by Matthew Cobb
The two men were well aware of this difference; in October 1948 Shannon wrote to Wiener:
I consider how much information is produced when a choice is made from a set – the larger the set the more information. You consider the larger uncertainty in the case of a larger set to mean less knowledge and hence less information. The difference in viewpoint is partially a mathematical pun. We would obtain the same numerical answers in any particular question.19
The explanation for these apparently opposite approaches, and the source of what Shannon called a pun was that whereas he was strictly focused on the relatively narrow issue of communication, Wiener wanted to create a grander, broader theory, which integrated communication, control, organisation and the nature of life itself. Wiener’s model therefore had to include meaning, something that Shannon opposed. This contrast between the two formulations was therefore far more than a mathematical pun – it reflected differences in the two men’s ambitions.
*
Wiener and Shannon’s ideas had an important influence on the scientific community. Information came to be seen as a characteristic of matter that could be quantified, preferably in terms of binary coding, while control and negative feedback seemed to be fundamental features of organic and engineered systems. One of the ways in which these ideas came to influence biologists was through the promise of creating automata and thereby testing models of how organisms function and reproduce.20 These links were explored at a symposium on Cerebral Mechanisms in Behavior that was held at the California Institute of Technology (Caltech) a month before the publication of Cybernetics, at the end of September 1948. The symposium was a small affair – only fourteen speakers, with a further five participants, one of whom was the Caltech chemist Linus Pauling.
1. Claude Shannon’s model of communication. From Shannon and Weaver (1949).
Von Neumann gave the opening talk, entitled ‘General and logical theory of automata’, and explored one of the defining features of life: its ability to reproduce. Von Neumann’s starting point was Alan Turing’s prewar theory of a universal machine that carried out its operations by reading and writing on a paper tape. But this was too simple for von Neumann: he wanted to imagine ‘an automaton whose output is other automata’.
Von Neumann argued that such a machine needed instructions to construct its component parts, and that these instructions would be ‘roughly effecting the functions of a gene’; a change in the instruction would be like a mutation. Von Neumann explained that a real gene ‘probably does not contain a complete description of the object whose construction its presence stimulates. It probably contains only general pointers, general cues.’ In contrast, the ‘fundamental act of reproduction, the duplication of genetic material’ could be conceptualised in terms of the copying of a paper tape – von Neumann was implicitly arguing that a gene was like one of Turing’s instruction-containing tapes.21 Although von Neumann did not make the link, this was a computer version of Schrödinger’s aperiodic crystal that contained a code-script.
Scientists who were not at the meeting were also becoming interested in the links between cybernetics and genetics. Wiener was in contact with the British geneticist J. B. S. Haldane, who had been following Wiener’s work with close attention. In mid-November 1948, Haldane wrote to Wiener as he struggled to apply the new concepts to his field:
I am gradually learning to think in terms of messages and noise. … I suspect that a large amount of an animal or plant is redundant because it has to take some trouble to get accurately reproduced, and there is a lot of noise around. A mutation seems to be a bit of noise which gets incorporated into a message. If I could see heredity in terms of message and noise I could get somewhere.22
Similar ideas were being developed by the University of Illinois physicist Sydney Dancoff, together with his colleague, the Austrian-born radiologist Henry Quastler. In July 1950, Dancoff wrote a letter to Quastler summarising the two men’s thinking on the links between genetics and information. The chromosome, wrote Dancoff, could be seen as a ‘linear coded tape of instructions.’ He went on:
The entire thread constitutes a ‘message’. This message can be broken down into sub-units which may be called ‘paragraphs,’ ‘words,’ etc. The smallest message unit is perhaps some flip-flop which can make a yes-no decision. If the result of this yes-no decision is evident in the grown organism, we can call this smallest message unit a ‘gene’.23
Dancoff and Quastler were forcing genetics into a binary mode, viewing organisms as an example of the automata imagined by Turing and von Neumann.
