Beyond chess: Deep green models rapid change for combat commanders

The USC Information Sciences Institute is playing a $7.6M part in a DARPA next-generation battle command and decision support technology

Screen shot of Arena War. Click here for more information.

Can an artificial intelligence program anticipate military surprises? The USC Information Sciences Institute is playing a $7.6 million part in a DARPA research effort called Deep Green aimed at creating a system that can do so, one that might help future combat commanders in the field anticipate enemy moves.

The same system would look around and recruit additional computing resources if the situation were too dire, the problem too difficult.

The Deep Green program, a next-generation battle command and decision support technology, is the vision of Col. John Surdu, who manages the program for the Information Processing Techniques Office of DARPA.

The system interleaves anticipatory planning with adaptive execution to help the commander think ahead, identify when a plan is going awry, and prepare options, before they are needed.

Deep Green will use a human operators hand drawn sketches and words to induce intent. It will generate options for all sides in an operation and predict the likelihood of multiple futures.

By presenting decisions early and allowing the commander to "see the future," Deep Green supports commander's visualization and adaptive execution, enabling correct, timely decisions by the commander.

Deep Green has several components, including novel interfaces for getting guidance from and presenting options to commanders, powerful simulations of the battlespace, and methods for efficiently searching the space of future options. The prime contractor, responsible for all these elements, is SAIC.

ISI researcher Paul Cohen, heading one of the two ISI groups subcontracting on the program, notes that the name is meant to recall Deep Blue, the famous IBM chess playing program that defeated world champion Garry Kasparov in a 1997 match, a landmark in the history of artificial intelligence.