How can uniframes classify disparate-seeming things, such as many different kinds of arches? Minsky says that this requires reformulating the description of the members of the category, which appears to require making a move to a more abstract (and therefore less structural?) description. One part of this involves dividing objects into essential and auxilliary portions. The other is to abandon old classification methods that no longer work. This is reformulation. "Reformulation is clearly very powerful -- but how does one do it?"
(NB: Free associating, refourmulation brings to my mind the notion of a paradigm shift. Will Minsky now proceed to describe how reformulation works?)
Pearl Street | Society Of Mind Home Page | Dawson Home Page |