Re: [unrev-II] Sowa on Semantic Nets --Fwd: [xtm-wg] Re: an introduction to the BCNGroup beadgames

From: Eric Armstrong (eric.armstrong@eng.sun.com)
Date: Mon Aug 20 2001 - 16:40:51 PDT


Super post, Jack.

Jack Park wrote:
>
> ...
> From: "John F. Sowa" <sowa@bestweb.net>
>
> ... Following are six of the most common kinds of semantic networks:
> > > 1. Definitional networks emphasize the subtype or is-a relation
> between a concept type and a newly defined subtype....
> >
> > 2. Assertional networks are designed to assert propositions...
> > >
> > 3. Implicational networks use implication as the primary relation
> for connecting nodes...
> >
> > 4. Executable networks include some mechanism, such as marker
> passing or attached procedures, which can perform inferences, pass
> messages, or search for patterns and associations...
> >
> > 5. Learning networks build or extend their representations by
> acquiring knowledge from examples...
> >
> > 6. Hybrid networks combine two or more of the previous techniques...
> >
Basically, we seem to have 5 kinds of mechanisms in the knowledge-
representation game:
  a. Definitions
  b. Assertions
  c. Implications
  d. Executables
  e. Abstractioning

That is a useful classification of intelligence-mechanisms, I think.
It is interesting to think that a combination of all of these could
produce something that equates to intelligence.

For example, in chess, the problem of "pattern recognition" (in the
sense that a player recognizes a pattern on the chessboard") would
seem to imply an abstraction, with attendant assertions and implications
based on the classification (aka "definition") that results. Finally,
an executable is needed to do the limited analysis we are capable of.

Of all the aspects, abstractioning is the trickiest. But were such
a module developed, it could be combined with the others to yield
a truly "intelligent" chess playing program that starts, as we all
do, from nowhere.

  Note:
  I suspect that "learning" is really a separate mechanism from
  "abstractioning", in that learning is really the adoption of
  someone elses abstractions. The "abstractioning" process, in
  turn, has 2 forms:
    a) The pattern discovery process, where you figure out what
       things have in common.
    b) The pattern recognition process, where you relate a new
       situation to ones that have been previously discovered
       or learned.

The model for intelligence, then, would have a "pattern repository",
which was loaded by a combination of discovery (reasoning) and
learning (adoption of other patterns, which interestingly enough
*also* involves reasoning). Once patterns are stored, they can
be compared to new situations, looking for a fuzzy match.

That process is pretty much what a new chess player goes through,
having no patterns at the outset, followed by some rudimentary
patterns that are seldom of any real help, followed by more and
more patterns that are increasingly more refined.

Machines of that kind would make it possible to objectively rate
one kind of pattern-recognition strategy against another, and
answer questions like, "What is the best way to think about X?",
where X may be the game, or a particular kind of situation.

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Do you need to encrypt all your online transactions? Secure corporate intranets? Authenticate your Web sites? Whatever
security your site needs, you'll find the perfect solution here!
http://us.click.yahoo.com/Bre3tC/Q56CAA/yigFAA/IHFolB/TM
---------------------------------------------------------------------~->

Community email addresses:
  Post message: unrev-II@onelist.com
  Subscribe: unrev-II-subscribe@onelist.com
  Unsubscribe: unrev-II-unsubscribe@onelist.com
  List owner: unrev-II-owner@onelist.com

Shortcut URL to this page:
  http://www.onelist.com/community/unrev-II

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/



This archive was generated by hypermail 2.0.0 : Tue Aug 21 2001 - 18:58:49 PDT