Alan Turing: The Enigma: The Book That Inspired the Film The Imitation Game
Page 48
These events were unknown on the Allied side, where roughly parallel but larger developments had been taking place. In Britain there was no such digital calculator, controlled by a sequence of instructions, except the Colossi. This was in marked contrast with the situation in the United States. The British success, frantic but triumphant, had been achieved at the last moment by individuals giving their all to wartime public service. The Americans, so much richer in capitalist enterprise, were years ahead in pursuing two different, perhaps slightly unimaginative, approaches to the Babbage idea – and in doing so even in peacetime, just as they were ahead in the analogue differential analyser of the earlier 1930s. For again it was 1937 when at Harvard the physicist H. Aiken began to realise it in terms of electromagnetic relays. The resulting machine was built by IBM and turned over to the US Navy for secret work in 1944. It was lavish and impressive in scale, but like Zuse’s machines, did not embody conditional branching, even though Aiken knew of Babbage’s plans. The instructions had to be followed rigidly from beginning to end. Aiken’s machine was also more conservative than Zuse’s in that it based the arithmetical machinery on the decimal notation.
The second American project was under way at Bell Laboratories. Here the engineer G. Stibitz had first only thought of designing relay machines to perform decimal arithmetic with complex numbers, but after the outbreak of war had incorporated the facility to carry out a fixed sequence of arithmetical operations. His ‘Model III’ was under way in the New York building at the time of Alan’s stay there, but it had not drawn his attention.
There was another person, however, who did make a thorough examination of these two advanced projects, and who like Alan had the mind with which to form a more abstract view of what was happening. This was that other mathematician of the wizard war, John von Neumann. He had been connected with US Army ballistic research as a consultant since 1937. From 1941 most of his time had been spent upon the applied mathematics of explosions and aerodynamics. In the first six months of 1943 he was in Britain, conferring on those subjects with G. I. Taylor, the British applied mathematician. It was then that he had first become involved in programming a large calculation, in the sense of organising how it could best be done by people working on desk calculators. Back in the United States, his entry in September 1943 into the atomic bomb project had taken him into similar problems with shock waves, whose prediction by numerical computation required months of slogging work. In 1944 he toured the available machines in search of help. W. Weaver, at the Office of Scientific Research and Development, had put him in touch with Stibitz, and on 27 March 1944 von Neumann wrote22 to Weaver:
Will write to Stibitz: my curiosity to learn more about the relay computation method, as well as my expectations concerning possibilities in this direction, are much aroused.
On 10 April he wrote again to say that Stibitz had shown him ‘the principle and working of his relay counting mechanisms’. On 14 April he wrote to R. Peierls at Los Alamos about the ‘shock decay problem’, saying that it could probably be mechanised, and adding that he was now also in touch with Aiken. In July 1944 there were negotiations to use the Harvard-IBM machine. But then everything changed. For the pressure of wartime demands had brought about the same technological revolution as had happened at Bletchley, and at exactly the same time. In quite another place, namely the engineering department of the University of Pennsylvania (the Moore School), work had begun on yet another large calculator in April 1943. This was the ENIAC – the Electronic Numerical Integrator and Calculator.
The new machine was designed by the electronic engineers J. P. Eckert and J. Mauchly, although von Neumann’s first knowledge of it, apparently something of an accident, came through talking on a railroad station with H. H. Goldstine, a mathematician associated with the project. Von Neumann seized upon the possibilities opened up by a machine that when built would perform arithmetical operations a thousand times faster than Aiken’s. From August 1944 he was regularly attending ENIAC team meetings, writing on 1 November 1944 to Weaver:
There are some other things, mostly connected with mechanised computation, which I should like a chance to talk to you about. I am exceedingly obliged to you for having put me in contact with several workers in this field, especially with Aiken and Stibitz. In the meantime I have had a very extensive exchange of views with Aiken, and still more with the group at the Moore School … who are now in the process of planning a second electronic machine. I have been asked to act as their adviser, mainly on the matters connected with logical control, memory, etc.
