[129]                               home                            [131]

Friday, September 24, 2004

 

The BCNGroup Beadgames

 

Background discussions on a proposed

Anticipatory Technology Challenge Problem

 

 

Discussions with Ben Goertzel,

Novamente architect

 

Ben,

 

The Anticipatory Technology Challenge Problem is one that I have worked on for a decade, as have you and a few others.  The loosely organized communities who have worked on this problem may have as many as 50 significant contributors.  Perhaps several thousand individuals have played around a bit with some of the required ideas: complexity, stratification, information induction, pattern recognition, and modeling of affordance being perhaps the most important ideas.  

 

ARDA's request for proposals for Challenge Problems is an ideal channel, which we hope can be used to produce

 

1) real software that works in real time to produce anticipatory knowledge about event evolution within a targeted event space (like the PTO awards of new software patents)

 

2) a open peer review of all work anywhere that suggests anticipatory capability

 

3) a written curriculum useable in colleges and in schools (K-12) that helps the American public understand the pragmatics of inconsistency and revision.

 

 

I asked you, this morning, if your software and technology, called Novamente, is "anticipatory".  Your answer is "indeed". 

 

I agree that the types of machine intelligence that you and your team has been developing, for a decade or more, has many of the properties that scholars like Robert Rosen, Howard Pattee, Peter Kugler, Robert Shaw, Michael Turvey and others would "require" to fulfill the branding "anticipatory".  But this is because I have had sufficient interaction with you over the years to make that judgment. 

 

But how would you validate the claim that some properties of anticipatory systems are now provided by your software.  I am suggesting that an evaluation of anticipatory technology needs to be part of the Challenge Problem and that a budget should exist to take care an objective evaluation of the Orb (Prueitt, Adi, Ewell), VivoMind (Sowa) and Mark 3 (Ballard) platforms, and other potential platforms like CoreSystem (Klausner), Cyc Corp (Lenat), and Novamente (Geortzel).

 

How might we provide this comparative evaluation within the one-year course of the Challenge Problem?

 

My suggestions, as of today Sept 26th, 2004, are in one page evaluation section of the MITi/OntologyStream proposal to ARDA.

 

I ask for comments.

 

Paul Prueitt

Director, BCNGroup Inc