The Defense Advanced Research Projects Agency has chosen five teams to participate in its human-artificial intelligence teaming program focused on multimedia analytics.
Boston Fusion, GE Research, Systems & Technology Research, the University of California San Diego, and the University of Illinois Urbana-Champaign will lead research projects under the Environment-driven Conceptual Learning program, DARPA announced Tuesday.
ECOLE aims to develop AI agents that can continuously gain knowledge from machine learning-based algorithms in order to analyze massive amounts of video, image and other multimedia data for national security missions. The program’s first and second phases will each take 18 months to complete, while a third phase specializing in geospatial intelligence is expected to take 12 months.
The research projects chosen for ECOLE include Boston Fusion’s exploration of the “masking” technique to test AI agents’ recognition of an object despite the absence or removal of some of its properties.
GE Research, meanwhile, will devise automated object and action curricula generation that would help AI discover an object’s properties. The company will also create methods to address user input and knowledge structure conflicts.