The ENIAC project was immensely impressive, quite enough to give people allowed to see it the sensation of seeing the future. It employed no fewer than 19,000 electronic valves. As such it surpassed the Colossi, with which in many ways it was comparable, though one difference was that in the summer of 1945 the ENIAC was still incomplete, and would come too late to have any use whatever in the war.
It required a greater total number of valves than the Colossi because it stored long decimal numbers – the more so because of the primitive system the designers had employed whereby ten valves were allocated to each decimal digit required, a ‘9’ being represented by the ninth of those valves being ‘on’. In contrast the Colossi operated on single pulses which represented the logical ‘yes’ or ‘no’ of holes in telegraph tape.
But this was a fairly superficial point of difference. Both alike demonstrated that thousands of valves, hitherto regarded as too unreliable for operation en masse, could be kept in simultaneous use.* And the ENIAC project was embodying the idea that Zuse, Aiken and Stibitz had missed. Like the Mark II Colossi, with their ability to automate acts of decision, the results of one counting operation automatically deciding what step was taken next, the ENIAC would have a form of conditional branching. It was designed so that it could be made to hop to and fro within the stock of instructions supplied to it, repeating sections as many times as the progress of the calculation showed was necessary, without the interference of human management. None of this went beyond what Babbage had envisaged – except that electronic components were so much faster, and that the ENIAC was (or was nearly) a reality.
Like the Colossus, the ENIAC had been designed for a specific task, that of calculating artillery-range tables. Essentially it was to simulate the trajectories of shells through varying conditions of air resistance and wind velocity, which involved the summation of thousands of little pieces of trajectory. It had external switches which would be set to store the constant parameters for a trajectory calculation, and further external devices for setting up the instructions on how to calculate the segments of motion. Then there would be valves to hold the intermediate working figures. In these arrangements it resembled the Colossi. But in both cases, people had quickly discovered the possibility for using the machines for a wider range of tasks than those for which they had been designed. The role of the original Colossus had been much extended by Donald Michie and Jack Good, and then the Mark II had once been set up to punch out a deciphered message, although this was done out of interest, not for the sake of efficiency. Even though a parasite on the German cipher machine, the flexibility offered by its instruction table facility was such that it could even be ‘almost’ set up to perform numerical multiplication. The ENIAC was flexible in a more serious way, and von Neumann had already discovered that it could be used, when ready, for Los Alamos problems.*
But the ENIAC had not been conceived as a universal machine, and in one important respect the designers had departed from Babbage’s line of development. Babbage had been proud of the fact that the planned Analytical Engine would be able to ingest an unlimited number of instruction cards. The Aiken relay machine enjoyed the same feature, although cards had been replaced by a sort of pianola roll. But on the ENIAC the state of affairs was different. Its operations, being electronic, would be so fast as to make it impossible to supply cards or tape quickly enough. The engineers had had to find a way to make the instructions availa
ble to the machine in electronic times of a few millionths of a second.
On the ENIAC they were arranging for this by a system of external devices which would set up the instructions for each job. It took the form of making connections with plugs as on a manual telephone exchange. (The Colossi had something very similar.) The advantage of this solution was that the instructions would in effect be available instantaneously, once the plugging work was done. The disadvantage was that the sequence of instructions was limited in length, and that it would take a day or so to do the plugging. It would be like building a new machine for each different task. Both ENIAC and Colossus were like kits out of which many slightly different machines could be made. Neither sought to embody the true universality of Babbage’s conception, in which the machinery would be entirely unchanged, and only the instruction cards rewritten.
