KIF

• Mail folder: Interlingua Mail
• Next message: Ramesh Patil: "Re: KIF "
• Previous message: sowa@watson.ibm.com: "Definitions and analytic truth"
• Reply: Ramesh Patil: "Re: KIF "
• Reply: Michael R. Genesereth: "kif"

From: dam@ai.mit.edu (David McAllester)
Date: Tue, 7 Jan 92 11:12:58 EST
Message-id: <9201071612.AA03111@raisin-bran>
To: sowa@watson.ibm.com
Cc: ramesh@isi.edu, interlingua@isi.edu
In-Reply-To: sowa@watson.ibm.com's message of Mon, 6 Jan 92 23:02:28 EST <9201070402.AA13794@life.ai.mit.edu>
Subject: KIF

I want to apologize for the somewhat harsh tone of my previous
message.  I have often defended KIF as an opportunity to think about
fundamental issues in knowledge representation.  I have not made any
great effort to become involved myself because my world view seems
quite different from that of the other researchers involved.  For
example, my main interest in definitions involves the semantics of
recursion.  I think this is an important and subtle issue that is
fundamental to knowledge representation.  There is a fundamental
difference between least fixed point and greatest fixed point
semantics.  Nonmonotone recursion is another can of worms (there may
not be any fixed point).  I have not yet found anyone involved in the
KIF project who shares my interest in the semantics of recursion.

Another interest of mine involves the value of higher order logic.
First order logic is far weaker that most AI researchers seem to
appreciate.  For example, first order logic can not express the concept of
transitive closure, or the concept of a finite path in a graph.
Any universal standard for a knowledge representation language should,
it seems to me, be able to express these concepts.  As with recursion,
I have not found people involved with KIF who are interested in role
of higher order logic in knowledge representation.

	David McAllester