I have just come in from a walk with our Professor, and he has led my staggering mind through mazes and mysteries of the truth about atoms and electrons and about that most elusive of God’s creatures, the ether. And all the while we were creeping across a wide white valley and up a pine clad ridge, and everywhere the snow crystals sparkled under our feet, flashing and vanishing mysteriously like our own fleeting inklings of the truth about electrons. The snow was very dry and powdery under foot, and beneath that soft white blanket was the bumpy frozen mud. The pine trees stood in black ranks watching us from the hill crest, and the faintest of faint breezes whispered among them as we drew near. The old Prof (he is only about thirty-five, and active, but of a senior cast of mind) won’t walk fast, and I was very cold in spite of my sheepskin coat; but after a while I grew so absorbed in his talk that I forgot even my frozen ears.… We crossed the ridge through a narrow cleft and laid bare a whole new land, white as the last, and bleaker. And over the new skyline lay our old haunts and the lines. Sounds of very distant gunfire muttered to us.”8
Richardson kept working on his numerical model, whenever there were moments to spare. “This billet is a barn like the last, but we are far more crowded together,” Stapledon noted on January 12, 1918. “Beside me sits Richardson, the ‘Prof,’ setting out on an evening of mathematical calculations, with his ears blocked with patent sound deadeners.”9 The model’s input was a tabulation of the weather conditions observed over Northern Europe over a six-hour interval between 4:00 a.m. and 10:00 a.m. on May 20, 1910, an “international balloon day,” when detailed records had been collected by Norwegian meteorologist Vilhelm Bjerknes, whose pioneering efforts to quantify our understanding of the atmosphere had inspired Richardson, and whose son Jacob Bjerknes would later be Thompson’s supervisor at UCLA.
“My office was a heap of hay in a cold rest billet,” Richardson reported. “It took me the best part of six weeks to draw up the computing forms and to work out the new distribution in two vertical columns for the first tim
e.”10 The extended computation was well suited to the long, drawn-out war. Surrounded by mud and death and shrapnel, Richardson worked to reconstruct the weather on a spring morning when balloons had drifted over the then-peaceful European countryside, treating the motions of the atmosphere as nature’s solution to a system of differential equations that linked the conditions in adjacent cells from one time step to the next.
Richardson used a method of finite differences that he had developed in 1909. “Both for engineering and for many of the less exact sciences, such as biology, there is a demand for rapid methods, easy to be understood and applicable to unusual equations and irregular bodies,” he had written in a report to the Royal Society in 1909.11 With boundary conditions as poorly defined as they usually were in meteorology, approximate answers were good enough.
The resulting prediction was at odds with what had happened on May 20, 1910, yet Richardson was correct in his belief that calculations would eventually supersede the existing synoptic methods of weather prediction, where “the forecast is based on the supposition that what the atmosphere did then, it will do again, now,” and “the past history of the atmosphere is used, so to speak, as a full-scale working model of its present self.”12 He completed the test forecast, and then, “during the battle of Champagne in April 1917, the working copy was sent to the rear, where it became lost, to be rediscovered some months later under a heap of coal.”13
After the war, Richardson published a detailed report, Weather Prediction by Numerical Process, so that others might learn from his mistakes. At the end of his account, he envisioned partitioning the earth’s surface into 3,200 meteorological cells, relaying current observations by telegraph to the arched galleries and sunken amphitheater of a great hall, where some 64,000 human computers would continuously evaluate the equations governing each cell’s relations with its immediate neighbors, maintaining a numerical model of the atmosphere in real time. “Outside are playing fields, houses, mountains and lakes, for it was thought that those who compute the weather should breathe of it freely,” he imagined, adding that “perhaps some day in the dim future, it will be possible to advance the computations faster than the weather advances, and at a cost less than the saving to mankind due to the information gained.”14
Twenty-six years later, Philip Thompson picked up where Richardson had left off. He remembers: “1946 was a year of ferment, for the formulation of the problem and the means of solving it were at last moving toward each other, albeit not by design.”15 Thompson “ground away at an old Monroe desk calculator, trying to figure out short-cuts and becoming increasingly depressed by the burden of hand calculation,” until, as he describes it, “one fine afternoon in the early autumn of 1946, Prof. Jørgen Holmboe called me in, said he was aware of what I was trying to do, and handed me an article from the New York Times.” The article announced the intention of Vladimir Zworykin of RCA and John von Neumann of the Institute for Advanced Study to collaborate on the construction of a high-speed electronic computer and its application to weather prediction and control. “Next day, I called my commander, Gen. Ben Holzman, and requested authorization to travel to Princeton to meet with von Neumann,” Thompson continues. “General Holzman grumbled a bit, but agreed to it if I traveled as extra crew on a military aircraft that was headed East anyway. The following day the arrangements were clear, and I made my way to Princeton, via B-29, bus, stagecoach, train, oxcart, and the PJ&B.”16 PJ&B was the two-car train, or “Dinky,” that shuttled from Princeton University to the main line at Princeton Junction and back.
