XVI. ARTIFICIAL INTELLIGENCE

The purpose of our work is to investigate ways of making machines solve problems that are usually considered to require intelligence. Our procedure is to attack the problems by programming a computer to deal directly with the necessary abstractions, rather than by simulating hypothetical physiological structures. When a method for solving a problem is not known, searches over spaces of potential solutions of the prob-lem, or of parts of the problem, are necessary. The space of potential solutions of interesting problems is ordinarily so enormous that it is necessary to devise heuristic methods (1) to replace the searching of this space by a hierarchy of searches over sim-pler spaces. The major difficulty, at present, is the excessive length of time required for building machinery or even for writing programs to test heuristic procedures. For this reason, the major part of our effort is going into the development of ways of com-municating with a computer more effectively than we can now communicate. This work has two aspects: development of a system for instructing the computer in declarative, as well as imperative, sentences, called the advice taker (2), and development of a programming language called LISP (3) for manipulating symbolic expressions that wil