Xref: utzoo comp.ai.neural-nets:1113 comp.ai:5055 Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!snorkelwacker!spdcc!merk!alliant!linus!mbunix!bwk From: bwk@mbunix.mitre.org (Kort) Newsgroups: comp.ai.neural-nets,comp.ai Subject: Rule-based systems and model-based reasoning Summary: On learning to build reasonable models. Keywords: Deep Reasoning, Models, Analogies, Metaphors, Car Talk Message-ID: <80267@linus.UUCP> Date: 13 Nov 89 09:41:45 GMT References: <1690@cod.NOSC.MIL> <77404@linus.UUCP> <13659@orstcs.CS.ORST.EDU> Sender: news@linus.UUCP Reply-To: bwk@mbunix.mitre.org (Barry Kort) Distribution: usa Organization: The Ferchachta Corp. Bedford, MA Lines: 20 In article tgd@aramis.rutgers.edu (Tom Dietterich) writes: > Rule-based systems (such as those described by Klatt) may attain > superior performance. The challenge is to come up with learning > methods that can match the performance of hand-crafted rule bases. I agree. Large rule-based systems are brittle, unwieldy, and difficult to evolve. I expect that model-based reasoning will eventually supplant rule-based systems. At the present time, model-based reasoning (also called deep reasoning) is an AI frontier. When I do diagnostic reasoning, I reason from a mental model or from a mechanical or computer simulation model. Most of the work is in the construction of the model. One of the reasons that I cannot diagnose the trouble with my Pontiac's carburetion system is that I don't have a clear mental model of how it works. --Barry Kort