But even when von Neumann joined the ENIAC team as ‘adviser’ in late 1944, Eckert and Mauchly had perceived a quite different solution to their problem. This was to leave the hardware alone, and to make the instructions available at electronic speeds by storing them internally, in electronic form. The ENIAC was designed to store its arithmetical working internally, and the point of the original Colossus had been that it would hold Fish key-patterns internally. It was a quite different matter to consider holding the instructions to the machine internally. Instructions were naturally thought of as coming from the outside, to act upon an inside. But the ‘second electronic machine’ mentioned in von Neumann’s letter to Weaver was intended to incorporate this new idea.
Every tradition of common sense and clear thinking would tend to suggest that ‘numbers’ were entirely different in kind from ‘instructions’. The obvious thing was to keep them apart: the data in one place, and the stock of instructions to operate on the data, in another place. It was obvious – but wrong. In March and April 1945, the ENIAC team had prepared a proposal, the Draft Report on the
EDVAC. The EDVAC – the Electronic Discrete Variable Calculator – was the planned ‘second electronic machine’. The report was dated 30 June 1945, and signed by von Neumann. It was not his design, but the description of it bore the mark of his more mathematical mind rising above the technicalities.
In particular, it articulated the very cautious, circumspect, but quite new idea, at which the ENIAC team had arrived in planning a better machine. It discussed the different kinds of storage that existing machines required: intermediate results, instructions, fixed constant parameters, statistical data, and then stated that23
The device requires a considerable memory. While it appeared, that various parts of this memory have to perform functions which differ somewhat in their nature and considerably in their purpose, it is nevertheless tempting to treat the entire memory as one organ.
But such a proposal, that of the ‘one organ’, was equivalent to adopting the ‘one tape’ of the Universal Turing Machine, on which everything – instructions, data, and working – was to be stored. This was the new idea, different from anything in Babbage’s design, and one which marked a turning point in proposals for digital machines. For it threw all the emphasis on to a new place – the construction of a large, fast, effective, all-purpose electronic ‘memory’. And in its way it made everything much more simple, less cluttered in conception. Von Neumann might well have seen it as ‘tempting’, because it was almost too good an idea to be true. But it had been in Computable Numbers all the time.
So the spring of 1945 saw the ENIAC team on the one hand, and Alan Turing on the other, arrive naturally at the idea of constructing a universal machine with a single ‘tape’. But they did so in rather different ways. The ENIAC, now already shown to be out-of-date in principle even before it was finished, had been something of a sledgehammer in cracking open the problem. And von Neumann had been obliged to hack his way through the jungle of every known approach to computation, assimilating all the current needs of military research and the capabilities of American industry. The result was something close to the Lancelot Hogben view of science: the political and economic needs of the day determining new ideas.
But when Alan Turing spoke of ‘building a brain’, he was working and thinking alone in his spare time, pottering around in a British back-garden shed with a few pieces of equipment grudgingly conceded by the secret service. He was not being asked to provide the solution to numerical problems such as those von Neumann was engaged upon; he had been thinking for himself. He had simply put together things that no one had put together before: his one-tape universal machine, the knowledge that large-scale electronic pulse technology could work, and the experience of turning cryptanalytic thought into ‘definite methods’ and ‘mechanical processes’. Since 1939 he had been concerned with little but symbols, states, and instruction tables – and with the problem of embodying these as effectively as possible in concrete forms. Now he could consummate it all.
And now the war was over, his motives were much nearer to those of G. H. Hardy than to the practicalities of the world’s work. They had more to do with the paradox of determinism and free will, than with the effecting of long calculations. Of course, no one was likely to pay for a ‘brain’ that had no useful application. And in this respect Hardy would have found justification of his views regarding the applications of mathematics. On 30 January 1945 von Neumann had written24 that the EDVAC was being designed for three-dimensional ‘aerodynamic and shock-wave problems … shell, bomb and rocket work … progress in the field of propellants and high explosives’. Such, in Churchill’s phrase, would be the ‘progress of mankind’. Alan Turing too would have to come a long way from the logic of Hilbert and Gödel if he was really to build a brain.
