Learning for Skill Acquisition and Refinement: Toward Exploring Everyday Physics

The present talk claims that "robotics" is not a test bed for AI but should involve a research frontier, which attempts to account for intelligibility of everyday physics underlying human activities such as perception, remembrance, planning, practices, and skill. In addition to traditional AI and neuro-network approaches, more of new domains that can account for any aspect of human intellectual behaviors must be exploited, and also more of new tools that actualize real implementation of intelligence in machines need to be devised. To aim at going on an expedition in this direction, this talk introduces one new domain and another new tool. The former is practice-based learning for skill refinement and the latter is a design tool of signal-based structured information base for skill acquirement.