"How do we make generalizations from fragmentary bits of evidence?" No one has solved the problem of generalization. Minsky's slant on this is uniframes, level bands, and now polynemes. (NB: For connectionist systems in general, generalization is a real problem. If Minsky is buying into a connectionist model, then his concern with generalization is justified.)
Pearl Street | Society of Mind Home Page | Dawson Home Page |