SUO: Re: article on the pitfalls of metadata
Jon,
I strongly agree with the following point:
> One of the main hangovers is analytic philosophy, that continues
> to promulgate the tardive delusion that you can conquer significant
> problems by dividing them into just two classes, the "already solved
> in principal" and the "illegitimate".
The so-called illegitimate problems are usually the most important
ones that really get to the core of the issue.
As an example, I believe that the worst mistake of the formalists
(whose papers I continue to read and even contribute to from
time to time) is the identification of logic with deduction
to the nearly total exclusion of any attention to induction,
deduction, and analogy.
I agree with them that deduction is important, and then they're
happy to accept me as a fellow traveler. But as soon as I add
the point that induction, abduction, and analogy are equally
important, they lose interest and classify me as somebody who
is "not in the mainstream".
Unfortunately, what they call the "mainstream" is a little
brook, which happens to flow into the Mississippi, which
later flows into the Gulf of Mexico, which is connected
to the much bigger oceans.
For some further thoughts on this issue, see the slides of
my talk at the UQAM Summer Institute on Cognitive Science:
http://www.jfsowa.com/talks/uqam.htm
Categorization in Cognitive Computer Science
Later that week, I presented a few more slides as an
intoduction to a panel session:
http://www.jfsowa.com/talks/uqam2.htm
Data Mining and Knowledge Discovery
These slides end with some embarrassing questions
about why none of the current work on machine learning
can begin to compete with a 3-year-old child.
John Sowa