Exclusive: John von Neumann’s letters to Norbert Wiener on computational machinery (1945)

For a special treat to all my readers I am putting these previously difficult to access letters online for the first time.

In this exchange I’ve selected two from Dr. von Neumann that I found to be especially interesting, and which I was able to secure special permission to reproduce from the copyright holder and the American Mathematical Society.

Please make the appropriate credits if you wish to further copy.

Letter 1 “

Professor Norbert Wiener
November 20, 1945
Mathematics Department
Massachusetts Institute of Technology
Cambridge 39, Massachusetts

Dear Norbert:

The negotiations at Princeton have come to a conclusion, with the result that the Institute of Advanced Study and the Radio Corporation of America (whose research laboratory is, as you know, in Princeton), with the cooperation of Princeton University, have decided to undertake a joint high-speed, automatic, electronic computer development. While this is to be a community effort, it is agreed between all parties that the completed computer is to be located at the Institute, and used exclusively as a research tool. I have been offered the over-all direction of this

As far as I can see, the project is adequately financed and has the necessary guarantees of manpower, availability of the RCA’s experience in various pertinent fields, etc.

Since I have always urged such an enterprize on the Institute for Advanced Study authorities, and insisted that it is of great importance in several vital respects, I feel that it is now my prime responsibility to see it through here.

I need not tell you how much I regret that this means that I cannot join you at Tech (MIT), especially since I am sure that without the decisive encouragement that I received from you and from the Tech authorities I would have hardly had the perseverance and the strength of conviction which are the minimum requirements in this project. On the other hand, I hope that we shall work together in the field just the same. I think that we should discuss the modalities as quickly as possible.

I am staying in Princeton continuously until November 30. Then I am going
for 2.5 – 3 weeks to Los Alamos. This trip, plus the Christmas vacation eliminate the month of December. Could you find the time to visit here before the end of this month, say between November 26 (Monday) and 29 (Thursday), both inclusive? I hope that you can do this and will accept our invitation to come as a guest of the Institute.

Hoping to see you soon, and very much looking forward to it,

Yours as ever,

J. von Neumann

Letter 2 “

November 29, 1946

Dear Norbert:

This letter represents an effort to do better than I estimated in my letter of November 25, in which I proposed that we might get together for the afternoon or evening of December 4 in Cambridge, and indicated only somewhat vaguely what the subject was that I would like to discuss with you. I am now trying to give you a more detailed advance notice, hoping that this will make our discussion on December 4 more specific.

Our thoughts – I mean yours and Pitts’ and mine- were so far mainly focused on the subject of neurology, and more specifically on the human nervous system and there primarily on the central nervous system. Thus, in trying to understand the function of automata and the general principles governing them, we selected for prompt action the most complicated object under the sun – literally. In spite of its formidable complexity this subject has yielded very interesting information under the pressure of the efforts of Pitts and McCulloch, Pitts, Wiener and Rosenblueth. Our thinking – or at any rate mine – on the entire subject of automata would be much more muddled than it is, if these extremely bold efforts – with which I would like to put on one par the very un-neurological thesis of R. Turing – had not been made. Yet, I think that these successes should not blind us to the difficulties of the subject, difficulties, which, I think, stand out now just as – if not more-forbiddingly as ever.

The difficulties are almost too obvious to mention: They reside in the exceptional complexity of the human nervous system, and indeed of any nervous system. What seems worth emphasizing to me is, however, that after the great positive contribution of Turing – cum – Pitts – and-MeCulloch is assimilated, the situation is rather worse than better than before. Indeed, these authors have demonstrated in absolute and hopeless generality, that anything and everything Brouwerian can
be done by an appropriate mechanism, and specifically by a neural mechanism and that even one, definite mechanism can be “universal”. Inverting the argument: 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. I know that this was well known to Pitts, that the ‘nothing’ is not wholly fair, and that it should be taken with an appropriate dose of salt, but I think that you will feel with me the type of frustration that I am trying to ress
(H. N. Russell used to say, or to quote, that if the astrophysicist found a general theory uniformly corroborated, his exclamation should be “Foiled again” Since no experimenta crucis would emerge.) After these devastatingly general and positive results one is therefore thrown back on microwork and cytology – where one might have remained in the first place. (This “remaining there” is, of course, highly figurative in my case, who have never been there.) Yet, when we are in that held the complexity of the subject is overawing. To understand the brain with neurological
methods seems to me about as hopeful as to want to understand the ENIAC with no instrument at one’s disposal that is smaller than about 2 feet across its critical organs, with no methods of intervention more delicate than playing with a fire hose (although one might ill it with kerosene or nitroglycerine instead of water) or dropping cobblestones into the circuit. Besides the system is not even purely digital (i.e. neural): It is intimately connected to a very complex analogy (i.e. humoral or hormnonal) system, and, almost every feedback loop goes through both sectors, if not through the “outside world (i.e. the world outside the epidermis or within the digestive system) as well. And it contains, even in its digital part, a million times more units than the ENIAC. And our intellectual possibilities relatively to it are about as good as some bodies vis-a-vis the ENIAC, if he has never heard of any part of arithmetic. It is true that we know a little about the syndromes of a few selected breakdowns – but that is not much.

