Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!decwrl!apple!lantz
From: lantz@Apple.COM (Bob Lantz)
Newsgroups: comp.ai
Subject: Re: Doug Lenat's "cyc" project
Message-ID: <44091@apple.Apple.COM>
Date: 19 Aug 90 21:43:06 GMT
References: <1990Jul31.034417.19350@nixtdc.uucp> <1990Aug19.145236.1@csc.fi>
Organization: Apple Computer Inc., Cupertino, CA
Lines: 26

ylikoski@csc.fi writes:

>Why not have the possibility to build incrementally (by means of the NL
>interface) not only propositional but also procedural information
>into the "cyc"?

I believe that although Cyc optimizes (i.e. hard-codes in lisp) most 
common inferences (i.e. procedural information,) the information in
the system, propositional and procedural, is also encoded in declarative
form.  In general, the system is designed to support multiple inference
schemes.  The rationale behind this is that certain inference schemes
may be better for certain problems, and a clear-cut "winning" method
has not been discovered.  Currently, general-purpose algorithms can
reach the same conclusions as the optimized algorithms, (because of the
declarative encoding,) but at a much slower rate.

The natural language interface is not complete enough to enter information
into the system, I believe. To complete the natural language system,
they need a more complete knowledge base, so there's a sort of bootstrap
problem which they are solving by encoding information by hand.  Lenat
and others have compared the current process with "teaching by brain 
surgery."  

>Antti (Andy) Ylikoski              ! Internet: YLIKOSKI@CSC.FI

Bob Lantz