|
|
Anthropology and Prediction
John Mcreery (JLM@TWICS.COM)
Sat, 9 Sep 1995 21:58:00 +0900
Mike Carson writes,
"I'd like to emphasize this point a little more. We can make
generalizations, noticing correlations between phenomena. For example,
we can correlate population growth and agriculture or perhaps social
stratification. But we do not create full EXPLANATIONS for these
correlations until we have discovered anomalies or exceptions to the
generalizations."
This reminds me of something I read, if my memory is accurate, in what is now
a very old book on AI programming in LISP by Patrick Winston. The reference
was to the "Blocks World," one of the first attempts to develop programs
capable of learning from experience. The point was, like Mike's, that the
best way to do it was start with a prototype and then use error-correction as
a way to refine it. I wonder if anyone, say Danny Yee (who, we know, is more
than a little computer-literate) could chase down what's happened to this
idea?
-------
Thanks, by the the way, to Danny for the reviews of "Rabbits & Spaghetti" and
the book on LaTex. One I'd really like to see him take on is _Neurophilosophy:
Toward a unified science of the mind/brain_ by Patricia Smith Churchland
(1992) Cambridge:MIT Press. Interesting book by an interesting woman who
decided that if, as a philosopher, she was going to think about the way the
mind works she ought to know neurology. So she went to medical school to add
and M.D. to her Ph.D.
Cheers,
John McCreery
|