by Matthew Cobb
Although Wiener’s insight excited his academic colleagues, his attempt to build an anti-aircraft device that could be engineered into a battlefield version was upstaged by a rival top-secret project, which was jointly run by MIT and a private company, Bell Laboratories. Under the deliberately misleading title the Radiation Lab (known as the Rad Lab), the project involved more than thirty scientists and in its first year alone had a budget of more than $800,000. Although it used a highly unrealistic prediction method – it assumed that the aircraft would fly in a straight line – the device made up for its lack of accuracy by firing a hail of shells around the predicted location, some of which would get lucky. In 1942 the Rad Lab project passed a practical test and more than 1,200 units were ordered by the US military. Although Wiener and Bigelow’s statistical predictor was marginally more accurate than their Rad Lab rival, it soon became apparent that the improvement over the Rad Lab version would not be worth the effort and Wiener’s project was cancelled in November 1943.12 The Rad Lab system, now called the M-9, incorporated some elements of Wiener’s predictive protocols and went into mass production. It eventually formed a central component of what was the first robot war – the clash between the German V-1 automatic rockets or doodlebugs and the Allies’ semi-automated defence systems, in the skies over southern England in 1944–45.13
Wiener wrote up his method for predicting the movement of objects and filtering out noise under the daunting title ‘Extrapolation, interpolation, and smoothing of stationary time series with engineering applications’. This duplicated document consisted mainly of pages and pages of Wiener’s fiendishly complex mathematics and was circulated to workers on the various anti-aircraft projects. Even the briefest chapter, which was only six pages long, contained thirty-seven equations. Stamped ‘Restricted’, the document was printed in 300 copies and was bound in pale yellow covers that carried a declaration threatening anyone who revealed its contents with the full force of the US Espionage Act. It soon became known to its many perplexed readers as the Yellow Peril – one US engineer later said, ‘copies should have been distributed to the enemy so that they would have to devote such time to it and enable us to get on with winning the war.’14 The document was eventually declassified and has since become a classic of its kind; in 2005 a copy sold at Sotheby’s for $7,200.
Wiener claimed that his method had many applications and could shed important light on the nature of communication. He argued that there was no difference between human communication and messages sent by a machine: ‘the records of current and voltage kept on the instruments of an automatic substation are as truly messages as a telephone conversation.’15 All communication possesses the same fundamental feature, argued Wiener – it has to contain what he termed variable information. One measure of that variability, he said, could be found in the mathematical theory of probability – all forms of communication could be understood in terms of a mathematical, probabilistic analysis of the information they contained. When Wiener’s two fundamental insights – the importance of negative feedback in shaping behaviour, and the existence of information – were combined, it implied that the feedback loops that lay at the heart of apparently purposive behaviour were carrying information.
*
Another MIT-trained mathematician called Claude E. Shannon was working on similar problems at the same time. Shannon was a shy young man with a lifelong love of Dixieland jazz who enjoyed tinkering with electronics and spoke with a slow drawl, a bit like James Stewart. In 1938 Shannon obtained his MSc from MIT for his work on applications of Boolean logic, which played a decisive role in the development of electronics. By 1940, Shannon had completed his PhD on ‘An algebra for theoretical genetics’, in which he developed a mathematical way of describing how genes spread in populations. As Shannon admitted, although the proof was novel, the results were not. He was not actually interested in genes at all – according to his doctoral advisor, Vannevar Bush, ‘he has only a fragmentary knowledge of this aspect of genetics’. His primary concern was with using statistics to describe the behaviour of genes in populations, not how they functioned or what they were made of.16
By 1942, Shannon was working for Bell Laboratories in their New York headquarters on West Street, overlooking the Hudson River. At the rear, on Washington Street, an overground subway line ran right through the building, like something out of a 1930s film of the future. Shannon was part of the cryptography group, studying the transmission of messages over the telephone. In January 1943, as Schrödinger was about to give his lectures in Dublin, the Bell Labs had a visitor from England – the mathematician and cryptographer Alan Turing, who had arrived in New York on the Queen Elizabeth in November. Turing began work at Bell Labs, investigating ways of setting up a securely encoded telephone link between Roosevelt and Churchill – this was later successfully implemented after Shannon’s theoretical demonstration that the code could not be broken.17
Although Turing did not work with Shannon, the two young men regularly had tea together in the cafeteria, where they discussed Turing’s ideas for a ‘universal machine’ that could perform any conceivable calculation. Shannon apparently surprised Turing by suggesting that such a ‘Brain’ would be capable of more than just doing complicated sums: ‘Shannon wants to feed not just data to a Brain, but cultural things! He wants to play music to it!’ Turing exclaimed in a letter.18 More significantly, the two men also exchanged their ideas about signal transmission, how to measure the content of communication, and how to incorporate uncertainty into their mathematical procedures. Turing had developed the concept of ‘decibans’, which were a measure of the uncertainty contained in a message; Shannon was on the brink of defining the ‘binary digit’ or ‘bit’, which could have two states – 0 or 1 – and was at the heart of postwar computing. At the beginning of March, Turing returned to England on the hazardous north Atlantic crossing, the only civilian on a 4,300-strong troop ship. The next time the two men met was after the war, in Manchester, where Turing was working on Baby, the world’s first stored-program computer.
