ACKNOWLEDGMENT This paper is based on a doctoral dissertation submitted to the Department of Electrical Engineering, Polytechnic Institute of Brooklyn, Brooklyn, New York. The work of this research was done at the David Sarnoff Research Center of the RCA Laboratories, Princeton, New Jersey. The author is grateful to Professor E. Mishkin of the Polytechnic Institute of Brooklyn for his interest and encouragement throughout this research. The author also wishes to express appreciation of his colleagues S. Amarel, J. Sklansky, and R. O. Winder of the RCA Laboratories for their helpful suggestions and comments. 1. 2. 3. REFERENCES J. Von Neumann, "Probabilistic Logics and the Synthesis J. Von Neumann, "Fixed and Random Logical Patterns and The W. S. McCulloch, "Agatha Tyche of Nervous Nets Lucky Reckoners, pp. 611-625, Mechanization of Thought Process, Vol. II, Her Majesty's Stationary Office, London, 1959, (presented at the symposium held in November, 1958 at the National Physical Laboratory, Teddington, England). 4. E. F. Moore and C. E. Shannon, "Reliable Circuits Using Less Reliable Relays," B.S.T.J. Monograph 2696, (also published in the Journal of the Franklin Institute, Vol. 262, Part I, pp. 191-208, Sept., 1956, Vol. 262, Part II, pp. 281-297, Oct., 1956. WHAT GOOD ARE ARTIFICIAL NEURONS? W. A. van Bergeijk L.D. Harman Bell Telephone Laboratories INTRODUCTION As recently as 10 years ago, the difference in outlook and interests between biologists concerned with the nervous system and engineers was as great as it ever had been. The biologists were trying to understand nervous systems through anatomical, physiological, pathological and psychological techniques. Their interests lay mainly in understanding for its own sake, in arriving at theories describing nervous systems. Applications, if any, were largely in the area dealing with pathology. The engineers, on the other hand, were inventing artifacts for definite purposes; they were busy designing the very latest advances in the art, among them digital computers. Their interest was mainly in making machines that would accomplish a given goal; the means by which this could be done have always been the engineer's bread and butter. By contrast, as we have noted, the biologist was interested in finding out how an existing piece of "machinery," the living nervous system, was in fact operating, and what it really could do. The attitudes of the two camps were, in a very real sense, diametrically opposed. The respective interests had not the slightest overlap, and the methods and philosophies of the two disciplines were as different as one could imagine. Then Norbert Wiener wrote "Cybernetics," Shannon and Weaver wrote "The Mathematical Theory of Communication" and the lid was off Pandora's box. The psychologists suddenly discovered that their stimulus patterns could be described in bits, and that a rat or human being could be considered as a "black box," an "information source" or a "channel." The physiologists were awed by the realization that the all-or-none nervous impulse they had known for a couple of decades was a "digital" pulse, and that the nervous system was really a miniaturized digital computer, "handling" all sorts of "data." The engineers could not help being infected with the excitement in the biological camp. The mutual reinforcement (a positive feedback, if you wish) generated a wildly oscillating wave of activity and projects which has, as of the present time, shown no detectable damping. Out of the inevitable flood of nonsense resulting from such a situation, some useful ideas and theories have emerged. At least, so it seems. Actually, we are still too much in the thick of the riot to be able to segregate reliably sense from nonsense, the useful from the useless, and sound speculation from fanciful project-making. |