In the meantime, Alan’s London lectures over the winter had left some observers cold:
On a series of Thursday afternoons in Dec 1946 & January 1947, Turing delivered a course of 7 lectures on the design of the ACE at the Ministry of Supply in London. I attended the first one or two, but then decided that I could employ my time more suitably than in hearing about the details of the ACE, which I regarded as a very odd machine.
The reason it was odd was the just-in-time programming designed around the delays inherent in delay-line memory. The oddity was noted by Dr M.V. Wilkes, who was setting up the Mathematical Laboratory in Cambridge, which was aiming to build a computing machine of its own. Wilkes had been invited to the Turing lecture series by Womersley, who had told Wilkes that ‘we are very anxious indeed to have all the help and co-operation we can from you. Could you come next Wednesday, lunch here, and talk the matter over with Turing and myself?’ This was in the spirit of the times, when the vision of the NPL Mathematics Division was to use computing machinery as a service bureau to assist the mathematicians, and the wider scientific community, with their computational problems. It was a good idea for there to be several bureaux around the country developing their own machines: several bureaux would ease capacity bottlenecks, and there would be sharing of different ideas developed in different laboratories. Wilkes sent over a four-page outline of his plan, which Alan Turing was asked by Womersley to criticise. There was just one problem. Alan Turing and Maurice Wilkes could not abide one another.
Mr Womersley
I have read Wilkes’ proposals for a pilot machine, and agree with him as regards the desirability of the construction of some such machine somewhere. I also agree with him as regards the suitability of the number of delay lines he suggests. The ‘code’ which he suggests is however very contrary to the line of development here, and much more in the American tradition of solving one’s difficulties by means of much equipment rather than by thought. I should imagine that to put his code (which is advertised as ‘reduced to the simplest possible form’) into effect would require a very much more complex control circuit than is proposed in our full-size machine. Furthermore certain operations which we regard as more fundamental than addition and multiplication have been omitted.
On the surface, and in public, strict British politeness was at all times maintained. However, it became notorious that they antagonised each other. Maurice Wilkes was an exact contemporary of Alan’s. His name appears four below Alan’s in the list of mathematics tripos finalists for 1934, as one of the mathematical superheroes now familiar to us as B-star wranglers. Wilkes’s background was very different, though: Wilkes’s father had been a ‘clerk’ (in fact chief administrator) for the Earl of Dudley, and a world away from a Raj upbringing which put the Turing parents in a position of command. Like Alan, Wilkes liked machines and tinkering with them; unlike Alan, he was good at it – he was more of a physicist than a pure mathematician, and did his thesis on radio in the ionosphere, a subject which was valuable during the war years when Wilkes worked on radar. Before the war, Wilkes had mastered the delicate machinery of differential analysers (his machine was also made of Meccano parts); in 1937 he became University Demonstrator in the Mathematical Laboratory at Cambridge. One might be snobbish about the status of Demonstrator, which is the most inferior of academic appointments, but Prof didn’t have any university post at all before 1948. After the war Wilkes was back in his Cambridge lab and preparing to create a computation service for the university. To Wilkes, service was an honourable calling. For Alan, such utilitarian ambitions may have seemed very drab; but they were, in fact, just like the plans of the NPL.
I.J. Good, who visited the NPL with Alan’s lifelong supporter M.H.A. Newman in 1946 to swap ideas about logical design, also despaired of the approach he was taking:
His really great handicap was that he could not forbear to work everything out for himself. He really had to invent the whole of mathematics for himself, and this is very fine and impressive you see, and all this, but I mean, there simply isn’t time for it, and every now and then one would come across this in this computing machine of his. It was a sort of nightmare. It worked in a sense but he didn’t think of all the things that had happened, and the human frailties of the operators, and he would always cling to his thing. He invented an extraordinary series of names for the 32 operations of the alphabet and he would never vary these things and he always put these extraordinary things into all his books and papers and everything, and everybody was supposed to learn these symbols.
Lousy computer
Just-in-time programming was very abstruse, and very fiddly, and not at all suitable for someone whose written work had a habit of being sloppy and inaccurate. Alan Turing needed an assistant. Jim Wilkinson had graduated from Trinity College, Cambridge, as senior wrangler in 1939. He was therefore as good a mathematician as you could get, and he was easily recruited for the project to build an electronic computer, as he later explained to Dr Christopher Evans:
EVANS: How did you first come into contact with Turing? Was it at NPL? WILKINSON: Yes. I’d heard of him at Cambridge but I knew him more as an eccentric than anything else. I didn’t really know very much about his work. But when Goodwin asked me if I would come to NPL I came over here, I chatted with Turing, and I soon became very enthusiastic about the project. Of course I didn’t know very much about the ideas that were in the air at the time then of building digital computers, so this was my first contact. And it seemed to me just the sort of thing that I was looking for to make the numerical solution to partial differential equations a reasonably effective proposition. At that point I decided that I would come to NPL. I came here in May of 1946, and I was assigned half-time to Alan Turing to work with him. We were to work on the logical design of an electronic computer and on the development of what he referred to as ‘tables of instructions’ for solving basic mathematical problems. The other half of my time was to be spent in the desk computing section with Dr Goodwin and Les Fox at Oxford, acquiring the necessary expertise in the numerical solution of mathematical problems.
