The Defense Advanced Research Projects Agency is looking to build artificial intelligence algorithms that could make decisions in complex military scenarios where two commanders would not agree on a single right answer.
The In the Moment program aims at building military personnel’s trust in AI systems to present a choice at moments where there is no objective ground truth, DARPA said Thursday.
Matt Turek, ITM program manager, explained that participants will create a framework for evaluating AI decision-making in difficult domains by comparing an algorithm’s choice with a human in multiple scenarios.
“The [Department of Defense] needs rigorous, quantifiable, and scalable approaches to evaluating and building algorithmic systems for difficult decision-making where objective ground truth is unavailable,” added Turek.
DARPA will select companies to test AI systems on a small-unit medical triage domain for two years and on mass-casualty events for 18 months, with potential awards following the two phases for technology maturation.
The agency will hold a virtual event on March 18th to discuss the opportunity with potential proposers.