Can computers be intelligent? Artificial intelligence and conceptual change

What is the extension of the predicate “intelligent”? This troublesome question is complicated by accomplishments in the field of artificial intelligence (AI), which collectively raise the question of whether intelligence can legitimately-and not just metaphorically-be ascribed to computers. Indeed, the problem is evaded by Marvin Minsky’s often used definition of A1 as “. . . the science of making machines do things that would require intelligence if done by men.”” We believe that developments in A1 are forcing a restructuring of the denotational span of the term “intelligent.” We shall argue against the claim that intelligence cannot be ascribed to computers because computers do not isomorphically emulate human intelligence; we shall also argue against ascribing intelligence to computers by recasting the definition of human intelligence into strictly information processing terms. Rather, we shall argue that one way of analyzing this question of computer intelligence is to treat the problem as a problem of language. More specifically, using philosophical theories of language and conceptual change, advanced by philosophers Hilary Putnam and Philip Kitcher, we shall argue that the extension of the term “intelligent” (or “intelligence”) can justifiably and systematically be made more general. We believe that our approach, grounded in behavioral considerations and the philosophy of language, represents a significant departure from traditional approaches, grounded in philosophy of mind and the new, interdisciplinary field of cognitive science. Indeed, before beginning our behavioraVlinguistic approach, we shall spend some time reviewing and criticizing a number of these traditional arguments. Our purpose for doing this is two-fold: (1) we believe the time is ripe for the collation of different arguments and the presentation of a general overview, in one place, of the work that has been done so far, although we do not claim