A Serious Strategy
... choosing to compete in AI
TL;DR: A good national AI strategy is about comparative advantage. It is not about trying to outspend the US or out-scale China. The UK already possesses world-class institutions, science, professional expertise and trusted public assets. We should organise national effort around those strengths and make deliberate choices about where the UK can lead.
I had a good deal of my political formation in the 1970s, and my instincts still bear the marks of that period. One legacy is a scepticism about government intervention in industry and technology, and an inclination towards a bottom-up, entrepreneur-led, market-oriented approach. In that view, the role of the state is largely to remove obstacles and otherwise stay out of the way.
I have come, although not completely, to moderate that position. National advantage in advanced technology sometimes needs coordinated investment at scale, sustained support for research, market-shaping interventions that help innovation cross the gap into deployment, regulation that enables rather than frustrates that transition, long-term investment in skills and talent, the removal of red tape, and intelligent use of government as both purchaser and first user. With science and technology understood as central to economic prosperity, national security and societal resilience, the case for an active national technology strategy is, I believe, overwhelming.
Of course, bad strategy is often worse than no strategy at all. Too much industrial and technology strategy is an attempt to satisfy every constituency simultaneously. Every hot technology appears and every sector is deemed strategically important. The aim is that each stakeholder finds something they recognise. The result is comprehensive, extensively analysed and carefully balanced, but often devoid of the central characteristic that distinguishes strategy: the readiness to make choices.
Strategy is fundamentally about comparative advantage. It requires an understanding of where a nation possesses capabilities that others do not, where those capabilities can realistically be strengthened, and where scarce resources are unlikely to produce reasonable returns. AI provides an excellent opportunity to think in those terms.
Both the previous and the most recent UK Governments have, quite properly, placed AI centrally in their economic thinking. Some progress, under the aegis of the now-defunct Department for Science, Innovation & Technology (DSIT), has already been made: the AI Security Institute (AISI) has established an international reputation; and public investment in compute capability has increased. The AI Opportunities Action Plan has sought to establish important elements of the supporting infrastructure including AI Growth Zones and the National Data Library. The new Government has moved with some rapidity to reaffirm AI as a central priority, establishing a Prime Ministerial AI Taskforce chaired by Lord Vallance and strengthening leadership at the political centre of government with a Minister for Artificial Intelligence, Kanishka Narayan, attending Cabinet.
This is all well and good as a demonstration of seriousness of intent, but what remains unclear is the strategic proposition that binds the initiatives and intentions together. It cannot simply be to build larger models than the US or deploy more compute than China. Those are competitions in which others enjoy structural advantages that will be difficult to overcome. Success cannot be measured by the number of initiatives announced or institutions established – strategy is not an inventory of activity.
The UK begins from a position of some strength – we have a range of valuable national assets that we can leverage and that can anchor our choices. With this in mind here is the @profserious 10-point (inevitably) Serious Strategy.
1. Make the NHS the world’s leading AI-enabled health platform
The NHS possesses something that almost no competitor can match, that is longitudinal, population-scale health data. Combined with assets such as Genomics England and UK Biobank, this creates extraordinary opportunities in diagnostics, personalised medicine, drug discovery and the optimisation of healthcare pathways. Clearly, the challenge is governance, trust and public legitimacy. Trusted research environments, privacy protections and arrangements that ensure the UK captures a share of the value created are a vital part of this. If we can address the challenge, health and life sciences could become one of the UK’s defining AI strengths.
2. Make the UK the world’s leading AI-enabled higher education system
UK universities are amongst our greatest strategic assets, although current financial sustainability pressures place that position at risk. They combine internationally recognised research, global educational reach, deep international partnerships, and a continuing ability, policy permitting, to attract talented students and researchers from around the world. Higher education is itself one of the UK’s most successful export sectors and an important source of economic value, international influence and soft power. The UK should set itself an ambitious objective: to become the world’s leading system for AI-enabled higher education. AI creates the opportunity for genuinely personalised learning, richer student support, new forms of professional and lifelong education, more effective assessment and feedback, and educational experiences that combine the best of human teaching with intelligent technology. The prize is to establish UK higher education as the international benchmark for the thoughtful application of AI, strengthening universities financially, extending the UK’s educational export advantage and soft power, and creating a distinctive national capability that few countries could readily replicate.
3. Make the UK’s professional economy the world’s leading adopter of AI
The UK’s professional economy – law, consulting, accountancy, insurance, architecture and other knowledge-intensive services – is already a globally successful export built on expertise, trust, judgement and long-established institutions, all of which make fertile ground for deploying AI. Financial services deserve particular attention: London’s combination of deep capital markets, regulatory sophistication, common law and global reach opens applications across many areas including compliance, fraud detection, underwriting, risk modelling and market supervision, as well as the chance to build entirely new international markets – notably the insurance of AI-related risks, drawing on London’s strengths in specialist insurance and financial innovation. Government’s role is to create the conditions for confident innovation through procurement, liability frameworks, regulatory clarity, and support from professional bodies; if the UK becomes the place where AI is deployed most effectively across professional services, it will have secured an enduring comparative advantage that extends well beyond the technology itself.
