( Index )

 

(The DARPA REAL BAA)

 

Proposal roadmap

(SAIC/OntologyStream proposal for Real World Reasoning due September 2003)

Using the Knowledge Sharing Core Concept

 

Goals: We seek a breakthrough in measured knowledge sharing fidelity based on differential ontology formation and real time ontology processing.  The initial 12-month period will establish a market driven method and a deployment/use ecosystem that will drive an anticipated advent of machine interoperability and agility within the semantic web. 

 

Benefits: Benefits include better intelligence findings from any large data stream, especially in practical terms related to detecting novelty, low salience changes, and broad emergent shifts that other methods miss.  The use process of the ecosystem will self synchronize and reinforce those processes that analysts actually use.  Intuitions about how to use advanced computational processes will be enhanced because the deep knowledge of how the analytic system works will not be held as proprietary secrets and will be provided in a transparent fashion via scholarship exposition from leading university faculty [1] .

 

Technical barriers: A specific class of technical barriers stem from specific paradigms in computer science and Information Technology, and from institutions traditionally involved in evaluation, procurement and deployment.  Both the specific paradigms and the evaluation, procurement and deployment practices are typical of Industrial Age Organizations [2].  These barriers inhibit collaboration and rapid innovation based on relevant natural (cognitive and social) science and on new advanced computational theory – including control theory, number theory, general systems theory and theory of algorithms.  Many of these technical barriers come from the market rejection of the strong form of Artificial Intelligence.  It is true, however that AI has given us some useful experience in machine representation of human knowledge.  But our group bypasses many of the technical barriers by redefining what is expected from algorithmic processes and pushing the cognitive load away from the computer, where it cannot be, into the human domain [3]. 

 


Elements of approach: The elements of approach have four independently managed components. 

(1) (Algorithms and science) Algorithms from competitive traditions are modified, selected, compared, combined, and tested rapidly, and are organized by hypotheses derived from a specific set of natural science theories about memetic expression, human reasoning, and human perception.  Those who have originated and continue to develop the theories advise the program and will provide internal scientific peer review over the results. 

(2) (Knowledge component) We provide distance learning modules on elementary algorithmic processes within the ecosystem so that anyone can learn how and what the internal processes do and do NOT do.  Social and cognitive sciences are very clear in the claim that distance-learning activities are necessary for adoption of transformational technology.  These are not traditional training programs but a clear and transparent presentation of the underlying principles and processes involved in the Knowledge Sharing Core.  

(3) (Infrastructure) We radically reduce dependencies on enterprise software such as .NET and J2EE and provide increased stability and access allowing innovation to be confidentially expressed directly into a common design language.  The radical reduction uses Open Source tools such as the Berkeley Database, and Linux kernels to build a solid foundation in software modules that have a commonality of functionality determined by the design language (called Cubicon [4]).

(4) (Patents and Intellectual Property) We believe that relevant parts of the intelligence property space can be organized to reflect scientific consensus and fairness to true innovators.  First we see the application of the Knowledge Sharing Core technologies in the development of a map of the memetic expression of innovators.  Second, we offer a common design language.  This design language allows (a) the easy expression of the innovation’s reduction to practice as computer process and (b) facilitates the rapid testing of the innovation within the Knowledge Sharing Core as one of many tools available for use and reinforcement via metrics on tools usage.

 

Rationale that builds confidence:  New algorithms, and new synergistic combinations of existing tools, are applied rapidly to outperform methods currently applied to high value intelligence problems. This program will rapidly build a community of practice involving between 20 – 30 leading scientists and engineers from participating software companies, many of which will collaborate openly in highly focused workshops with low cost to the project.

 


Nature of expected results:  More analysts are willing to use these new tools compared to other research systems, not only because the results are better and more obvious, but because the role of the analyst is fully recognized and enhanced rather than reduced.

 

Risk if work is not done: In the absence of this alternative, elaboration (at great expense) of the general artificial intelligence paradigm will continue.  Many scientists recognize this paradigm as founded on outdated notions about the nature of human reasoning.  Lack of fidelity and real time usability in processing of semantic data guarantees continuation of a broad class of intelligence failures.

 

Criteria for annual progress evaluation:  Level of adoption by intelligence analysts in pilot settings.  Findings in pilot operations are corroborated by other existing means offering less fidelity.  One measure of progress is the number of patent applications from ecosystem members.  However, we anticipate the development of a revenue model based on the notion of micro-transaction reporting.  The Knowledge Sharing Core will be self-sustaining within five years.

 

Cost for each performance year

 

FY 04

FY 05

FY 06

FY 07

FY 08

Total

1,600K

800K

600K

600K

400K

4M