To appear in Proceedings of Turing 1990 Colloquium. It is hard to define a sharp boundary between computational and non-computational processes if the computational processes are supposed to provide a non-circular explanation for cognitive abilities. The only well-defined notion of "computation" -- what can be done by a Turing machine or a mathematically equivalent mechanism -- is not adequate for the key role in explaining the nature of mental processes, because it is both too general and too restrictive. Instead we need a new theory-based taxonomy of mechanisms and corresponding process types. Computations in the Turing-equivalent sense are most likely but a sub-class of the processes required for intelligence. The more general analysis of mechanisms for intelligent systems starts with the notion of a system with independently variable, causally interacting sub-states that have different causal roles (e.g. some belief-like). For certain architectures, the sub-states will have a semantic interpretation for the system. The relevant concept of semantics is defined in terms of a kind of Tarski-like structural correspondence, with indeterminacy reduced by a semantics based on environmental states that can close causal loops involving behaviour generated by desire-like and belief-like sub-states.
This paper is not available online