Al: Inventing WILLIAM Znstztute { bill-clancey@ a New Kind of Machine J. CLANCEY for Research u-l erg) on Leurrung, 2550 Hanover Street, Palo Alto, CA 9430-I During the 1980 s, related hypotheses about cognition emerged in computer science and neuroscience: (1) AI researchers began to realize that knowledge engineering the process of codifying human knowledge in expert systems makes a solid contribution to computer modeling, but such programs neither replicate nor explain the full capability of human intelligence. (2) Exclusively symbolic theories of cognition inadequately distinguish between an observer s description of patterns in intelligent behavior and the underlying mechanism that relates sensory and motor processes in biological systems. Symbolic theories view reasoning in words and other symbols as both the product and engine of thought. representations such as (3) Word-based rules and scripts enable knowledgebased programs to effectively model and control complex systems in the world. But a mechanism built only out of networks of words and their relations is limited in its ability to conceive and coordinate behavior in multiple sensory modalities in time. Expert systems cannot dance or smell a change in the weather. approach to engineer(4) A means-ends ing an intelligent machine
/lp/association-for-computing-machinery/ai-inventing-a-new-kind-of-machine-WA4lR4aUpe