Shannon was part of the Bell Labs team working with the MIT Rad Lab on fire control. Like Wiener, Shannon’s job was to come up with a method for predicting the location of the target, and the two men discussed this question several times. Bigelow later recalled that Wiener was extremely generous, exchanging ideas with the younger man, sharing his insights. Eventually Wiener’s generosity began to wear off and – perhaps because of his amphetamine abuse – he started to react in a paranoid fashion to Shannon’s visits, telling close friends that Shannon was ‘coming to pluck my brains’, and doing his best to avoid the visitor.19
Although Shannon was clearly inspired by some aspects of Wiener’s work in the Yellow Peril, he had already begun thinking about the nature of communication and how to describe it mathematically. He was not the first person to do this; in the late 1920s Ralph Hartley and Harry Nyquist had studied how telegraph messages were transmitted, but they did not approach the problem from a probabilistic point of view, nor did they include random variation – noise – as a factor affecting transmission accuracy. In 1945, Shannon wrote a document for the D-2 division of the NRDC, entitled ‘A mathematical theory of cryptography’, in which he summarised his ideas about communication and what it involved. He called the stuff that was communicated ‘information’, and described the nature of its fundamental unit, which he termed the ‘bit’. For obvious reasons the paper was immediately stamped ‘secret’, but after the war a version of it was published and it was eventually declassified in 1957.20
The final element of the ideas about information that were coming into form around the war years was an exploration of Maxwell’s Demon by the German physicist Leo Szilárd. This thought experiment was devised in 1871 by the British physicist James Clerk Maxwell, with the aim of showing how it was theoretically possible to violate the second law of thermodynamics. Maxwell imagined a demon that without effort could open a door between two chambers, allowing the more
energetic molecules into one side, thereby increasing the temperature in that chamber and decreasing entropy in the system – something the second law said was impossible. Szilárd’s solution, which he devised in 1929, was that the demon would have to be able to measure the speed of the molecules, and to do this would require the expenditure of energy and therefore an increase in entropy. If the demon and the chamber were taken as a whole, the entropy of the system would not decline, and the second law remained intact. Although Szilárd did not use the term information, his theoretical discussion linked entropy and measures of knowledge in a way that proved significant.
*
At the beginning of 1945, Wiener and fellow mathematician John von Neumann organised a meeting of the newly formed Teleological Society. The aim of the society was to study ‘how purpose is realised in human and animal conduct and on the other hand how purpose can be imitated by mechanical and electrical means.’21 Von Neumann was a mathematician and a pioneer of game theory – mathematical models that describe and predict simple behaviours. He played an important role in developing the models that came to dominate much of postwar economics, and which also constituted a strand in the thinking of evolutionary biologists. Above all, von Neumann played a leading role in the Manhattan Project.
Eight months after the meeting of the Teleological Society, the world changed utterly, in two terrifyingly destructive flashes of light. On 6 August 1945, the US dropped the first atomic bomb on Hiroshima, causing unimaginable devastation, instantly killing up to 80,000 people, with a similar number condemned to a slow death over the following months. Three days later, on 9 August, the city of Nagasaki was destroyed by a second bomb, which used plutonium rather than uranium and employed an implosion ignition procedure developed by von Neumann.
The Manhattan Project was a success, but many of the scientists involved were horrified at their part in the destruction. The prime reason behind the Manhattan Project – fear of being beaten to the bomb by the Nazis – had been eradicated by the surrender of Germany in May 1945. Furthermore, it soon became apparent that the Germans had not been close to success. All except the most naive or unworldly scientists came to recognise that the development and deployment of the atomic bomb showed that the Allies had something else in mind – the bomb was used to threaten the USSR. Von Neumann was quite comfortable with this. He had helped decide which two Japanese cities were to be smashed; as a committed anti-communist he accepted that the Hiroshima and Nagasaki bombs were primarily warnings to the USSR, and considered that the attendant death and destruction were quite justified.22
Wiener took a very different attitude. He was concerned about the moral issues raised by the use of the bomb against Japan, and by the potential for infinitely greater destruction in the future, to the extent that he considered abandoning science altogether. As he told a friend in October 1945:
Ever since the atomic bomb fell I have been recovering from an acute attack of conscience as one of the scientists who has been doing war work and who has seen his war work a[s] part of a larger body which is being used in a way of which I do not approve and over which I have absolutely no control. … I have seriously considered the possibility of giving up my scientific productive effort because I know no way to publish without letting my inventions go to the wrong hands.23
Wiener’s wartime experience had convinced him that science should be as open as possible, and should not be tied to the private sector or the military. Von Neumann, in contrast, was keen to put his snout as deep as possible into the trough of the military–industrial complex that was beginning to dominate the US economy. His twin aims were to obtain funding to build the computer he had dreamt up, and to counter what he saw as the threat from the USSR.
