Professional Career
Summary of professional accomplishments
-
2009 Two courses taught using Disciple-LTA and Disciple-COG. I participated in the preparation and delievery of a course on intelligence analysis with Disciple-LTA for intelligence analysts. I also participated in the delievery of the Spring 2009 session of the "Case Studies in Center of Gravity Determination" course at the US Army War College, course which includes a laboratory section based on Disciple-COG. .
-
2008 Book used as course textbook at US Army War College. I co-authored a textbook for the US Army War College course "Case Studies in Center of Gravity Determination". I participated in the preparation and use of Disciple-COG in that course.
-
2007 Disciple use at US Army War College and US Air War College. I participated in the preparation and use of the Disciple-COG in courses at these institutions.
-
2006 Disciple use at US Army War College. I participated in the preparation and use of the Disciple-COG and Disciple-LTA systems in two courses: "Case Studies in Center of Gravity Determination" and "Military Applications of Artificial Intelligence."
-
2006 Guest Co-Editor for AI Magazine Special Issue on Mixed-Initiative Assistants. Together with George Tecuci and Michael Cox I was guest co-editor for the Special Issue of the AI Magazine, on the topic of Mixed-Initiative Assistants.
-
2005 Intelligent System Demonstration at AAAI-2005, the Twentieth National Conference on Artificial Intelligence, Pittsburgh, Pennsylvania, USA, July 9-13, 2005, with the demo: Mihai Boicu, Gheorghe Tecuci, Cindy Ayers, Dorin Marcu, Cristina Boicu, Marcel Barbulescu, Bogdan Stanescu, William Wagner, Vu Le, Denitsa Apostolova, Adrian Ciubotariu. A Learning and Reasoning System for Intelligence Analysis.
-
2005 Disciple use at US Army War College. I participated in the preparation and use of the Disciple RKF/COG system in several courses: "Case Studies in Center of Gravity Determination" Winter and Spring sessions, and "Military Applications of Artificial Intelligence" Spring session. We performed the first successful knowledge base development experiment for Intelligence Analysis.
-
2004 Disciple use at US Army War College. I participated in the preparation and use of the Disciple RKF/COG system in the course: "Case Studies in Center of Gravity Determination," Winter and Spring sessions.
-
2003 Disciple experimentation at US Army War College. I participated in the preparation and completion of the successful experimental use of the Disciple RKF/COG system in several courses: "Case Studies in Center of Gravity Determination," Winter and Spring sessions, and "Military Applications of Artificial Intelligence," Spring session. We performed the first ever successful parallel knowledge base development and integration experiment.
-
2002 Intelligent System Demonstration at AAAI-2002, the Eighteenth National Conference on Artificial Intelligence, Edmonton, Alberta, Canada, July 30-August 1, 2002 with the demo: Boicu M., Tecuci G., Marcu D., Stanescu B., Boicu C., Balan C., Barbulescu M. and Hao X., Disciple-RKF/COG: Agent Teaching by Subject Matter Experts.
-
2002 Disciple experimentation at US Army War College. I participated in the preparation and completion of the experimental use of the Disciple RKF/COG system in several courses: "Case Studies in Center of Gravity Determination," Winter and Spring sessions, and "Military Applications of Artificial Intelligence," Spring session.
-
2001 Disciple experimentation at US Army War College. I participated in the preparation and completion of the experimental use of the Disciple RKF/COG system in several courses: "Case Studies in Center of Gravity Determination," Winter and Spring sessions, and "Military Applications of Artificial Intelligence," Spring session.
-
1999 Knowledge Acquisition Experiment at BCBL. I participated to the successful knowledge acquisition experiment that took place over one week, in August 1999, at the US Army Battle Command Battle Lab, in Ft. Leavenworth, KS, as part of the DARPA's HPKB program. In this experiment, four military experts with no prior knowledge engineering experience received very limited training in the use of Disciple-COA and then each succeeded in significantly extending the knowledge base of Disciple-COA, receiving only very limited support from a knowledge engineer. To our knowledge, this is the first time that such an experiment has been conducted, and its success demonstrates the feasibility of our long term research objective. At the end of the experiment, the domain experts completed a detailed questionnaire where they gave very high scores to our research. For instance, LTC John N. Duquette, Chief of the Experimentation Division of BCBL stated: "The potential use of this tool by domain experts is only limited by their imagination - not their AI programming skills."
-
1999 Best Results at DARPA's Course of Action Challenge Problem. Four research teams participated in the DARPA's HPKB challenge problem called "Course of Action Critiquing": 1) A joint Teknowledge-Cycorp team, 2) The "Expect" research group from the Information Science Institute of the University of Southern California, 3) The "LOOM" research group the Information Science Institute of the University of Southern California, and 4) the "Disciple" research group from LALAB/GMU. Confirming the results obtained during the DARPA's 1998 evaluation, the GMU team distinguished itself by obtaining the best evaluation results among all the participating teams. Moreover, by receiving a score of 114%, it exceeded even the performance of the domain experts that performed the evaluation. This high performance was due to the fact that Disciple-COA was trained by a very competent expert and, as a result, generated many new solutions that were not anticipated by the evaluators. This is a very significant result because it demonstrates that a very knowledgeable expert can train Disciple to exhibit much of his or her expertise. During the evaluation, we again demonstrated a very high rate of knowledge acquisition, the knowledge base of Disciple increasing by 46% during 8 days (from a size of 6229 simple axioms equivalent to a size of 9092 simple axioms equivalent).
-
1998 Selection for EFX'98. Based on the evaluation results in the HPKB program, the Disciple-workaround system was selected by DARPA and Alphatech to be further extended and was integrated by Alphatech into a larger system that supports air campaign planning by the JFACC and his/her staff. The integrated system was one of the 17 systems selected from 84 systems to be demonstrated at EFX'98, the Air Force's annual show case of the promising new technologies.
-
1998 Best results at DARPA's Workaround challenge problem. During the first part of the HPKB program we developed the Disciple-Workaround integrated system that demonstrated a capability by which a knowledge engineer can rapidly build a knowledge base capturing knowledge from Military Engineering manuals. During the 17 days of DARPA's 1998 evaluation, the knowledge base of Disciple was increased by 72% with almost no decrease in performance (Cohen et al., 1998: The DARPA High-Performance Knowledge Bases Project, AI Magazine, 19(4), 25-49). The resulting workaround reasoner also achieved the best scores among all the teams that participated in the workaround challenge problem (see the 1998 Workaround Challenge Problem Evaluation). In a journal paper describing the various approaches to the workaround challenge problem, John Kingston of Edinburgh University states: "The rapid development of GMU's system highlighted its integrated knowledge acquisition tool as being its greatest strength. Over the fortnight of testing, GMU added 150 concepts, 100 tasks and 100 problem-solving rules to their knowledge base, representing a 20% increase in concepts, a 100% increase in tasks and a 100% increase in rules. This rate of knowledge acquisition suggests that GMU's system may indeed be able to achieve one of the Holy Grails of knowledge acquisition: rapid, accurate and direct knowledge entry by an expert without intervention from a knowledge engineer. GMU's system was also capable of reasoning about most aspects of the workarounds problem - indeed, it generated a few (correct) solutions that had not been considered by the expert."