Speaker: John Laird , Electrical Engineering and Computer Science Department, University of Michigan
Toward Human-level AI
Abstract: Although the creation of integrated, human-level AI has not been a major goal of the field during the last 20 years, some progress is being made. In this talk I will give an overview of extensions we have added to the Soar architecture that give it significant new capabilities: episodic memory, reinforcement learning, and models of emotions and feelings. Episodic memory has largely been ignored by AI and cognitive science, but plays a key role in our ability to remember prior events. Reinforcement learning has been extensively studied in Machine Learning, but not within the context of large knowledge base systems with complex state structures.
Our research on emotions and feelings explores how affective reasoning intermixes with symbolic cognitive reasoning. I will review the design of Soar, and describe how these new capabilities are integrated with it, as well as present preliminary results of using these capabilities on simple tasks.
Short Bio: For details, please see the speaker home page.
Host: Michael Young, Computer Science, NCSU