In Europe, too, the scientific community discussed the new ideas about information and control, with each country taking a slightly different approach to the question. In September 1950, the Royal Society of London organised a three-day conference on ‘information theory’ – this was not a term used by Shannon or Wiener, although it has since become widespread. Around 130 attendees crowded into the small lecture theatre in Burlington House on Piccadilly to discuss the mathematical and electronic aspects of the field. Shannon gave three talks at the meeting, but there was little exploration of Wiener’s cybernetic approach. There were no geneticists to be seen. The only person to speak about genes was Alan Turing, and he was more interested in how natural selection alters the shape of organisms than in thinking about how heredity works.24
The degree to which information had inserted itself into scientific thinking and ordinary language was revealed at the end of 1950, when the British zoologist J. Z. Young gave the prestigious Reith Lectures on the BBC, under the title ‘Doubt and certainty in science’. The first three radio lectures were about how biologists study brain function, and they teemed with the word ‘information’.25 Listeners were presented with this new vision as though it were the only way of understanding how nervous systems work. What it really showed was that the way that biologists thought about life had been transformed.*
In France the Nobel Prize-winning physicist Louis de Broglie gave a lecture series in spring 1950 under the general title ‘Cybernetics’, and in 1951 a congress was held in Paris, funded by the Rockefeller Foundation, that was attended by over 300 people including Wiener and McCulloch. After the congress was over, Wiener remained in Paris to give several talks on the subject at the prestigious Collège de France. There were also articles in French journals such as Esprit and the Nouvelle Revue Française while, for the general public, science journalist Pierre de Latil wrote a lively book called La Pensée Artificielle, which explained the nature and genesis of cybernetics, full of useful diagrams and photos, but focusing on feedback and showing how French engineers had come up with the concept in the fifteenth and nineteenth centuries.26 De Latil’s book embodied the contrast between the French and British approaches to cybernetics: whereas the British focused on information, the French emphasised the control and robotics aspects of the subject. Strikingly, de Latil’s book did not refer to Shannon at all. In the UK and the US ‘information’ was widely discussed in popular science magazines such as The Times Science Review and Scientific American, both as an abstract concept and in Shannon’s mathematised version.27
Whatever the contrasts between different subject areas and different countries, in Britain, America and France, everyone in science knew that a conceptual revolution was taking place. Not everyone was impressed, however. In 1948 Max Delbrück was invited to one of the cybernetics conferences. It was the only such meeting he attended. Not a man to mince his words, Delbrück later recalled that he found the discussion ‘too diffuse for my taste. It was vacuous in the extreme and positively inane.’28
*
In 1950, public interest in cybernetics was cranked up even further when Wiener published a second book – this time without a single equation – in which he outlined the potential changes that society would have to face as a result of increased automation. With the unwieldy title On the Human Use of Human Beings, Wiener’s new book explained how society should respond to the looming cultural and econo
mic developments that would follow the introduction of automation and the development of computers in the second half of the twentieth century. It spanned a wide range of culture in a free-wheeling style, dealing with language, law and individuality, exploring their changing meaning in the context of machines that could apparently embody aspects of purposeful behaviour. Wiener was concerned that top-down social control was becoming typical of all economic and political systems across the planet. As he explained: ‘I wish to devote this book to a protest against this inhuman use of human beings.’29 Professor Cyril Joad, the BBC’s favourite philosopher, hated the book. His review in the Times Literary Supplement was scathing, criticising Wiener’s ‘incoherent and slovenly language’ and branding the book as ‘highly dangerous’.30
Despite Joad’s fears, Wiener’s remarkable book had a lasting influence because it showed that everything – including humans – could ultimately be reduced to mere patterns of information. Wiener again emphasised the links between the latest technological developments and the way in which organisms behave and function: ‘It is my thesis that the operation of the living individual and the operation of some of the newer communication machines are precisely parallel.’31 If humans were essentially machines in both form and function, then it should be possible to define a human being in terms of the information they contained, by calculating ‘the amount of hereditary information,’ he claimed. If that information could be represented in some way, then it would even be possible to transmit it by electronic means, preserving the identity of the individual, argued Wiener.32 Although he realised that such a procedure would remain in the realms of science fiction for the foreseeable future, Wiener had made his point: a human being was fundamentally no different from any other form of organised matter. In the end, it was information.
Wiener was not the only person to be thinking along these lines. In July 1949, Shannon sketched a list of different items and their ‘storage capacities’. He considered that a ‘phono record’ contained about 300,000 bits of information, one hour of broadcast TV contained 1011 bits, and the ‘genetic constitution of man’ was a mere 80,000 bits.33 Nothing became of these wildly inaccurate guesstimates – the sketch remained in the Shannon archives until it was recently discovered by James Gleick – but they show how the concept of information could be applied to virtually anything. In May 1952, J. B. S. Haldane wrote a letter to Wiener in which he announced that he had ‘worked out the total amount of control (= information = instruction) in a fertilized egg’.34 It is not known what number Haldane came up with, nor on what basis he made his calculation – he never published his answer to this conundrum.
Henry Quastler was bolder. In March 1952 he organised a symposium on Information Theory in Biology, which was held in his Control Systems Laboratory at the University of Illinois. The rising star of bacterial genetics, Joshua Lederberg, had been invited to attend, but he was wary because the meeting was funded by the US Office of Naval Research. Lederberg was concerned that the discussions were to be recorded and might involve matters that could be the subject of a future security classification.35 Lederberg was not being paranoid – the McCarthyite witch-hunts were getting into their stride, leading to US academics having to swear ‘loyalty oaths’ or risk losing their jobs; one ill-judged comment could lead to disaster.
