Evaluate the implications of transitioning to a mainly autonomous and uncrewed fleet for the components of naval fighting power (physical, conceptual and moral).

Transitioning to autonomy will occur over three main phases; As-is, Transition Phase and Autonomy as Usual. The Transition Phase will be the most challenging, as a significant amount of energy is expended for seemingly little result. Data collection, cleaning, processing and labelling systems will need to be set up, MLOps processes set to work, training completed across the RN and live trials and validation/verification studies completed. This ‘latent heat’ will take time to show results. During the Transition Phase, the RN will need to spend more resources to achieve seemingly the same day-to-day results, but must do so with the knowledge that these AI Building Blocks will manifest large benefits.

Transitioning to an autonomous fleet will predominantly result in dramatically reduced costs in the Physical component. These Physical savings of time and money should be used to enable the changes and improvements that the Moral and Conceptual components of an uncrewed fleet will require. The presumed temptation is that savings will be used to reduce headcount; but a better alternative would be to invest in the resilience of AI/ML systems, AI/ML training and providing the headspace for RN personnel to undertake the wargaming and leadership training that will be required to get the best out of these new Physical capabilities.

Physical Component. Ships’ Companies will monitor and deliver operational effects from distributed control centres. This means sailors that previously spent 6 months on deployment will rarely spend time away from home. Rather, ‘Front Line’ personnel may actually conduct watch rotations, in far smaller teams (~30 for a fleet of Uncrewed Surface Vessels (USV)), travelling home while off-watch. Maintenance will be managed at remote bases closer to the AOO, undertaken by either RN maintainers or USV manufacturers on a cheaper, ‘Platform/Effect as a Service’ subscription-style basis. With simple systems produced at scale, sustainability of the physical systems should improve.

Subsequently, all RN personnel will need training in AI/ML techniques and capabilities, potentially in a way similar to First Aid training today; everyone receives basic AI/ML training, with more detailed technical training for platform operators. At higher levels, Role 1, 2 and 3 AI/ML production,  maintenance or validation/verification/testing services, akin to Role 1-3 medical facilities, will provide ever more specialist levels of advice, potentially with Role 3 AI/ML Maintenance Facilities brought to readiness during high-tempo operations. All will need to take time to understand likely AI/ML failure modes, validation and verification techniques and attack vectors into AI/ML systems. Cyber defence skills will also need to be reinforced to protect autonomous systems from remote attack. AI/ML training will not only be important for operational capability reasons; they will also provide vital Motivation Factors for technically-minded individuals to continue learning and deploy cutting-edge skills, reducing the risk of voluntary outflow.

Pay will need to be carefully considered. Remote systems operators may demand higher compensation for the greater effect-per-person they deliver and the higher technical skillset they hold; while those in roles that cannot be physically removed from the Front Line (e.g. Boarding, CASD) will demand comparatively higher compensation for the extra physical risk and time away they experience. The reduction in overall costs that autonomous systems should bring may allow both groups to be adequately rewarded; but this may not be possible until real savings emerge during the Autonomous as Usual Phase. Reward during the Transition Phase will have to be managed carefully, as the ‘Total Reward’ recommendations in the Haythornethwaite Review make clear. The RN Clearance Diver Branch is arguably undertaking this transition first, as divers are replaced by autonomous systems.

Conceptual. The likely Conceptual changes are already broadly understood. Affordable mass and the use of many disposable platforms should result in improved Situational Awareness and in Commanders being willing to take increased risk with regards to autonomous platform safety. Following an extended period in the RN’s history where there have been so few warships that they have all essentially needed to be treated as a Mission Essential Unit, this should come as a welcome opportunity to reiterate core doctrine, especially the idea that units can be sacrificed if it results in a battle-winning opportunity, e.g. unlocking the ability to effectively fire first. Conversely, a deeper understanding of AI/ML ethics, particularly understanding the consequences of decision-making when seemingly innocuous parts of the Kill Chain have been automated, will also be required; if three stages of your sensor collection and processing chain only have 95% accuracy, how will that change your target identification process?

The lower demands on Ship Company time (due to being able to work from home port) should result in more opportunities to undertake professional education, leadership training, experimental wargaming (or, with cheaper platforms, even real-world, live-fire wargames) as formed units or Divisions. The savings delivered by autonomous systems should be reinvested in MWC, both in headcount and high-fidelity wargame environments; especially environments that allow full AI/ML decision-making and cyber-attack interventions to be played out. This will be vital. Not only will new ways of fighting with current autonomous systems need to be learned, but rapid experimentation, understanding and dissemination of tactics will be fundamental if the RN is to keep pace as these autonomous technologies improve at logarithmic rates. New AI/ML-powered autonomous systems will also require data of real-world decision-making for training purposes; only a vast quantity of Human-in-the-Loop wargames can provide this (perhaps with wargame data shared among AUKUS allies). Failure to reinvest savings here will see the RN fall even further behind the ‘pace of relevance’ at an increasing speed.

Again, the Transition Period will be challenging. Initial models (e.g. object detection models that automate EO/IR detection) will take time to improve, and will need human oversight for a number of years to come. But an understanding of the potential unlocked by these models when they reach full potential (perhaps up to 90% precision/recall scores) needs to be understood now; as does an understanding of how the RN will deal with the 10% of remaining edge cases that AI/ML will not be able to handle.

Moral. The opportunities for physical leadership of large teams, teams experiencing physical danger and those requiring inspirational acts of courage will be reduced. The opportunity to develop and demonstrate human leadership skills will fall, as fewer individuals will be required to deliver the same effect. Indeed, a core component of RN Moral strength -that all individuals are exposed to the same level of physical harm- will greatly reduce. More trust will need to be placed in the ability of contractors to create and maintain robust, efficient and capable platforms, as RN personnel will be operating them from a distance and less likely to be conducting maintenance. However, were the RN to assume a greater responsibility for autonomous platform maintenance, these Moral issues would be resolved. A revamped WE and ME Branch would deal with the maintenance of vast numbers of autonomous vessels, providing fast turnaround services in upthreat, shore-based locations (akin to a shore-based RFA Repair Ship). This may not save as much money as fully contractor-provided support, but would provide the RN with greater confidence in and expertise of their autonomous fleets. It would also maintain the conditions under which traditional RN cultural strengths (strong leadership of technical experts in physically dangerous, isolated and tiring conditions) could continue to be leveraged. In any event, a culture that provides the ability to empathetically and effectively lead smaller, highly technical, diverse teams in solving complex, fast-changing problems will be even more important than to the RN of today (not unlike modern civilian tech start-ups). Helpfully, the growth of autonomy, both for uncrewed platforms and use of AI/ML to automate mundane back-office functions, should lead to more time to undertake high-value leadership training; provided that savings accruing to autonomy are not seen as carte-blanche to reduce headcount.