From Big Medical Encyclopedia

ARTIFICIAL INTELLIGENCE — the section of cybernetics studying possibilities of modeling of intellectual activity of the person on electronic computers. Modeling at the same time is understood as creation of such programs which would allow to solve the problems demanding from the person of an intuition, experience, cogitative activity on the COMPUTER.

Any machine program however difficult it was, represents the accurate description of a certain sequence of elementary actions. If the computer, working according to the program, successfully solves any intellectual problem (e.g., well plays chess or correctly classifies difficult images), then this program can be considered as the hypothesis explaining the mechanism of the corresponding intellectual process or as one of possible models of this process.

Works in the area I. and. are important as for studying of process thinking (see), and for automation of various areas of human activity demanding from the person of an intuition and experience.

In the 70th 20 century programs for control of robots, for drawing up production schedules, information search in directory systems, the automatic analysis of images, etc. are developed. In process of accumulation of knowledge of process of thinking and improvement of methods of work with computers this list will gradually extend.

Researches in the area I. and. develop in two directions: clarification of mechanisms of intellectual activity and creation of as much as possible powerful methods allowing to solve successfully on computers complex intellectual challenges.

In researches of the first direction subject to modeling is not the net result of intellectual process, and process. E.g., during the development of programs for the proof of mathematical theorems the main objective (from the point of view of this direction) consists not in that the car could prove as much as possible difficult theorems, and in that the machine proof reminded reasonings of the mathematician proving the same theorem as much as possible. The Universal Solver of Tasks program developed an amer belongs to number of the most interesting models created within this scientific direction. scientists Newell and Simon (A. Newell, H. A. Simon) in the late fifties. In this work it is shown that for the solution of quite wide class of simple logical and mathematical tasks examinees use the same mechanism — achievement of the set purpose by generation and implementation of the sequence of simpler «local goals». The specified mechanism it was possible to realize in the form of the program for the computer. There is also a number of the programs modeling such elements of intellectual activity as storing of words, formation of concepts, decision-making. Experiments with these programs showed rather good coincidence to results corresponding psikhol, experiments. However on this way it was not succeeded to promote further understanding of some elementary mechanisms, to-rymi process of thinking and which in itself are insufficient for the solution of really complex challenges far is not exhausted.

Supporters of the second direction, without demanding obligatory similarity of computation process to reasonings of the person, aim during the development of programs as much as possible to use specific features of computers. The most effective methods and programs managed to be developed for some classes of so-called discrete tasks, i.e. tasks for which solution in principle it is enough to consider final (though, as a rule, astronomically big) number of options. Games like chess, the proof of mathematical theorems, problems of medical diagnosis belong to such tasks (see. Diagnosis machine ), drawing up production schedules. The problem of the solution of these tasks on computers comes down to development of the methods of selective search allowing the car to look through only rather small number of the most «perspective» options in the course of the solution of a task. The program for a game in checkers having ability to study, i.e. to accumulate experience with increase in number of the played parties can be an example. This program in the beginning losing to beginners after training won against one of the best players of the USA.

Bibliography: Arbib M. A metaphorical brain, the lane with English, M., 1976; Computers and thinking, under the editorship of E. Feygenbaum and J. Feldman, lane with English, M., 1967, bibliogr.; Sleygl Dzh. P. Artificial intelligence, Approach on the basis of heuristic programming, the lane with English, M., 1973; Simon H. And. The structure of ill structured problems, Artificial intelligence, v. 4, p. 181, 1973, bibliogr.

A. L. Chernyavsky.