(Technical Portion of NIMA Proposal) (e-mail Dr. Prueitt) (The Anticipatory Web)
on
On September 14, 2002 we asked for scientific comment on the following, and comment by government program managers.
There are four areas of innovation.
One innovation is an extension of
Latent Semantic Indexing, as is pointed out and as two of the reviewers make note
of. Categorical abstraction is a generalization of LSI. Moreover, the
use of a visualization of Generalized LSI (G-LSI) as small Peircean logic atoms
fits into a extensive literature on situational logics. The scholarly
background extends into the Russian applied semiotics literatures, which is
clearly innovative and largely laying outside of the western approaches to
Intelligent Control (as defined by Albus and Meystel at NIST).
A second innovation is in the use of
in-memory data structures that allow a convolution operator to pass
quickly over massive data to actually produce the categorical
abstraction that is then viewed.
A third innovation is in the interface between
the "synthetic perception" of categorical invariance and a standard
knowledge base such as the Cyc knowledge base.
The SAIC/OntologyStream proposal was developed
to reflect the BAA, and in the BAA there are five areas of focus. We where asked to not
focus on all aspects under the BAA direction that a team would be put
together.
For example, one corporation that was
deemed fundable and then funded is Cyc Corp.
They have an existing knowledge bases system of very high quality.
We anticipated that Cyc Corp or a similar knowledge base vendor will be
selected and that we would be asked to provide interoperability between that
knowledge base and the three innovations that we have. Had there no been anyone who was offering a
knowledge base, then our advisor Dr. Richard Ballard was prepared to introduce
the Mark 3 knowledge base.
It is made clear in the proposal that low
level bit stream events are aggregated into intrusion events and that the
intrusion events are aggregated into incidents and that incidents are compared
with policy guidance.
The clarity of the presentation rests on
the notion of "stratified processes", and this is a forth
basic innovation. This innovation is grounded in the cognitive neuroscience that
breaks into three layers of separated processes:
·
Memory process of invariance extracted (convolved) from
experience
·
The non-equilibrium formation (emergence) of an autopoietic
envelop as selective attention within awareness, and
·
Anticipation that forms from the mapping of the environment
of threats and vulnerabilities.
This "tri-level" architecture
for machine intelligence is the fifth area of innovation. This is
difficult to learn science, but then again the question we raise is about
the government in-ability to fund this degree of innovation. The case is
made that without a breakthrough, that Predictive Analysis Methodology
will be simply another branded term that is used to sell software and services
to the government in this time of national crisis.
In our proposal, we have a deep
grounding in the cognitive neurosciences, in theory of human perception and in
human factors research communities.