September 17, 2004
Background discussions on a proposed
Anticipatory Technology Challenge Problem
Richard Ballard and John Sowa,
One never knows about funding, however the BCNGroup/Ontologystream proposal for a Challenge Problem to create Anticipatory Technology:
http://www.ontologystream.com/beads/nationalDebate/challengeProblem.htm
may be funded at 400K for one year, and it could be funded at a higher level if one or both of you were to define a role within the context of the Challenge Problem. We have champions, as requested in the RFP, in two agencies.
Dick, you suggest that Mark 3 could be developed based on what you now know and your existing programming design within 12 months and 500K.
The ACQUAINT Program seems ideal for Mark 3
http://nrrc.mitre.org/arda_explorprog2005_cfp.pdf
If you modified your knowledge acquisition process so that information sources were about concepts expressed in text, or not... but directed at being anticipatory in the sense that a question anticipates the answer. Even as a senior advisor to the project, at any level of funding that you need, we may be able to get a Knowledge Foundations Mark 3 derivative so that the questions are converted into convolutions of Orbs. (see notational papers on Orbs)
http://www.bcngroup.org/area2/KSF/Notation/notation.htm
Remember, that the Orbs are really simply an encoding of the elements that come to occur in n-tuples into this key-less hash table (or something else if your data structures are better than mine.) How the n-tuples come into being is not relevant to the Orb formalism. So some n-tuples can be from the linguistic analysis of Arabic and some can be from a structured and classified XML resource on where known terrorist live. Orb arithmetic is very easy due to the encoding and the use of mathematics based on linear orders of things inside the Orb.
I do not know if this is something that you feel should be done. Perhaps you will write briefly about your feelings. In any case the question-answer model of computer human interaction is something you have worked with.
John, My hope is that you would not object if asked about a positive funding decision for the BCNGroup/Ontologystream challenge problem.
Better would be that you were able to start with the concept that structural co-occurrence of data should be separated from interpretation, at least during a process of measurement.
If course, measurement depends on instrumentation and instrumentation is already involved in some type of prior expectation. As you know, the human brain processing visual and other sensory input using feed back and feed forward chains of reactions.
So the found "structure in data" is defined by "instrumentation". Our parsing of text is an example of this. But, significantly one does not have to parse text to make an Orb system of the type that I have developed. One can have some production of "constructions of primitives". The constructions can then put into a linear order and parsed to produce a graph.
These graphs, or sub-graphs, can then be viewed, and human(s) need not complicate what is viewed. The graph can be algorithmically compared, as suggested by your work on analogy reasoning, using similarity measures, and graphs can be transformed into other graphs, using subsetting and convolutions.
Ken Ewell and I have located an Angel investor for the Provenance product:
http://www.bcngroup.org/beadgames/InOrb/provenance.ppt
and this will keep the core team functioning.
But there is a community of others that struggle unnecessarily with difficult, and wonderful, theory and application; while also having to struggle financially.
This community of others is the reason for the planning process leading, perhaps, to a National Project.