The delay in getting the ACE built created unwelcome tensions at the NPL, a fact which was still being discussed thirty years later:
EVANS: I’d like to ask you the atmosphere in the other parts of the lab and how other people – senior people – felt about this particular project. Who were the people who were basically ‘for’ research of this kind, and who were the people who were inclined to be a bit dismissive of it?
WILKINSON: If we take the Division itself, most of the real work in Mathematics Division in those days was done by the desk computing section. They were interested in the project and very impressed with Turing as a mathematician, he was a very impressive mathematician. He picked up ideas on numerical analysis. (He was a lousy computer incidentally, one of the worst performers on a desk computer it’s ever been my misfortune to work with.) But his brilliance was immediately obvious and they believed it to be a practical project.
In a later speech, Wilkinson also described his experience of working with Alan Turing:
It was impossible to work ‘half-time’ for a man like Turing and almost from the start the periods spent with the computing section were rather brief. The joint appointment did, however, have its useful aspect. Turing occasionally had days when he was ‘unapproachable’ and at such times it was advisable to exercise discretion. I soon learned to recognize the symptoms and would exercise my right (or, as I usually put it, ‘meet my obligations’) of working in the computing section until the mood passed, which it usually did quite quickly.
Wilkinson was speaking in 1970. His comments on Turing’s output of academic papers – by which professorial productivity is measured – are interesting as a reflection of what people then knew:
I feel bound to say that his published work fails to give an adequate impression of his remarkable versatility as a mathematician. He had only twenty published papers to his credit, written over a peri
od of some twenty years. Remarkable as some of these papers are, this work represents a mere fraction of what he might have done if things had turned out just a little differently. In the first place there were the six years starting from 1939 which he spent at the Foreign Office. He seemed not to have regretted the years he spent there and indeed we formed the impression that this was one of the happiest times of his life. Certainly it was there that he gained his knowledge of electronics and this was probably the decisive factor in his deciding to go to N.P.L. to design an electronic computer rather than returning to Cambridge. Mathematicians are inclined to refer to this period as the ‘wasted years’ but I think he was too broad a scientist to think of it in such terms.
A second factor limiting his output was a marked disinclination to put pen to paper. At school he is reputed to have had little enthusiasm for the ‘English subjects’ and he seemed to find the tedium of publishing a paper even more oppressive than most of us do. For myself I find his style of writing rather refreshing and full of little personal touches which are particularly attractive to someone who knew him. When in the throes of composition he would hammer away on an old typewriter (he was an indifferent typist, to put it charitably) and it was on such occasions that visits to the computing section were particularly advisable.
According to Wilkinson, the decision not to set up a hardware section at the NPL from the outset ‘appeared to me to be a deplorable decision even at the time’. Nothing had yet come of the attempt to outsource the engineering to other governmental bodies. Finally, Sir Charles Darwin brought the hardware development back in-house, and to the disgust of Alan Turing it was proposed to scale it back and have an engineer called Harry D. Huskey build a small-scale ‘test assembly’. Huskey was an American who had worked on the ENIAC, and might therefore be assumed to know his stuff. The test assembly – what these days might be called a ‘proof of concept’ – would be fit for testing the engineering and the feasibility of Alan’s programs, but it wouldn’t be up to any serious computing task. From Alan’s viewpoint the test assembly was a waste of effort.
On the one hand the Test Assembly was to be a small computer in its own right, involving much more equipment than was strictly necessary to test the fundamentals of Turing’s design, and yet on the other it fell far short of being the ACE. Probably Turing saw Huskey’s project as diverting effort from his own. According to Wilkinson, Turing ‘tended to ignore the Test Assembly’, simply ‘standing to one side’. Woodger1 described how he ‘was writing a program for [Version H] when Turing came in … looked over my shoulder and said, “What is this? What’s Version H?”. So I said, “It’s Huskey’s.” “WHAT!” … [T]here was a pretty good scene about that.’
In fact, the test assembly would evolve, after more machinations, into a small working machine called the ‘Pilot’ ACE. That, however, would take a few more years.
Robots and scrap iron
With the atmosphere at NPL becoming increasingly tense, and with no machine to test out the trial programs he and Jim Wilkinson were writing, Alan needed something else to do.
Sir Edward V. Appleton, G.B.E., K.C.B., D.Sc., F.R.S.,
Department of Scientific and Industrial Research,
Park House,
24, Rutland Gate,
London, S.W. 7
23rd July, 1947
Dear Appleton
As you know Dr. A. Turing, S.P.S.O.2 , is the mathematician who has designed the theoretical part of our big computing engine. This has now got to the stage of ironmongery, and so for the time the chief work on it is passing into other hands. I have discussed the matter both with Womersley and with Turing, and we are agreed that it would be best that Turing should go off it for a spell. Though I have other work he might do here, I judge that it is not quite suited for him, and that a different action would be better.
