Head , Learning Machines Group Applied Physics Laboratory

SUMMRY Stanford Research Insti tute proposes a program that will ultimately lead to the development of machines to perform reconnaissance mi ssions that are presently considered to require human intelligence. Significant but uncoordinated research studies applicable to this field are being carried out in a number of laboratories. We propose-in a balanced program of experimental , theoretical , and exploratory research-to coordinate the efforts of a large group of specialists wi thin and outside the Insti tute , directed towards the orderly development of theory, technique , and hardware. Several key problem areas have been identified and will serve as the basis of our ini tial work. These include , among others , research in adaptive pattern recognition , in learning systems , and in mul ti-level goal-seeking systems. In addi tion to the necessary explorat ion of these areas by analyses and simulation , it is intended to develop a machine facility to permit operation with models that will embody de-si red concepts , but wi 11 avoid superf luous engineering design. thus concurrently implementing theoretical results , major practical problems will not be overlooked , costs of real solutions may be assessed and the operating models can also serve a major purpose in the generation clarification, and simplification of complex ideas. The machine facili ty will consist primarily of a high-speed data processing system controlling a simple , mobile vehicle equipped wi visual and tactual sensors. It is planned to demonstrate wi th this facility a goal-seeking automaton learning to perform non-trivial missions in a complex and changing environment-missions that would normally be considered to requi re intelligence.