The Future of Defense Task Force says the Defense Department needs a Manhattan Project for artificial intelligence, but the department may be thinking about AI the wrong way. Bryan Clark, former special assistant to the Chief of Naval Operations and former director of the CNO’s Commander’s Action Group, discussed this on “Government Matters” Wednesday.
In an article in Breaking Defense, Clark and his colleague Dan Patt explain why artificial intelligence should be thought of as a technique as opposed to a technology. “What we’ve seen in the Department of Defense is a trend towards treating AI as a system or as a technology that you buy,” Clark said. “The problem is, AI is really just a technique. It’s an information processing methodology.”
Clark explained that artificial intelligence is able to provide similar capabilities as a human but in a very narrow form; i.e., it only performs one task at a time. “The idea is to start treating AI as a form of intelligence like you would treat the subordinates on your military team,” Clark said. And with AI handling something like maneuvering an airplane, for example, this would leave pilots free to focus more on using higher reasoning to achieve victory over the adversary.
Despite the fact that AI is now proliferated widely throughout the military, Clark explained that “those R&D efforts aren’t really transitioning into operationally useful applications very quickly.” This is why, Clark argues, a paradigm shift from thinking of AI as a “thing” to thinking of it as a methodology is necessary.
In addition, it is essential to come up with “organizational changes that allow these R&D efforts to transition into an acquisition system,” Clark said. He pointed to what the Joint Artificial Intelligence Center (JAIC) is doing as a good example of this process, where they are acquiring AI services to incorporate into software.