He wants to extend his work on the machine still further towards the biological side. I can best describe it by saying that hitherto the machine has been planned for work equivalent to that of the lower parts of the brain, and he wants to see how much a machine can do for the higher ones; for example, could a machine be made that could learn by experience? This will be theoretical work, and better done away from here. The proposal then is that he should be allowed to be away for a year, which he would spend at Cambridge, where he is a fellow of King’s. […]
Yours sincerely,
C.G. DARWIN
A machine that can learn might seem rather far-fetched. But once you have grasped the idea that a computer program is just so much data, it is really not so hard to believe in intelligent machinery, however peculiar the idea might have seemed to an audience listening to an Alan Turing lecture in 1946.
It has been said that computing machines can only carry out the processes that they are instructed to do. This is certainly true in the sense that if they do something other than what they were instructed then they have just made some mistake. It is also true that the intention in constructing these machines in the first instance is to treat them as slaves, giving them only jobs which have been thought out in detail, jobs such that the user of the machine fully understands what in principle is going on all the time. Up till the present machines have only been used in this way. But is it necessary that they should always be used in such a manner? Let us suppose we have set up a machine with certain initial instruction tables, so constructed that these tables might on occasion, if good reason arose, modify those tables. One can imagine that after the machine had been operating for some time, the instructions would have altered out of all recognition, but nevertheless still be such that one would have to admit that the machine was still doing very worthwhile calculations. Possibly it might still be getting results of the type desired when the machine was first set up, but in a much more efficient manner. In such a case one would have to admit that the progress of the machine had not been foreseen when its original instructions were put in. It would be like a pupil who learnt much from his master, but had added much more by his own work. When this happens I feel that one is obliged to regard the machine as showing intelligence.
Accordingly, in September 1947 Alan went back to Cambridge to think about machines and whether, and how, they could think. The plan was that Alan would spend his sabbatical year writing his ‘theoretical work’ on Intelligent Machinery. In fact, by the time he got to King’s the thinking had already been done. Alan’s paper was written up in that September. The operations of computing machines are based on components which react in different ways according to the inputs they receive: arithmetical processes such as adding and dividing are achieved by assembling small sub-groups of components linked together. In his paper on Intelligent Machinery, Alan Turing had begun to consider other implications of assembling groups of tiny decision-making units. What if they were connected together unsystematically – into what you might call ‘unorganised’ machines? The paper goes on to show that unorganised machines can be conditioned – essentially to learn. Alan Turing had invented the idea of the neural network.
The whole idea of ‘intelligent’ machinery was causing sniggers at the numerically focused NPL. ‘Turing is going to infest the countryside with a robot which will live on twigs and scrap iron!’
MAN AS A MACHINE
One way of setting about our task of building a ‘thinking machine’ would be to take a man as a whole and to try to replace all the parts of him by machinery. He would include television cameras, microphones, loudspeakers, wheels and ‘handling servo-mechanisms’ as well as some sort of ‘electronic brain’. This would be a tremendous undertaking of course. The object, if produced by present techniques, would be of immense size, even if the ‘brain’ part were stationary and controlled the body from a distance. In order that the machine should have a chance of finding things out for itself it should be allowed to roam the countryside, and the danger to the ordinary citizen would be serious. Moreover even when the facilities mentioned above were provided, the creature would still have no co
ntact with food, sex, sport and many other things of interest to the human being. Thus although this method is probably the ‘sure’ way of producing a thinking machine it seems to be altogether too slow and impracticable.
Instead we propose to try and see what can be done with a ‘brain’ which is more or less without a body providing, at most, organs of sight, speech, and hearing. We are then faced with the problem of finding suitable branches of thought for the machine to exercise its powers in. The following fields appear to me to have advantages:
(i) Various games, e.g., chess, noughts and crosses, bridge, poker
(ii) The learning of languages
(iii) Translation of languages
(iv) Cryptography
(v) Mathematics.
His boss was not impressed.
Dear Turing
I have received your paper which has been signed and sent out.
I may say that I read it through with some attention and interest, but spent most of the time cursing you for giving me such a perfectly smudgy copy to read. Next time I hope somebody else and not myself with [sic] be the sufferer, but I think the best plan would be to get better carbon paper.
I hope you are enjoying life at Cambridge.
Yours sincerely,
C G Darwin
The facsimile of the paper, helpfully made available online by the present-day NPL, is rather neat – much too well-presented to be one of Alan’s own typing efforts (such as the draft in the King’s College Archive). It’s tempting, therefore, to comment that Sir Charles should have checked his own letter before complaining about typing. Actually the problem was a bit more serious than giving Sir Charles a carbon rather than the top copy. The paper bore no relation to the role of the NPL’s Mathematics Division as the nation’s computing bureau; apparently Sir Charles’s real view was that it was a ‘schoolboy’s essay’. Intelligent Machinery was not to be published for another twenty years.
Prof Page 18