Saturday, January 22, 2005
Center of Excellence Proposal
à
Evolution towards
the National Project
Peter (Stephenson)
you said
"I have a line on a very large DARPA grant that stands a good chance of funding a three year project on attack mechanics and deployed artificial immune systems. Lots of details to tell you so we can talk in the morning. I'm headed home from the airport now."
</quote>
As anyone can see from the new Challenge Problem BAA, DARPA is deeply committed to artificial intelligence, and this new approach we have taken is an alternative viewpoint.
Our approach is "structural". No magic. The abstractions developed do not become self aware.
In our approach one is required to gain structural (co-occurrence plus other) knowledge of real events, and this acquisition occurs through instrumentation, measurement, encoding (as categories) and interpretation (of event patterns).
The "magic" is that the natural world is organized as if in layers, like chemistry is organized from the elements of a small set of atoms. (C. S. Peirce; D. Pospelov and others have worked under this assertion).
OntologyStream's new contract to DHS will allow this to be demonstrated in a high profile way without anyone really becoming aware that we are making a significant step towards Tim Berners Lee's notion of a two-sided semantic web.
A separate effort by Nathan, you and I will result in some type of analysis about the "real" prototypes that are part of the cause of events, attacks or the formation of vulnerabilities, can be persisted as defined concepts. We have to remember the history that lead Tom Adi to his discovery of the Q matrix underlying Arabic use.
We are following a similar path... instrumentation, measurement, encoding (as categories) and interpretation (of event patterns). This "concept" or "event" role up is also demonstrated in the NdCore from Applied Technical Systems and my generalization of part of their work:
Then the set of concepts can be encoded into representation of elements of those concepts that we make explicate. Ontology Web Language is a sufficient means to do this, at least imperfectly.
The paradigm of AI may shift in the direction of Semantic Web (which has explicitly disowned AI) and bring with it the funding that must follow mistakes.. to create the illusion that it was not an mistake.
This then is a prediction, on my part.
AI will gravitate towards SW, as SW ontologies begin to serve as mediators to XML repository exchanges. The machines will "know" each other and the usefulness of agile situational organization of data (originally expressed as XML data) will be claimed as a consequent of the AI effort.
I can live with this, for reasons that the SW makes it clear that the origin of intelligence is a human doing something about the situations that they face.
I suggest that, Nathan's work needs a week more to complete some tasks, and then we should look serious at what he has done (without necessarily making in pre-judgments). We will all three examine what is being measured and then describe these new results in an extension of the cA, eC, gFt paradigm.