The speakers at the symposium showed how scientists were trying to apply the new concept of information to biology. One participant discussed Linus Pauling’s models of the molecular structure of the protein keratin in terms of the information they contained, while another explored the information content of a zygote, which he argued was ‘a set of instructions coded into the fertilized egg as dictated by the genetic constitution’. There were even two attempts to calculate the information contained in organisms. Henry Linschitz used molecular and energetic calculations to conclude that ‘the information content of a bacterial cell’ was around 1013 bits.36 In a joint paper that was completed after Dancoff’s untimely death, Quastler outlined what he accepted were ‘crude approximations and vague hypotheses’ and then calculated that a human genome contains at most 1010 bits of information. This calculation took as its starting point the view that each gene and its different versions, or alleles, is ‘an independent source of information, with an entropy which depends on the number of allelic states, or different messages’. Quastler admitted that he knew ‘neither the number of genes nor the average number of allelic states’, both of which would appear to be essential for such a calculation to be valid. Unperturbed, Quastler concluded: ‘this is an extremely coarse estimate, but it is better than no estimate at all.’37 Not everyone agreed. In 1965, after the genetic code had been cracked but before it had been completely deciphered, Michael Apter and Lewis Wolpert returned to Dancoff and Quastler’s figures and dismissed them as ‘so arbitrary as to make the values obtained meaningless’. Their conclusion about the pointlessness of Dancoff and Quastler’s calculation was cutting:
We believe that, on the contrary, they are not better than no estimate at all, since such estimates are liable to be misleading and to breed a false confidence.38
Applying the new concepts of information to genetics was proving to be more difficult than many people expected.
*
It seemed that cybernetics and communication theory were going to sweep away everything in their path, changing the whole of science. But then some of its principal proponents were thrown into turmoil by personal and political events. The backdrop to the developments in cybernetics, and indeed the source of much of its funding, was the Cold War. In February 1949, the US lost its monopoly on nuclear weapons when the USSR exploded its first atom bomb. In 1950, the Cold War began to heat up as the Korean War broke out and the US fought a proxy war against the Russians and the Chinese. Shocked by these developments, the anti-communist John von Neumann pressed the US government to focus all its research effort on building a hydrogen bomb. Thanks in part to his lobbying, a major development programme began in which he was heavily involved, leaving little time for his other interests. The project culminated in the explosion of the first H-bomb in November 1952, with a yield that was nearly 1,000 times more destructive than that of Hiroshima. Nine months later, the USSR exploded its own thermonuclear device, and the arms race began in earnest. Von Neumann abandoned his interest in creating self-reproducing automata and spent much of the rest of his life until his death in 1957 working on intercontinental ballistic missiles, applying his mathematical genius to the potential destruction of the human race.39
In 1951 Wiener’s collaboration with Pitts and McCulloch, which had been at the heart of the development of cybernetics for nearly a decade, came to an abrupt end as he broke off all contact with the two younger men. Wiener had been severely irritated by a trivial but jocular letter sent by Pitts and McCulloch, but the source of the crisis appears to have been his wife’s malicious and entirely untrue suggestion that Pitts and McCulloch had seduced Wiener’s 19-year-old daughter Barbara. The cybernetics group was severely weakened and never recovered from the blow.40
Other thinkers were taking up the banner of cybernetics. In 1950 Hans Kalmus, a geneticist based at McGill University in Montreal who worked closely with J. B. S. Haldane, published a brief article in The Journal of Heredity entitled ‘A cybernetical aspect of genetics’. Kalmus had read Cybernetics and had been struck by what he called ‘certain unifying principles’ that shed an interesting light on heredity. Kalmus suggested that a gene ‘is a message, which can survive the death of the individual and can thus be received repeatedly by several organisms of different generations.’ Kalmus even claimed there was a parallel between what he called the racial memory of genes and the recently developed ability of computers to store information.41
Kalmus went on to argue that there was no contradiction between the widely held enzymatic view of gene function and the cybernetic vision he was putting forward. Even more ambitiously, Kalmus tried to show how cybernetic concepts of feedback could expla
in interactions between genes at the genomic and populational levels, and interactions between genes and environmental factors – climate, other organisms and so on. However, Kalmus had little to say by way of detail and his ideas had no discernible influence. The first person to cite his article was … H. Kalmus, in 1962.42
As cybernetics and information theory became fashionable in fields far removed from automata and electronic communications, it became a target for ridicule. One example was a satirical letter to Nature that was cooked up over a well-oiled lunch in the Italian Alps, in September 1952.43 Boris Ephrussi and Jim Watson were dining with one of Ephrussi’s colleagues, Urs Leopold, and they decided to write a brief spoof letter, taking the mickey out of a recent review by Joshua Lederberg in which Lederberg had rather pompously suggested that several well-established terms in bacterial genetics should be replaced by fancy new words that he had invented.44
Ephrussi and Watson’s ‘joke’ consisted of satirically suggesting that words such as transformation, induction, transduction and infection, all of which had recently come into currency in bacterial genetics, should be replaced by the single term ‘inter-bacterial information’. Several points could have alerted the reader that this was not intended to be taken seriously. The suggested change did not make sense – ‘information’ could not be a grammatical substitute for ‘transformation’. Furthermore, the brief letter closed by reassuring the reader that their preferred term did ‘not imply necessarily the transfer of material substances’ – but the only alternative to a material transfer would be something like a radio broadcast between bacteria. The final phrase was equally facetious, as it highlighted ‘the possible future importance of cybernetics at the bacterial level’, conjuring an apparently surreal comparison of single-celled organisms with the most complex machines on Earth at the time.45