Taking AI to the Edge in Defence
Planning my team's activities for the coming year has required us to do some thinking about the near-term activities we can do, to help ensure that UK Defence can exploit AI at scale and pace.
While conversations around the cutting-edge of AI tend to focus on the performance of the latest 'frontier' AI models, exploiting AI for competitive advantage in Defence requires us to push forward in a number of directions, including people and processes as well as technology.
Recently I've been describing these different directions as the four 'edges' of AI in Defence. These four edges are a deliberate simplification, intended as a way to articulate at high-level what is needed to push AI forward in Defence. They are not intended as an exhaustive list or a deliverable plan.
The idea is that to truly exploit AI for advantage, AI will need to push forward on all four edges. Note that, as with all content on this site, what follows is personal reflection not a statement of MOD policy or intent.
- The Technological Edge
Although model performance isn't everything, Defence nonetheless needs to be able to exploit both frontier AI models and new ways to exploit these models (agentics, multi-modal etc). And - while the debate around Artificial General Intelligence (AGI) and its potential to disrupt defence and security is somewhat beyond the scope of this post - it's clear that Defence needs to stay abreast of AI developments.
This means increasing the speed at which we can deploy transformational AI into Defence, so we can keep pace as the frontier moves.
- The Operational and Tactical Edge
Defence needs to get AI into the hands of warfighters, whether in HQs or closer to the frontline. That means dependable tools for warfighters at every level and in every domain. It also means using AI far away from the usual public cloud services, requiring useful, robust AI tools that can be surfaced in rugged portable devices and that can function in deployed, disconnected or degraded environments.
This is not a trivial ask, likely involving fundamental changes to secure networks and upgrades to legacy systems as well as edge devices. Note that bringing AI to the tactical edge is not the same thing as 'edge compute' although the latter is likely an important part of the former.
- The Organisational, Doctrinal and Process Edge
To thrive in an AI age, Defence needs to be an AI-native organisation that operates to an AI-enabled tempo and mindset. We will need answers to the questions that this shift poses: Is AI adoption widespread to the tips of the fingers, teeth and arms of the whole force or centred around a few pioneers? Do our people - from the front-line to the back office - have the skills they need to use AI safely? Are we using AI to fight, train and work in new ways - or just using it to speed up the old ways?
Are UK Commanders confident taking decisions based on the outputs of AI tools - or do they just use it to confirm their existing biases? AI will need to be baked into our concepts, doctrine and training as we evolve.
- The Regulatory and Ethical Edge
The UK has well-established ethical principles for the use of AI in Defence. But potential adversaries could draw their ethical lines in a very different place. This does not mean stepping outside our ethical principles - but it places greater emphasis on our AI safety and assurance tools and processes, which will need to evolve rapidly as AI technology, regulation and practice move forward.
By way of analogy: The RAF has a strong culture of aviation safety. This dedication to safety does not prevent the RAF from flying - quite the opposite. In fact it is one of the things that enables the RAF to fly in difficult and dangerous circumstances, beyond the envelope of civil aviation. It may feel uncomfortable, but Defence likely needs to adopt a similar posture with AI. The better your guardrails, the closer to the edge you can skate.