Date: Wed, 20 Nov 1996 19:16:33 GMT Server: Apache/1.0.3 Content-type: text/html Content-length: 1995 Last-modified: Thu, 14 Mar 1996 20:31:24 GMT
The Metacat project is an attempt to computationally model certain key aspects of human cognition. It has its foundations in an earlier project called Copycat, a computer model of high-level perception and analogy-making. The central theme underlying Copycat is the idea of nondeterministic, stochastic processing distributed among a large number of small computational agents, which work on different aspects of an analogy problem simultaneously, at different speeds, thereby achieving a kind of differential parallelism. All processing occurs through the collective actions of many agents working together, without any higher-level, executive process directing the overall course of events. Thus, Copycat lies firmly within the paradigm of emergent computation. At the same time, however, it incorporates many ideas from the more traditional paradigm of symbolic AI, inhabiting a kind of middle ground between these two opposites. Current research is concerned with extending the model in a way that will allow it to create much richer representations of the analogies it makes. This involves the idea of `self-watching' -- the ability to perceive and remember patterns that occur in its own processing as it solves analogy problems. Based on this ability, Metacat will be able to understand and explain its answers in a way that Copycat cannot, and will eventually be able to perceive analogies between analogies.
Associated Faculty: Douglas Hofstadter
Associated Graduate Students: Jim Marshall
Affiliated Projects: Letter Spirit (Douglas Hofstadter & John Rehling)
Support:
This research is supported by funding for the Center for Research on Concepts and Cognition (CRCC) provided by the IU College of Arts and Sciences.