Thompson met with von Neumann, “overawed” but managing to stay on topic and explain what he had been doing on his desk calculator at UCLA. “After about half an hour he asked if I would like to join his Electronic Computer Project,” Thompson recalls. “Then he asked how my assignment should be arranged. I suggested that he call Gen. Holzman and request it. He called, talked for a few minutes, held the phone and said Gen. Holzman would like to speak with me. The conversation was very short and one-sided. It went something like, ‘Well, I guess you better go back and get your gear. Orders will follow.’ ”17
Thompson arrived at the Institute in December 1946, moving into one of the Mineville apartments at the foot of Olden Lane. “He was tall and very aristocratic,” remembers Akrevoe Kondopria. “And he was very good looking, almost like Peter O’Toole … and he wore a uniform. I gather he was the one who took the sugar.” The meteorological group consisted mostly of temporary visitors. “I shared an office with Paul Queney, from the Sorbonne, a little office under the eaves of Fuld Hall,” remembers Thompson. “We had a hard time communicating because his English was no better than my French.”18
At the Institute, meteorologists were almost as suspect as engineers. “Any study of the weather, even a study leading to eventual scientific control, was primarily an empirical rather than a theoretical science, and as such belonged in an engineering school rather than in an institution devoted to the liberal arts,” argued Marston Morse.
Except for von Neumann and Veblen, the mathematicians “approved of this step with some reluctance,” the minutes report. Resigned to the inevitable, Morse warned that “if such a study were embarked on in connection with the electronic computer project, great care should be taken to separate it from the work of the Institute as such.”19
Meteorology had been part of the computer project from the start. In mid-1945, Vladimir Zworykin had begun to see meteorology as an opportunity for RCA. Whether Zworykin enlisted von Neumann or von Neumann enlisted Zworykin remains unknown. “I remember in late 1945 or early 1946 reading a rather fantastic proposal of Zworykin’s for the construction of an analogue computer which would scan two-dimensional distributions of weather data projected on a screen and then compute the future weather by analogue techniques,” Jule Charney later explained. “By varying the input continuously and observing the output one could determine how most efficiently to modify the input to produce a given output. Johnny was in contact with Zworykin at that time, and perhaps his interest in weather computation and weather modification began then.”20 Von Neumann and Zworykin went to Washington, D.C., together to sell their plan.
“In the late summer of 1945, after the end of the war in Europe and in Asia, John von Neumann … and Vladimir Zworykin … called on me at the Navy Department,” remembers Lewis Strauss. His visitors described the digital storage tubes being developed at RCA, and how “observations on temperature, humidity, wind direction and force, barometric pressures, and many other meteorological facts at many points on the earth’s surface and at selected elevations above it … could be stored in the ‘memories’ of these tubes.” From this digital representation, “a pattern or harmonic system might be developed which eventually would enable such a data storage device to predict weather at extremely long range.”21 The numerical model would be captured in vacuum tubes from which all traces of the real atmosphere had been withdrawn.
Zworykin drafted an eleven-page “Outline of Weather Proposal,” dated October 1945, which suggested that computerized forecasting “would be a first step in any attempt in the control of weather, a goal recognized as eventually possible by all foresighted men.” With sufficiently detailed knowledge, “the energy involved in controlling the weather would be very much less than that involved in the weather phenomenon itself.” Von Neumann appended a cover letter, adding that “the mathematical problem of predicting weather is one which can be tackled, and should be tackled, since the most conspicuous meteorological phenomena originate in unstable or metastable situations which could be controlled, or at least directed, by the release of perfectly practical amounts of energy.”22
Von Neumann and Zworykin proposed that the Institute for Advanced Study, the Radio Corporation of America, and the navy collaborate, and with Strauss firmly on board, the IAS computer project was launched. “They pointed out the military advantages of accurate long-range weather intelligence, and this seemed to justify the cost of such a venture, es
timated at about $200,000,” says Strauss. “Had the decision to make the computer not been taken in 1945, the thermonuclear program might have been delayed long enough for the Soviets to have had the first weapons. This was far from the minds of anyone when Von Neumann initiated the project.”23
Thermonuclear weapons, however, were very much on von Neumann’s mind in late 1945, although this would have been kept secret from Zworykin and, for the time being, even from Strauss. Preparations were already under way for the thermonuclear calculation that would begin running on the ENIAC on December 10, 1945, and, acutely aware of the limitations of the ENIAC, the weaponeers were scrambling to start building its successor without delay. Meteorology offered both a real problem and a perfect cover for the work on bombs.
The first public announcement of the project was made by the New York Times after a meeting between Zworykin, von Neumann, and Francis W. Reichelderfer, chief of the U.S. Weather Bureau in Washington, D.C. The “development of a new electronic calculator, reported to have astounding potentialities … might even make it possible to ‘do something about the weather,’ ” the Times reported. “Atomic energy might provide a means for diverting, by its explosive power, a hurricane before it could strike a populated place.”24
With the details of the ENIAC still restricted, the Times reported vaguely that “none of the existing machines, however, is as pretentious in scope as the von Neumann–Zworykin device.” Von Neumann and Zworykin proposed to build not just one computer, but a network of computers that would span the world. “With enough of these machines (one hundred was mentioned as an arbitrary figure) area stations could be set up which would make it possible to forecast the weather all over the world.”25
Reichelderfer was upset that news of the project had been leaked to the press. Eckert and Mauchly were upset that the New York Times had made no mention of the ENIAC, but had mentioned the proposed IAS/RCA computer, which did not even exist. They felt scooped by von Neumann, as they had over the authorship of the EDVAC report, and prevented, by the secrecy imposed on their own project, from voicing a response.