Re: Doug Lenat's THE BIG PICTURE

sowa@west.poly.edu (John F. Sowa)
Date: Thu, 8 Jun 1995 12:47:20 +0500
From: sowa@west.poly.edu (John F. Sowa)
Message-id: <9506081647.AA01032@west>
To: cg@cs.umn.edu, hovy@isi.edu, lenat@cyc.com, srkb@cs.umbc.edu
Subject: Re: Doug Lenat's THE BIG PICTURE
Cc: doug@csi.uottawa.ca, fritz@rodin.wustl.edu
Content-Length: 2192
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

Ed Feigenbaum writes,

>Our many and various AI programs have great form but most of them dont
>do anything very interesting because we have not given them the content
>(the knowledge). As one of this list's e-correspondents put it, it's too
>hard (the implication being that we should therefore run away from it).

I agree with the above point wholeheartedly.  But I have some reservations
about the next:

>The content-full KB is on the critical path to machine intelligence. So
>let's stop philosophizing about it, or logicizing about it. Let's just do
>it!

Although it's important to put content into our programs, it's also
essential to put it in a usable form.  We already have tons of machine
readable dictionaries, thesauruses, and enormous volumes of source text.
There is a lot that can be done with them, and to quote Leibniz again,
"those who have laid out all sorts of notions under certain headings or
categories have done something very useful."  But these things would be
even more useful if they were put in a common form that was suitable
for something more sophisticated than spelling checkers.

Leibniz certainly appreciated the importance of content, but he would
be the last to "stop philosophizing or logicizing about it."  He was
a pioneer in symbolic logic, he inventied binary arithmetic, he designed
and built the first mechanical calculator to do multiplication & division,
he represented compound concepts by products of prime numbers, and he used
his calculator to compute subtypes and supertypes in his concept lattice
(besides his other hobbies like inventing calculus and serving as a
diplomat for the Duke of Brunswick).

Some people are good at collecting and analyzing large volumes of data,
and other people are good at developing high-powered theories.  We need
both kinds of people, and it's important to get them to work together,
attend the same workshops, talk to one another, and apply the best
theories to the best data.

I sympathize with Lenat's complaint about isolating the content people
from the theorists.  We should solve that problem by encouraging them
to pay attention to one another instead of complaining about each other.

John Sowa