My description is intentionally exaggerated and belittling, but don’t you think that there is an element of truth in it?

Next: If we go to lower organisms from man with 10^10 neurons to ants with 10^6 or to some sub-animal with, say, 10^2 neurons – we lose nearly as much as we gain. As the digital (neural) part simplifies, the analogy (humoral) part gets less accessible, the typical malfunctions less known, the subject less articulate, and our possibilities of communicating with it poorer and poorer in content.

Further: I doubt that the “Gestalt” theory, or anybodies verbal theory will help any. The central nervous system is complicated, and therefore its attributes and characteristics have every right to be complicated. Let not our facile familiarity with it, through the medium of the subjective consciousness, fool us into illusions in this respect.

What are we then to do? I would not have indulged in such a negative tirade if I did not believe that I see an alternative. In fact, I have felt all these doubts for the better part of a year now, and I did not talk about them because I had no idea as to what one might say in a positive direction.

I think now that there is something positive to be said, and I would like to indicate in which direction I see it.

I feel that we have to turn to simpler systems. It is a fallacy, if one argues, that because the neuron is a cell (indeed part of its individual insulating wrapping is multicellular), we must consider multicellular organisms only. The cell is clearly an excellent “standard component”, highly flexible and suited to diferentiation in form and in function, and the higher organisms use it freely. But its self-reproductivity indicates that it has in itself some of the decisive attributes of the integrated organisms- and some cells (e.g. the leukocytes) are self-contained, complete beings. This in itself should make one suspicious in selecting the cells as the basic “undefined” concepts of an axiomatism. To be more par terre: Consider, in any field of technology, the state of affairs which is characterized by the development of highly complex “standard components”, which are at the same time individualized, well sited to mass production, and (in spite of their “standard” character) well suited to purposive differentiation. This is clearly a late, highly developed style, and not the ideal one for a first approach of an outsider to the subject, for an effort towards understanding. For the purpose of understanding the subject, it is much better to study an earlier phase of its evolution, preceding the development of this high standardization – with differentiation. I.e. to study a phase in which these “elegant components do not yet appear. This is especially true, if there is reason to suspect already in that archaic stage mechanisms (or organisms) which exhibit the most specific traits of the simplest representatives of the above mentioned “late” stage.

Now the less-than-cellular organisms of the virus or bacteriophage type do possess the decisive traits of any living organism: 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 possibly the greatest step that may at all be required.

I would, however, put on “true” understanding the most stringent interpretation possible: That is, understanding the organism in the exacting sense in which one may want to understand a detailed drawing of a machine – i.e. finding out where every individual nut and bolt is located, etc.

It seems to me that this is not at all hopeless. A typical bacteriophage, which can be multiplied at will (and hence “counted” – by the colonies it forms on a suitable substrate- as reliably as elementary particles can be “counted” by a Geiger-counter) is a phage that is parasitic, I think, on the Bacillus Coli. It has been extensively worked with, e.g. by Delbrueck at Vanderbilt. It is definitely an animal, with something like a head and a tail. Its dimensions are I think, ca. 60 microns. 25 microns x 25 microns, i.e. its volume is 60 x 25 x 25 x 10^-21 cm^3 = 3.7 x 10^-17 cm^3.
The density may be taken to be 1, hence its mass is about 3.7 x 10 grams. i.e. the same as about 2.5 x 10^7 H atoms. Since the average chemical composition of these things is usually about one C or N or O per one or two H, the average atomic weight of its constituents is about 6. Hence the number of atoms in it is about 4 x 10^6. Furthermore, it is known from the behavior of physiological membranes, that they are monomolecular – or oligomolecular – Langmuir-layers, which exercise their function in a highly mechanical way. E.g. the so-called “active permeability”: The peculiar ability to “permit” ions to pass through the membrane against an electrical field – an activity which clearly must, and demonstrably does, require an energy supply from the metabolism – and which is therefore better described as “pushing the ions across” than as “permitting them to pass”. I understand that here an ion simply gets seized by the opposite-polarity end of one of the rare (charged) radicals in the membrane, which then turns around and deposits the ion on the other side of the membrane. Very similar things can be said about the functioning of the “phosphate bond”, which seems to be the main physiological device for localized, short time energy storage – i.e. the equivalent of a spring. Thus one can really talk of “mechanical elements”, each of which may comprise 10 atoms or more. Thus the organism in question consists of six million atoms, but probably only of a few hundred thousand “mechanical elements”. I suppose (without having done it) that if one counted rigorously the number of “elements” in a locomotive, one might also wind up in the high ten thousands. Consequently this is a degree of complexity which is not necessarily beyond human endurance.

The question remains: Even if the complexity of the organisms of molecular weight 10^10 is not too much for us, do we have the observational means to ascertain all the facts? Or to be more lenient: If we do not possess such means now, can we at least conceive them, and could they be acquired by developments …

Letter 2 will be continued in my next post.