The Draft Report on the did carry a more theoretical burden (one reflecting von Neumann’s interests) in that it drew attention to the analogy between a computer and the human nervous system. The use of the word ‘memory’ was an aspect of this. In its way, it was ‘building a brain’. However, its emphasis was placed not on an abstract thesis concerning ‘states of mind’, but on the similarities between input and output mechanisms, and afferent and efferent nerves respectively. It also drew upon the 1943 paper written by the Chicago neurologists W. S. McCulloch and W. Pitts, which analysed the activity of neurons in logical terms, and used their symbolism for describing the logical connections of electronic components.
McCulloch and Pitts had been inspired by Computable Numbers, and so in a very indirect way, the EDVAC proposal owed something to the concept of the Turing machine. One of its curiosities, however, was that it made no mention of Computable Numbers, nor made precise the universal machine concept. Yet von Neumann had been familiar with it before the war, and must surely have recognised the connection when he liberated himself from the assumption that data and instructions had to be stored in different ways. According to S. Frankel,25 who worked at Los Alamos on the atomic bomb and was one of the first to use the ENIAC,
in about 1943 or ’44 von Neumann was well aware of the fundamental importance of Turing’s paper of 1936 ‘On computable numbers’ … Von Neumann introduced me to that paper and at his urging I studied it with care…. he firmly emphasised to me, and to others I am sure, that the fundamental conception is owing to Turing – insofar as not anticipated by Babbage, Lovelace and others.
So the Wizard might well have learnt something from Dorothy. However, the essential point about these two initiatives, American EDVAC and British, was not the rather tenuous connection between them. It was their very marked independence.26
Whatever ideas had flowed westwards, the Draft Report on the EDVAC was the first to put them together in writing. So once again, British originality had been pipped at the post by an American publication – and at a time when everyone was looking to the west. The Americans had won, and Alan was a sporting second. This time, however, American priority was nothing but an advantage to the Turing plans, for it provided the political and economic impetus that his own ideas alone could never have enjoyed.
&nbs
p; Indeed, it was probably only the existence of the ENIAC and the EDVAC idea that made possible the next stage of Alan Turing’s life. For in June he had a telephone call at Hanslope. It was from J. R. Womersley, Superintendent of the Mathematics Division at the National Physical Laboratory.
Womersley was a new man in a new post in a new organisation. The National Physical Laboratory was not new; it had been set up in shabby, suburban Teddington in 1900, the British response to state-sponsored German scientific research. Its site was carved out of Bushy Park, itself largely given over to Supreme Headquarters, Allied Expeditionary Forces. It was the most extensive government laboratory in the United Kingdom, enjoying a high reputation within its traditional sphere, that of setting and maintaining physical standards for the benefit of British industry. Its current Director, installed in 1938, was Sir Charles Galton Darwin, grandson of the theorist of evolution and himself an eminent Cambridge applied mathematician. His major contribution had been in the field of X-ray crystallography, and like Humpty Dumpty, who was able to explain the Jabberwocky, he was regarded as the27 ‘interpreter of the new quantum theory to experimental physicists’. Large, awesome, and remote, during the war he had spent a year as director of what became the British Central Scientific Mission in Washington, and had been the first scientific adviser to the British Army.
The Mathematics Division, however, was new. Indeed, it was a computational equivalent of the planned welfare state, the product of a calculators’ Beveridge Report. In about March 1944 a proposal28 had been mooted for an independent Mathematical Station, and this suggestion, a fine example of wartime planning for peace, went to a large interdepartmental committee, itself a manifestation of cooperation and coordination unthinkable in peacetime days. The government accepted the principle of continuing the funding found necessary in war, and a centralised, rationalised institution was planned to take over the various ad hoc offices which had done the mostly dreary work of numerical computation for military purposes. Sir Charles Darwin had persuaded the committee to establish it as a division of the NPL.