Computation and Intelligence in Problem Solving

The concept of computation, as well as the related concepts algorithm, Turing machine, computability, and computational complexity, correspond to a specific mode of using computers to solve problems. This computational mode assumes the sufficiency of knowledge and resources with respect to the problem to be solved. From the view point of Artificial Intelligence, an intelligent mode of problem solving is introduced, where the problem solving process cannot been seen as computation anymore. A system working in this mode is briefly described. Finally, these two modes of problem solving are compared. “Computation” is a fundamental concept in computer science. Artificial Intelligence (AI) is usually taken as a branch of computer science, and it has also, to a large extent, inherited the theoretical heritage associated to the concept of computation. In this paper we argue that it is improper to treat intelligent problem solving as computation, though this conclusion does not make it impossible to build intelligent computer systems. Actually it is the contrary: we believe that real AI can be realized only when we go beyond the concept of computation. An intelligent reasoning system, NARS, is introduced, whose problem solving activities cannot be referred to as computation, though the system is still implemented in an ordinary computer. 1 Problem Solving as Computation Turing machine is a mathematical model of computer. A Turing machine M has a finite number of states, and among them there is one initial state