The Relevance of Connectionism to AI: A Representation and Reasoning Perspective

Abstract : In this paper it is argued that not only is connectionism relevant to knowledge representation and reasoning, but it also provides an ideal computational architecture for intelligent systems. To justify this claim, certain critical features that any computational architecture capable of supporting intelligent behavior must possess are identified, and then it is shown that the core features of connectionism correspond exactly to these features. It is also argued that connectionism cannot be viewed merely as an implementation paradigm because its core features influence our conceptions of representation and reasoning in important ways: designing connectionist models for solving complex tasks leads to the identification of constraints on the conceptual structure. The above issues are discussed with reference to two connectionist systems that perform reasoning tasks with extreme efficiency. Keywords: Artificial intelligence.