A Distributed Activity Processing System for AI

Abstract

This paper is concerned with the development of a, primarily hardware, knowledge system with common sense like capabilities which eventually may prove more useful than 'logical deduction' systems. To this end the following section makes some observations about human intelligent behaviour and reformulates these in 'machine terms'. After this, two themes are developed (i) a simple set of control primitives for a distributed system, and (ii) the development of "priority" and "relevance" schemes to the above stated goal. What has been omitted from this paper is a detailed discussion of processor-processor deadly embraces and general network message saturation (although these are touched upon). This is a deliberate omission on the grounds that space precludes detailed discussion of this. The authors will be commenting else where on these aspects.