Despite their profound differences, the two men continued to work together, most notably in the organisation of a conference that took place in March 1946 under the cumbersome title ‘The Feedback Mechanisms and Circular Causal Systems in Biology and the Social Sciences Meeting’ (more commonly known as the Macy conference, after the sponsors, the Josiah Macy Jr Foundation). The attendees were basically the same crowd as the people who had heard Wiener outline his negative feedback vision in 1942, with the addition of the ecologist G. Evelyn Hutchinson, the sociologist Paul Lazarsfeld and some others.24 Wiener and von Neumann presented their project of electronic computer brains, with von Neumann drawing a parallel between the human nervous system and the digital, stored-program computer he was constructing in Princeton.
Seven months later, in October 1946, the New York Academy of Sciences held a special meeting on ‘Teleological mechanisms’ at which Wiener spoke, outlining the ideas in the Yellow Peril that had been withheld from public view the year before.25 Wiener explained that underlying all examples of negative feedback control there was a single unifying idea, which he called the message – all control systems involved communication, and could be understood using the same conceptual framework. Inspired by Schrödinger’s What is Life?, Wiener made a link between information and entropy, going even further than Szilárd’s discussion of Maxwell’s Demon, in that he defined entropy as ‘the negative of the amount of information contained in the message’. This was ‘not surprising’, Wiener went on, because ‘Information measures order and entropy measures disorder. It is indeed possible to conceive all order in terms of message.’ The laws governing communication, he argued, were ‘really identical’ with the second law of thermodynamics. So, for example, once a message has been created, subsequent operations can degrade it but cannot add information. The arrow of entropy points only one way, and all that life can do is to temporarily halt the process; it cannot truly reverse it. One of the main explanatory frameworks used by postwar science – the role of information in biology – was emerging and was now connected with the fundamental measure of order on a cosmic scale.
A month later, von Neumann took a step towards linking the study of control systems with visions of how life reproduces itself. He was increasingly convinced that Wiener’s focus on modelling human behaviour was a mistake: the human brain was far too complex. At the end of November 1946, von Neumann wrote a long letter to Wiener outlining a startlingly different approach, which dominated science for decades to come.26 He began with a self-criticism, pointing out that through their shared enthusiasm for studying the human central nervous system, ‘we selected … the most complicated object under the sun – literally.’ But the problem was even greater than the mere complexity of the brain, argued von Neumann. He felt they had first to understand the underlying molecular mechanisms before they could hope to understand higher level activity:
nothing that we may know or learn about the functioning of the organism can give, without ‘microscopic’, cytological work any clues regarding the further details of the neural mechanism.
Von Neumann’s solution was radical. He concluded that they should focus on what he termed
the less-than-cellular organisms of the virus or bacteriophage type … They are self-reproductive and they are able to orient themselves in an unorganized milieu, to move towards food, to appropriate it and to use it. Consequently, a ‘true’ understanding of these organisms may be the first relevant step forward and possible the greatest step that may at all be required.
Von Neumann’s grasp of virus biology was flimsy – he told Wiener that a virus was ‘definitely an animal, with something like a head and a tail’ (in fact viruses are not even alive by most definitions). Despite being a poor biologist, von Neumann’s suggestion that simple systems can reveal principles that apply to more complex forms of organisation was absolutely right. He calculated that each virus consisted of around 6,000,000 atoms, and ‘only … a few hundred thousand “mechanical elements”.’ Von Neumann explained to Wiener that it should be possible to understand the interaction of these components, although he recognised that even this proposal was challenging:
Even if the complexity of the organisms of molecular weight 107–108 is not too much for us, do we
not possess such means now, can we at least conceive them, and could they be acquired by developments of which we can already foresee the character, the caliber, and the duration.
Von Neumann suggested to Wiener that they should study the ‘physiology of viruses and bacteriophages, and all that is known about the gene-enzyme relationship’. His explanation for this approach was that viruses could give an insight into genes, which suggests he understood more about viruses than implied by his statement that they were animals:
Genes are probably much like viruses and phages, except that all the evidence concerning them is indirect, and that we can neither isolate them nor multiply them at will.
Von Neumann was becoming interested in genes because one of the essential features of life that fascinated him was its ability to replicate itself. Indeed, from his previous thinking about self-reproducing automata, von Neumann now felt sure that ‘self-reproductive mechanisms’ in living things could be understood in terms of the framework that he and Wiener had been developing:
I can show that they exist in this system of concepts. I think that I understand some of the main principles that are involved. … I hope to learn various things in the course of this literary exercise, in particular the number of components required for self-reproduction. My (rather uninformed) guess is in the high ten thousands or in the hundred thousands, but this is most unsafe.
