Artificial Intelligence, Cognitive Science, and Explanation-Based Learning (talk recording available)
Dr. Gerald DeJong
Tuesday, April 21, 2009, 01:00pm - 02:00pm
University of Illinois at Urbana-Champaign, Computer Science Department
Artificial intelligence (AI) has always been an exceptionally diverse field. However, one clear trend has emerged over the last several decades. In mainstream AI, machine learning has steadily grown in importance. >From a constituent subdivision, co-equal with planning, vision, natural language processing, robotics, and so on, machine learning has become the engine underlying these subdivisions and driving many of AI's recent advances. We will examine this trend as a natural maturing of the discipline.
Concomitantly, mainstream AI has gradually but steadily grown apart from cognitive science. With technological maturity one expects such departures; airplanes fly better without flapping their wings. But we will argue that there is limited justification for this position. By examining recent cognitive developmental results, we find evidence that even very young humans employ algorithms that in many respects are far more advanced than current machine learning algorithms.
We will examine a new statistical version of explanation-based learning (EBL) as one direction that may bridge the widening gap between AI and cognitive science. This approach holds promise for addressing current issues in machine learning (including relational learning, structural learning, and feature construction) as well as for suggesting new directions for developmental investigations.
Note: If you would like to receive email announcements about the colloquium series, please contact the Business Office to have your name added to our announce lists at firstname.lastname@example.org.