To: stan.franklin@memphis.edu Subject: Our WDL will start a revolution? Bcc: dlall wdl Stan, I have updated your abstract for WDL. Thank you for your response on sensing! NSF suggested that we start our email discussion before the workshop starts; now it seems a good time. I take the liberty to start. I have provided a Bcc of this email to all the workshop participants. Every participant is encouraged to participate in this pre-workshop email discussion. I volunteer to forward all different views (but not "take me off this list" kind of message) to this email list. The discussion material is also needed for composing our WDL report. If any of you likes to be taken off from this email discussion list, please send an email to wdl@cse.msu.edu. Many important issues that automatic development can fully address include the concept grounding issues, invariance issues, generalization issues, etc. That is a major reason why automatic machine mental development is fundamentally different from any other existing approaches in the AI field. Our WDL will have developmental psychologists who will give their views about how concepts are grounded in real-time, online, sensorimotor experience when children learn new concepts. Off-line spoon-fed machine learning is popular in AI. Does this mode allow sufficient grounding? Further, neuroscientists will report their new foundlings that are probably shocking to AI/robotics guys: Rewiring of vision signals to auditory cortex in animals and then the animals can be trained to do vision tasks (yes, using auditory cortex)! Why are most our machine guys have a hard time using very different methods for vision and speech while our brain does them successfully under general self-organization principles? What are these self-organization principles? Our workshop topics are organized to address these very fundamental issues. Some may feel that the scope of our workshop is too wide. Without such a wide scope, we may never know that most of us in the AI field (robotics, computer vision, speech, language, planning, decision making, you name it) need a fresh, unified way! I found that several recent books about intelligence are very much in line with the work of automatic (mental) development for robots that my students and I are doing. These books suggest a new way of understanding human intelligence. For example, Howard Gardner argues that intelligence is multiple: linguistic, logical-mathematical, music, bodily-kinesthetic, spatial, interpersonal, intrapersonal ... (ref: his two well known books: "Multiple Intelligences" (1993), "Intelligence Reframed" (1999) ). I would add: these multiple intelligences are different behaviors of a tightly integrated THING. When robotics is viewed together with Gardner's multiple intelligence, the new plasticity results from neuroscience, and the studies in developmental psychology, we found that robotics/AI field as well as our way of understanding natural intelligence need a revolution --- this is a mission of this workshop that we proposed to NSF and DARPA: discuss, envision and plan for this revolution. Not only machines, understanding human intelligence also requires clearly specified, fully implementable, experimentally tested, end-to-end algorithms for automatic development. This workshop will not be like any other workshop ... John Weng ---------------------------------------------- Juyang (John) Weng, Associate Professor 3115 Engineering Building Department of Computer Science and Engineering Michigan State University East Lansing, MI 48824-1226 Tel & Fax: 517-353-4388 Email: weng@cse.msu.edu URL: http://www.cse.msu.edu/~weng/ ----------------------------------------------