The RN is hosting a panel session – chaired by Commodore Rachel Singleton, head of the Defence AI Centre at UK Strategic Command – addressing the role of AI in how the navy can exploit integrated technologies to improve capabilities.
Beyond DSEI’s discussions, the RN is already working to define its requirements for new technologies like AI, especially in enabling the navy to enhance its operational impact.
“What we need from AI is not AI itself. It’s a means to an end, and the end is autonomy,” Rear Admiral James Parkin, the RN’s Director Develop, told Naval News in an interview on 24 August.
“Autonomy means, effectively, no human beings in the loop. In cases where you need a human being in the loop for ethical purposes, AI provides decision-grade information to allow that human being to have the cognitive burden that’s suitable to what our brains can provide.”
Rear Admiral James Parkin, the Royal Navy’s Director Develop
In sum, Rear Adm Parkin continued, the navy is seeking to use AI to harness data and enable decisions to be taken at the ‘speed of relevance’ to support operational requirements.
In previous times, sailors working in an ‘ops room’ faced the challenge of determining whether an incoming target was a missile or an aircraft (in the latter instance, possibly a civilian aircraft). Today, Rear Adm Parkin explained, technology has helped solve that challenge, although with the solution presenting its own issue. “The problem is resolving itself with the proliferation of complex, high-speed weaponry – but it means we have to be able to respond in a high-threat, high-speed world,” he said. “AI is going to be able to do all of that and deliver autonomy in wartime, whilst also helping us operate autonomously in peacetime.”
Addressing the threat posed by crewed or uncrewed platforms operating in swarms is a task prevalent in both wartime and peacetime. It can be used to counter the threat in both contexts.
“If you have a swarm of 20 reconnaissance aircraft, you don’t need 20 pilots on the ground with a joystick. It can be done by one, or it can be done by none, in order to deliver the effect.”
Rear Admiral James Parkin, the Royal Navy’s Director Develop
“Previously, we’ve been concentrating on how we deliver output-based AI; now, I think we’re at the stage where we can define outcome-based AI,” Rear Adm Parkin continued. Here, he explained, an AI-enabled uncrewed system can be given a task to complete, with a set of parameters directing the system to only refer back to the operator if a prescribed level of ethical or ‘difficult-to-understand’ circumstances is encountered.
For example, he explained, using AI to ‘pattern spot’ in a task like intelligence, surveillance, and reconnaissance (ISR) can be an optimal application of AI capability, analysing raw data to exploit and disseminate information. At sea, such capability could evaluate sonar data in the underwater domain, or conduct bulk analysis of radar and automatic identification system (AIS) track pictures on the surface.
“You can point out that the track saying something is a merchant vessel was saying it was something else seven weeks ago,” said Rear Adm Parkin. “The system’s been tracking it electronically without any human intervention, and noted there was something wrong.”
“It’s what AI is good at doing – detecting the presence of the abnormal, or the absence of the normal, allowing the human being to be freed up to make the decisions that only humans can do,” he continued. AI also can be used to monitor lower-level, but mission-critical, tasks: for example, equipment and platform condition monitoring to assess fatigue; or optimising command-and-control network connections and integration. In the latter instance, Rear Adm Parkin explained, AI will enable the systems themselves to understand how to improve connections in system and network management terms.
