The Joint Artificial Intelligence Center (JAIC) has been utilizing resources to work with private industry to understand best practices and techniques when working with different datasets to best implement AI.
“Currently, I spent about three months talking to about 20 different small and big AI companies as a way to kind of fully understand how to best interact, how to change talent, as well as techniques,” Chief of Operations at JAIC Col. Stoney Trent said Thursday, June 6 at the C4ISRNET annual conference.
Col. Trent mentioned that commercial groups aren’t entirely proficient at incentivizing for rigorous testing and that it could represent a business risk for them to do so.
“What the Defense Department has to offer in the space is encouraging an incentive structure for better testing tools and methods that allows us to understand how a product is going to perform when they’re under conditions of national consequence,” Col. Trent said.
Further, Col. Trent said that he believes adversarial interaction today is a much lower-scale priority than designing the AI. He explained that with AI, JAIC wants to take a step-by-step approach to building up AI functions, because without careful considerations in the testing, problems can be multiplied heavily. Understanding how the model is evolving is key to doing any online learning, Col. Trent says.