Boeing’s submarine could be equipped with an AI
Boeing
The US Navy is quietly developing armed robot submarines controlled by onboard artificial intelligence. The vessels could potentially kill without explicit human control.
The Office of Naval Research is carrying out the project, known as CLAWS, which it describes in budget documents as an autonomous undersea weapon system for clandestine use. CLAWS will “increase mission areas into kinetic effects”, say the documents – military-speak for destroying things.
In particular, CLAWS will equip robot submarines with sensors and algorithms to carry out complex missions on their own, such as evading anti-submarine systems and finding targets, just as a human submarine commander would.
The details of the project are secret, only provided to US congressional committees. The US Navy hasn’t even said what the acronym CLAWS stands for, and declined to comment when approached by New Scientist.
Such submarines would be a step up from current capabilities, which include two kinds of craft. There are small, uncrewed underwater vehicles (UUVs) that are widely used for surveys and mine disposal. These are controlled directly by an operator via a communication tether because radio signals can’t travel through water.
Limited intelligence
There are also existing autonomous submarines that have limited intelligence, so can only carry out simple tasks without human guidance, like placing a sensor on the seabed at a particular location. More complex tasks require a remote operator and a communications link.
However, the US Navy has recently ordered much larger, long-range robot submarines made by Boeing, called Orca. Images of the new vehicle suggest it will be around 25 metres long and weigh more than 50 tons. It will be capable of 90-day missions before running out of fuel.
The CLAWS project was previously revealed in 2018, when the US Navy requested proposals to “improve the autonomy and survivability of large and extra-large unmanned underwater vehicles”, which would include Orca. But there was no mention then of weapons, only a requirement for sensing and decision-making, giving the impression that CLAWS submarines would be used for gathering ocean data.
One image of future plans for the Orca robot sub shows it armed with 12 torpedo tubes. It was assumed that these would be under human control, but with CLAWS, Orca could find and sink targets on its own. Essentially it could act as an intelligent naval mine in an area frequented by enemy vessels that attacks anything which passes.
Stuart Russell, a computer scientist and campaigner against killer robots at the University of California, Berkeley, sees CLAWS as a dangerous development. Unlike mines, robot submarines have potentially unlimited range.
Significant step
“Equipping a fully autonomous vehicle with lethal weapons is a significant step, and one that risks accidental escalation in a way that does not apply to sea mines,” says Russell.
Submarine expert H.I. Sutton, editor of underwater warfare website Covert Shores, also sees a potential danger.
“One challenge with weaponising UUVs will be identifying the target,” says Sutton, meaning an increased risk of killing friendly forces or non-combatants. “This problem already exists with mines but will be exacerbated by the increased levels of autonomy.”
CLAWS is allocated $26 million in this year’s US Navy budget and another $23m next year. This is expected to take it to the level of a working prototype. If successful, CLAWS could then be deployed on Orcas and other large robot submarines, which are expected to be delivered from June 2022.
Want to get a newsletter on all the latest advances in AI? Register your interest and you’ll be one of the first to receive it when it launches.