9. Newell, A., Shaw, J. C., and Simon, H. A. Empirical explorations of the logic theory machine: a case study in heuristic. Proceedings of the Western Joint Computer Conference, 1957. 10. Whitehead, A. N. and Russell, B. Principia Mathematica. Vol. I. 11. Gelernter, H. Realization of a geometry theorem proving machine. In Information Processing. Paris: UNESCO, 1960. 12. Garfinkle, S. Heuristic solution of first year algebra problems. Institute of Industrial Relations, Univ. of California, 1960, dittoed. 13. Reitman, W. R. Notes on conversations with Soviet investigators of thinking and problem solving. Carnegie Institute of Technology, 1960, dittoed. 14. Tonge, F. M. Summary of a heuristic line balancing procedure. The RAND Corporation, 1959, paper P-1799. 15. Gere, W. S., Jr. Heuristics in job shop scheduling. Carnegie Institute of Technology, 1960, dittoed. 16. Clarkson, G. P. Thesis proposal. Carnegie Institute of Technology, 1960, dittoed. 17. Feldman, J. Analysis of predictive behavior in a two-choice situation. 1959. Ph.D. dissertation, Carnegie Institute of Technology, 18. Feigenbaum, E. An information processing theory of verbal learning. The RAND Corporation, 1959, P-1817. 19. Feigenbaum, E. and Simon, H. A. stimuli. Psychological Review, 20. Gregg, L. W. and Simon, H. A. Comment: the distinctiveness of in press. Paper to be read at meetings of the Psychonomic Society, Chicago, 1960. 21. Laughery, K. Thesis proposal. Carnegie Institute of Technology, 1960, dittoed. 22. Hovland, C. I. and Hunt, E. B. Computer simulation of concept attainment. Behavioral Science, 1960. 23. Newell, A., Shaw, J. C. and Simon, H. A. Chess-playing programs and the problem of complexity. IBM Journal of Research and Development, 1958. 24. Lindsay, R. K. The reading machine problem. CIP paper no. 33. Carnegie Institute of Technology, 1960, dittoed. 25. McCarthy, J. Programs with common sense. In Symposium on Mechani zation of Thought Processes. Vol. I. London: Her Majesty's Stationery Office, 1959. 26. Minsky, M. L. Some methods of artificial intelligence and heuristic programming. In Symposium on Mechanization of Thought Processes. Vol. I, London: Her Majesty's Stationery Office, 1959. 27. Newell, A., Shaw, J. C., and Simon H. A. Report on a general problemsolving program. In Information Processing. Paris: UNESCO, 1960. 28. Newell, A., Shaw, J. C., and Simon, H. A. The processes of creative thinking. The RAND Corporation, 1958, paper P-1320. 29. Newell, A. et al. Information processing language V manual. The RAND Corporation, 1960, papers P-1897 and P-1918. 30. Reitman, W. R. Heuristic programs, computer simulation, and higher mental processes. Behavioral Science, 1959. 31. Miller, G. A., Galanter, E., and Pribram, K. H. Plans and the Structure of Behavior. New York: Henry Holt and Co., 1960. 32. Holland, J. H. On iterative circuit computers constructed of micro-electric components and systems. Proceedings of the Western Joint Computer Conference, 1960. 33. Newell, A. On programming a highly parallel machine to be an intelligent technician. Proceedings of the Western Joint Computer Conference, 1960. MECHANICAL REALIZATION OF PATTERN RECOGNITION Paul Metzelaar In the group of problems represented by the title "Mechanical Realization of the Higher Functions of Living Systems," the Pattern Recognition problem area is perhaps the most difficult. This paper presents an outline of the work done in this area. The primary emphasis is here on machines designed to allow performance of some useful recognition function, rather than to serve as a structurally realistic model of a biological recognition mechanism. Designers with this functional approach have frequently worked with conventional or heuristic computer programs and have not necessarily attempted to use nerve-net or other biological concepts to implement their designs. Living systems are rather clearly better suited to the recognition of patterns than are our present day machines. A child can distinguish its parents from other people long before it can analyze anything in a logical way. With our machines, just the reverse is true--they can very rapidly and accurately perform arithmetic and logical operations, but have not yet done comparably well at distinguishing people or other entities in the outside world. In the performance of the great majority of pattern recognition tasks, our machines are still many times slower and less versatile than are people. One area where a comprehensive set of functions ordinarily performed by living systems are needed, is in deep space exploration. It seems clear that both the first and the farthest exploration of space will continue to be made by unmanned exploratory probes. This is not because the intelligence of human beings are not needed, but rather because the cost of supporting them and the hazard of the voyage become too great. To nevertheless realize the functions that humans should perform on these missions, the application of biological principles to electronic mechanisms or bionics seems most appropriate. The space probe as a whole, must in many ways perform like an individual biological organism. Perhaps a first requirement is that our space probe be able to contribute to its own survival on its mission or to contribute to the chance of survival of subsequent missions. For instance, we may require that our explorer be able to recognize where it is in order to make corrections to its trajectory. Such a navigational capability could, for instance, enable it to explore the close neighborhood of a planet or moon without crashing. With its various cameras and other sensing devices, our probes will be taking photographs and measuring the radio frequencies, optical and ionizing radiations present. Unusual incidence of gas clouds, belts of dust particles and meteorites are of course also of great importance to its survival. As any biological organism, our space probe must conserve its energy. It can only afford to turn on the engines to make necessary flight path corrections. In communicating the results of its measurements back to earth, it should not continuously keep its radio transmitters turned on, it should transmit only when there is information worth sending. Periods of measurements and information transmission will alternate with inactivity in analogy with the waking and sleeping of biological organisms. The space probe may travel so far that direct radio communications can no longer be received. Rather than to measure and record all data received on a long mission, we should like our explorer selectively to remember what significant patterns of data have been encountered. Our space vehicles must be able to cope with their internal environment as well as the external environment. For many of the space probes, there will be no maintenance men close by to effect repairs. The job of keeping a complex exploring mechanism operating for years or even several months without a repairman will require a degree of component and system reliability that is at present beyond the state of the art. For instance, to put a space probe up with a small airborne digital computer of conventional design as the control element, it has been estimated that we need a one-hundred-fold increase in reliability of the best present components to insure against failure. It has been estimated that even if the one-hundred fold improvement in reliability were achieved, a one-year trip would still have a 10% failure probability due to component unreliability alone. Evidently machines are needed that are especially designed for the mission in two respects: they must be able to perform their external functions, and yet be able periodically to check their internal functioning to check the good health of the components and circuits used. The diagnostic function should be augmented by the capability to bypass any component in which patterns of failure are recognized. |