*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Lifelong Interactive Learning for Open-ended Co-creative Embodied Performance
Date: Monday, August 7, 2017.
Time: 10:00 - 12:30 (EDT)
Location: TSRB 223
Mikhail Jacob
Ph.D. Student
School of Interactive Computing
College of Computing
Georgia Institute of Technology
Committee:
Dr. Brian Magerko (Advisor, School of Literature, Media, and Communication, Georgia Institute of Technology)
Dr. Ashok Goel (School of Interactive Computing, Georgia Institute of Technology)
Dr. Mark Riedl (School of Interactive Computing, Georgia Institute of Technology)
Dr. Mary Lou Maher (Department of Software and Information Systems, University of North Carolina Charlotte)
Abstract:
Improvisation is the gold standard for co-creativity between humans and computers due to the real-time performance, adaptability of agent responses to the current situation, and open-endedness of content knowledge required. A significant challenge for the creation of human - computer narrative improvisation through movement (movement improv) is the knowledge-authoring bottleneck. The knowledge-authoring bottleneck is the difficulty of acquiring expert knowledge and the complexity in storing it efficiently for future use. For movement improv, this knowledge includes actions that a character can perform and action selection knowledge to choose amongst them during the improv. The significant open-endedness of the domain and lack of full-body gestural datasets with formally annotated action semantics motivate novel learning approaches for this content knowledge.
Prior research has explored solutions to the knowledge-authoring bottleneck for open-ended abstract movement improv between humans and virtual characters in the LuminAI interactive art installation. The virtual character in LuminAI learned the content knowledge required to improvise through imitation learning and case-based reasoning over its lifetime. The current research proposes that virtual characters can learn to perform open-ended movement improv with humans that is mutually judged novel, high quality, and surprising, by making use of imitation learning, embodied knowledge representations, and computational models of creativity, in order to mitigate the knowledge-authoring bottleneck involved.
