Towards a Graphical Cognitive Architecture for Virtual Humans (and Intelligent Agents/Robots)

A cognitive architecture provides a hypothesis about the fixed structure
(and its integration) underlying intelligent behavior, whether in natural or
artificial systems. The overall goal of this effort is to leverage graphical
models – with their ability to uniformly yield state-of-the-art algorithms
across symbol, probability and signal processing – in developing a new
architecture that goes significantly beyond today's best in providing, and
tightly integrating together, the capabilities required for virtual humans
(and intelligent agents/robots). The current focus is on a graphical mixed
(i.e., statistical relational) architecture that supports hybrid processing
through its grounding in a continuous representation. Aspects of memory,
problem solving, perception, imagery, learning and natural language have
been demonstrated to date in this architecture, although some are still mere
beginnings. The talk will introduce the desiderata for this graphical
architecture, explain the basics of its operation, and highlight progress on
some of these capabilities.
Paul S. Rosenbloom is a Professor of Computer Science at the University of
Southern California (USC) and a Project Leader at USC's Institute for
Creative Technologies. He spent twenty years at USC's Information Sciences
Institute, including a decade leading new directions and a stint as Deputy
Director. Earlier he was an Assistant Professor of Computer Science and
Psychology at Stanford University, and a Research Computer Scientist at
Carnegie Mellon University. He received his B.S. in Mathematical Sciences
(with distinction) from Stanford University and his M.S. and Ph.D. in
Computer Science from Carnegie Mellon University. He is a Fellow of the
Association for the Advancement of Artificial Intelligence (AAAI). Prof.
Rosenbloom's research focuses on cognitive architectures; he was a co-PI of
the Soar Project for fifteen years, and is currently developing a new
approach based on graphical models. He has also been working to understand
the nature and structure of computing as a scientific domain.