Freeman's results give a great deal of insight on how grounding is
actually implemented. Werbos', Barto's, Skaggs', et al., work on Heuristic
Dynamic Programming (including Paul Werbos' DHP--dualistic heuristic
programming, which is HDP with the goal state information being partials
with respect to actions, rather than absolute state relative to the goal)
give insight into how the motor response is probably controlled. Is there
some way of merging the two processes? If so, it provides a way of
representing the mammalian brain as a complex stimulus/response system.
(Very complex!) If not, then what's present at the "center" defined by
incoming stimuli and outgoing actions?
Internet: erwin at trwacs.fp.trw.com