The Navy wants to integrate robotics and artificial intelligence (AI) into more of its workload, from warfighting missions to non-combat support roles, and is seeking fleet input on what types of AI research are worth investing in.
The service already has, in varying stages of completion, its Unmanned Carrier-Launched Airborne Surveillance and Strike (UCLASS) in the air, a Large Diameter Unmanned Underwater Vehicle (LDUUV) under the sea and the Swarmboat unmanned vehicle on the surface. The warfighting value for these platforms is clear, the service said.
“Traditionally in the Department of the Navy the focus is on the warfighting mission, and rightfully so, but maybe not so much on the support side,” Bob Kozloski, deputy director of Task Force Innovation and deputy chief of the Office of Strategy and Innovation, told USNI News on Wednesday.
“So how does artificial intelligence and robotics fit into some of the operational support functions or even management?”
Kozloski’s office released a memo on June 5 outlining the path forward in identifying opportunities to integrate robotics and AI into Navy operations and for leveraging commercial development.
“The private sector is investing heavily in AI and robotics automation for decision-making and physical implementation tasks,” according to the memo.
“The DON could benefit from considering how to adapt recent private sector advances in fields such as machine learning, natural language processing, ontological engineering, and automated planning for naval applications.”
“When I see this memo, what I see is a demand signal from the very top to more heavily leverage what ONR (the Office of Naval Research) is already doing,” ONR director of research Lawrence Schuette said during the same interview. He added that Google has created some buzz by buying up companies working on AI and robotics, but the Navy had already invested in many of the companies bought by Google and others.
“A Google car drives well on a crowded busy street, but will it run off-road in shifting conditions with vehicle failures but yet must continue, which is really where the DoD environment’s going to go?” he said.
“So I think what you see is the DoD environment tends to be the commercial environment plus plus. And I think that’s really the difference. So we’re leveraging what they’re doing, they’re leveraging what we’ve provided them in early research, and then we’ve got to take it the rest of the way.”
Schuette said the Navy defines AI as “that thinking system that finds itself someplace it’s never been anticipated to be and is able to figure out what to do.” The goal of the memo is to get the fleet thinking about how this principle could be applied to their work – not just through unmanned planes or boats, but the basic idea of a computing system that learns as it works.
The Navy’s crowdsourcing website, Hatch, will today begin a 30-day campaign to solicit ideas from sailors, Kozloski said. The top ideas will be forwarded to a study group called for in the memo, and that group will analyze the ideas and report back to the Navy secretary by the end of the calendar year.
In addition to identifying opportunities to integrate AI and robotics, the group’s report will also include cost-effectiveness analysis for the ideas and resources needed to implement these ideas.
Concerns have been raised on how much autonomy to give weapon systems in the last several years.
In April 2013, the United Nations called for a moratorium on lethal autonomous systems, saying that systems with artificial intelligence should not make life-or-death decisions. The report mentioned the Northrop Grumman X-47B as a lethal system with some degree of autonomy.
However, in this current crowdsourcing effort, the Navy is seeking ideas for autonomy outside this idea of “lethal autonomous robotics” and more in line with sailors’ day-to-day functions.