Separability Hypothesis

In this paper we argue for a Separability Hypothesis as a working hypothesis for AI. There is no a priori equivalence of the terms “mind” and “cognition.” Phenomenologically, a mental state has a cognitive content as well as subjectivity and emotions. The Separability Hypothesis asserts that an architecture for intelligence is a homomorphism of the architecture for mind. None of the technical problems that AI is working on today require machines which have anything other than a cognitive content to them. We discuss how this Hypothesis protects AI from a number of its critics, and lets the field get on with its technical concerns. Is Intelligence Separable from other Mental Phenomena? In both cognitive science and AI, the terms “mind” and “cognition” (or “intelligence”) are generally presumed to refer to one and the same thing. This is not an explicit doctrine of the fields in question, but more of a tacit assumption. Thus, both people within and outside AI take the questions, “Can machines have a mind?” and “Can machines be intelligent?” to be equivalent questions. A point of this paper is to argue that there is no a priori identity of these things. Both methodologically and philosophically it 1 Parts of this paper originally appeared in B. Chandrasekaran and S.G. Josephson, "Architecture of Intelligence: The Problem and Current Approaches to Solutions," Current Science, Vol. 64, No. 6, March 1993, pp. 366 380, and in Artificial Intelligence and Neural Networks: Steps Toward Principled Integration, V. Honavar and L. Uhr, editors, Academic Press, pp. 21-50, 1994. The current paper is a much revised version of the earlier discussions