The Defense Advanced Research Projects Agency will host a proposers day on June 14 to discuss with industry an upcoming broad agency announcement to develop and test mathematical methods for guaranteeing the capabilities of generative artificial intelligence models used across the Department of Defense.
DARPA said Wednesday the Artificial Intelligence Quantified, or AIQ, program aims to test the hypothesis that mathematical foundations of AI, combined with advances in modeling and measurement, will enable guaranteed quantification of generative AI capabilities.
“AI has achieved near human-level performance in domains including text generation, game playing, and such, which raises the prospect of widespread integration with human partners in the military and society,” said Patrick Shafto, program manager of AIQ at DARPA. “And at the most general level, we’re interested in determining how to ensure AI systems will have the properties needed to solve various problems.”
The program will be divided into two technical areas. Technical Area 1 will focus on providing foundations for understanding and guaranteeing capabilities across levels while Technical Area 2 will develop methods for evaluating AI models.