Sunday, May 25, 2008

John McCarthy's Common Sense and Artificial Intelligence about Evidence and Inference

John McCarthy, one of the four 1956 Dartmouth conference founders of artificial intelligence, has a great many things to say that have a bearing on how ordinary people (including lawyers) must reason from evidence to judgments about states of the world (i.e., about factual questions). See, e.g., John McCarthy, Making Robots Conscious of their Mental States (24 July 1995 - 15 July 2002).
  • I find it interesting that McCarthy, working from a starting point different than Lotfi Zadeh's, agrees with Zadeh that many of the concepts that we human beings use and must use to make sense of our word and to make our way through and in the world are "approximate" in a way that the standard probability calculus cannot capture. See, e.g., John McCarthy, Approximate Objects and Approximate Theories (Feb. 2, 2000)
  • But what does the construction of robots have to do with the validity of different logical theories? Answer: How the mind works and must work should ultimately be tested against "reality." The attempt -- or, in any event, thinking very, very carefully carefully about how we would would have to proceed -- to build a working robot with the intelligence of a human being, a robot that actually mimics some of the intelligences of the human animal, is a better constraint than pure armchair theorizing, don't you think?

    No comments: