What Kind of Computation is Human Cognition? A Brief History of Thought (Episode 1/2)

Since the naming of the field in 1956, AI has been dominated first by symbolic rule-based models, then early-generation neural (or “connectionist”) models, then probabilistic models operating over symbolic structures, and now deep learning models. The next generation of models, many of us suspect, will be neurosymbolic systems. Projecting the future of the field is aided by understanding the forces that drive this swing of the pendulum between symbolic and neural representation, forces that have their origins in antiquity. In this 2-part series, this struggle for dominance in approaches to modeling intelligence will be viewed in the historical context of the dialectic in theories of human cognition that has opposed empiricist and rationalist perspectives. This is actually just one of over a dozen oppositions in theories of intelligence that collectively define the search space in which cognitive theory development plays out.

(These lectures were commissioned as a 2-hour synopsis of a course: Foundations of Cognitive Science. They do not present current research, but are primarily pedagogical.)

Date:
Haut-parleurs:
Paul Smolensky
Affiliation:
Microsoft Research

Taille: Microsoft Research Talks