Symbol Processing Systems, Connectionist Networks, and Generalized Connectionist Networks

Many authors have suggested that SP (symbol processing) and CN (connectionist network) models offer radically, or even fundamentally, different paradigms for modeling intelligent behavior (see Schneider, 1987) and the design of intelligent systems. Others have argued that CN models have little to contribute to our efforts to understand intelligence (Fodor & Pylyshyn, 1988). A critical examination of the popular characterizations of SP and CN models suggests that neither of these extreme positions is justified. There are many advantages to be gained by a synthesis of the best of both SP and CN approaches in the design of intelligent systems. The Generalized connectionist networks (GCN) (alternately called generalized neuromorphic systems (GNS)) introduced in this paper provide a framework for such a synthesis. 1. Symbol Processing Architectures for Intelligent Systems The symbol processing approach to the design of intelligent systems was summarized by Newell (1980) and Newell & Simon (1972) in terms of what they called the physical symbol systems and by Fodor (1976) in terms of what he called the language of thought. In this framework, perception and cognition are tantamount to acquiring and manipulating symbolic representations. It is suggested that the symbol structures that constitute such representations have a counterpart in the physical structure of the brain and/or the brain’s internal states. Models of intelligent systems developed within this framework typically are (but do not have to be) based on the von Neumann serial stored program model of computation. Popular interpretations of this definition are often overly restrictive, and appear to exclude (for no good reason) systems that perceive, learn, and reason with non-symbolic (e.g., iconic or analogic) representations, or using numerically-encoded probabilistic or fuzzy inference structures. The following sections critically examine popular conceptions and major strengths and weaknesses of SP models of intelligent systems. hhhhhhhhhhhhhhhhhhhhhhhhhhhhh This is a draft of a paper that is currently under review. Comments and suggestions for improvement will be appreciated. This work was partially supported by the University of Wisconsin-Madison Graduate School and the Iowa State University College of Liberal Arts and Sciences.