eas@utcsrgv.UUCP (Ann Struthers) (11/19/84)
THEORETICAL ASPECTS SEMINAR Tuesday, November 29, 1984 3:00 PM Sandford Fleming Building 1105 Professor Hector Levesque Department of Computer Science University of Toronto "Tractability and Belief" The current design practice in Artificial Intelligence is to base the behaviour of complex systems on inference: some propositions that a system is to believe are represented explicitly, and their implications are calculated as neded to fill in the details. For sufficiently expressive representation languages, however, it is impossible to decide in an acceptable amount of time just what is implicit in a set of propositions. This talk will attempt to motivate some compromises that will allow the management of what a system believes to be kept tractable. Although some prelimiary results will be presented, the emphasis will be on research directions.