I like that Sims approaches the problem of AI from the perspective that life is a consequence of the world, that life is the world discovering itself. He specifies a learning semantics (genetic algorithms) and a learning syntax (motivation functions and virtual embodiment in time) for his creations. His specifications are functor-like in that they determine a structure on the world that when probed gives information about the world, more or less finely. Through process come functions like crawling, reaching, or defending. Some how these functions follow from motivation, learning and the world. Is it reasonable to interpret them as dependent functions of the underlying motivation functions, the motivations acting as a generalized grobner basis?
To Glen's point, or perhaps the point of the Bengio paper, if we watch long enough and the virtual world has sufficient analog to our own, we can begin to experience a transparency of understanding. Still perhaps, the understanding is not of the agent but of the world.
.-. .- -. -.. --- -- -..-. -.. --- - ... -..-. .- -. -.. -..-. -.. .- ... .... . ... FRIAM Applied Complexity Group listserv Zoom Fridays 9:30a-12p Mtn GMT-6 bit.ly/virtualfriam unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com archives: http://friam.471366.n2.nabble.com/ FRIAM-COMIC http://friam-comic.blogspot.com/