On Sat, Aug 17, 2002 at 10:47:44AM -0500, Melissa Fitzgerald wrote: > I wonder whether the built-in strictness of programming languages > is a cause, or merely a symptom, of the difficulty we have had in > developing artificial intelligence. I'm not aware of any computer's I'd hazard more like a symptom, but not really that. We want our computers to be precise and predictable; this calls for easily resolvable, ambiguity free languages. AI, or dealing with humans, is a different problem than computers have usually had to solve... "Error: user is ill-defined." > ability to make sense of symbols with multiple, simultaneous and > contradictory meanings. (Is this something that fuzzy logic This is something I've thought about, trying to figure out what my cognitive science research will be. The work out of Douglas Hofstadter's lab, by Melanie Mitchell and others, has aspects of this: the programs are all about finding analogies, and at least while 'thinking' the programs can contain multiple pressures competing over the 'meaning' or best analogy. (Sample problems might be: if abc goes to abd, what does ijk go to? Or xyz? Or mrrjjj?) One can argue for ijk->ijl, or ijk->ijd, or ijk->ijk... the Copycat program has to figure out how to describe the abc->abd transformation. But yeah, I'd like to work on ambiguity resolution in language. -xx- Damien X-)