Technical Report Number
Computing Methodologies, Software, Computer Applications
The goal of Artificial Intelligence, broadly defined, is to understand and engineer intelligent systems. This entails building theories and models of embodied minds and brains -- both natural as well as artificial. The advent of digital computers and the parallel development of the theory of computation since the 1950s provided a new set of tools with which to approach this problem -- through analysis, design, and evaluation of computers and programs that exhibit aspects of intelligent behavior -- such as the ability to recognize and classify patterns; to reason from premises to logical conclusions; and to learn from experience. The early years of artificial intelligence saw some people writing programs that they executed on serial stored--program computers (e.g., Newell, Shaw and Simon, 1963; Feigenbaum, 1963); Others (e.g., Rashevsky, 1960; McCulloch and Pitts, 1943; Selfridge and Neisser, 1963; Uhr and Vossler, 1963) worked on more or less precise specifications of more parallel, brain--like networks of simple processors (reminiscent of today's connectionist networks) for modelling minds/brains; and a few took the middle ground (Uhr, 1973; Holland, 1975; Minsky, 1963; Arbib, 1972; Grossberg, 1982; Klir, 1985). It is often suggested that two major approaches have emerged -- symbolic artificial intelligence (SAI) and artificial neural networks or connectionist networks (CN) and some (Norman, 1986; Schneider, 1987) have even suggested that they are fundamentally and perhaps irreconcilably different. Others have argued that CN models have little to contribute to our efforts to understand cognitive processes (Fodor and Pylyshyn, 1988). A critical examination of the popular conceptions of SAI and CN models suggests that neither of these extreme positions is justified (Boden, 1994; Honavar and Uhr, 1990a; Honavar, 1994b; Uhr and Honavar, 1994). Recent attempts at reconciling SAI and CN approaches to modelling cognition and engineering intelligent systems (Honavar and Uhr, 1994; Sun and Bookman, 1994; Levine and Aparicioiv, 1994; Goonatilake and Khebbal, 1994; Medsker, 1994) are strongly suggestive of the potential benefits of exploring computational models that judiciously integrate aspects of both. The rich and interesting space of designs that combine concepts, constructs, techniques and technologies drawn from both SAI and CN invite systematic theoretical as well as experimental exploration in the context of a broad range of problems in perception, knowledge representation and inference, robotics, language, and learning, and ultimately, integrated systems that display what might be considered human--like general intelligence. This chapter examines how today's CN models can be extended to provide a framework for such an exploration.