4. Establish cyber defence as a national AI specialism
The UK has recognised strengths in cyber security anchored in national capabilities, notably GCHQ and the National Cyber Security Centre (NCSC). As the Director of GCHQ recently set out in her Bletchley speech, AI is beginning to enable adaptive defensive cyber systems able to learn continuously from attacks and respond at machine speed, and there is an emergent blueprint for developing a national ‘Cyber Shield’ based on this. It is an area in which the UK possesses both technical understanding and institutional credibility. Developing AI-enabled defensive cyber technologies into a recognised national specialism would strengthen national resilience, reinforce wider defence capability and create valuable export opportunities in an area of growing international demand.
5. Create an AI assurance industry
The UK’s work on AI safety has established an international reputation. The next step should be to convert that reputation into an economic capability. Evaluation, testing, certification, independent audit and assurance could become internationally significant professional services in much the same way that the UK has previously developed strengths in financial audit, standards and specialist insurance. Trusted deployment is likely to become an increasingly valuable capability as AI becomes embedded within key sectors of the global economy.
6. Turn regulation into a strategic asset
We have experienced regulators and a flexible common law system that has repeatedly demonstrated the capacity to respond to technological change. Much international discussion has presented regulation as though it were principally a constraint upon innovation. That may have some elements of truth but is too narrow a view. Well-designed regulation reduces uncertainty, increases confidence and enables investment. In highly regulated sectors it is often a precondition for innovation. The UK has an opportunity to become the best place in the world to develop and deploy AI in regulated environments. Health, financial services, legal services and other professions all depend upon trust, accountability and clear standards. A regulatory environment that combines predictability with flexibility could become a significant source of comparative advantage.
7. Use government as an intelligent anchor customer
Government is uniquely placed to accelerate markets through procurement. The NHS, HMRC, DWP, Defence and other major public services collectively represent one of the largest potential customers for advanced AI systems anywhere in the world. Thoughtful procurement can improve public services while simultaneously helping UK firms to develop products, demonstrate capability and achieve international scale. Government also has an opportunity to become an exemplar of effective AI deployment within complex organisations, demonstrating how AI can improve decision-making, productivity, and citizen services whilst securing public trust. It has often underestimated, perhaps squandered, the strategic value of its own purchasing power.
8. Build upon the UK’s scientific strengths
AI has the potential to accelerate scientific discovery across a very broad range of disciplines. The UK should concentrate where it already possesses internationally recognised excellence. Weather prediction through the Met Office, life sciences, engineering, advanced materials and fusion research all offer opportunities in which AI can grow existing capability and strengthen established leadership. The objective should be to compound strengths that already exist through the application of AI. Strategy rarely succeeds by attempting to create entirely new areas of excellence from nothing, but rather by extending those that already exist or are at any rate nascent.
9. Make the UK’s creative industries AI leaders
The UK’s creative industries are amongst our great economic and cultural strengths. Film, television, music, publishing, advertising, architecture, design, fashion and games already combine creativity with technology to produce globally successful businesses and international cultural influence. AI will inevitably become part of that creative ecosystem. The issue is not to protect intellectual property, however loud those voices are, but rather to ensure that British creators and creative businesses are amongst the first to exploit AI to increase productivity, develop new forms of creative expression and reach global markets. The UK could reasonably aspire to become the natural home for the combination of creative talent and AI, demonstrating that technological innovation and uniquely human creativity can be mutually reinforcing.
10. Build AI capability through professional institutions
Almost every technology strategy eventually arrives at a proposal for reskilling. The UK possesses something that it consistently undersells: chartered professional bodies and well-established institutions responsible for maintaining professional standards. These organisations should become central to the development, certification and continual renewal of AI capability across the workforce. AI will reshape professions rather than replace them, and professional institutions are well placed to ensure that standards change alongside technology. Embedding AI within established professions is likely to produce capability that is deeper, more trusted and more enduring than a succession of short-term training initiatives.
Taken individually, none of these proposals is particularly novel but their potential lies in the way they way they could interlock and deliver a strategy built upon comparative advantage rather than technological fashion. Each starts with capabilities that the UK already possesses and asks how AI can strengthen, extend and compound them over the coming decade. There are, of course, other potential candidates. Defence technology, space, quantum technologies, advanced manufacturing and energy could be in the mix. Strategy is not, however, about producing the longest possible list.
AI is the immediate subject of this article, but the strategic questions it raises run wider. They concern the capabilities the UK intends to sustain, the institutions on which it means to build, and where it chooses to compete. These are questions of national strategy as much as technology policy. Whether the Serious Strategy I have sketched is the right one is, in a sense, beside the point. Reasonable people (and @profserious subscribers are, of course, reasonable) will disagree about the priorities – they will favour different sectors, or weigh particular opportunities differently – and that is as it should be. What matters is that we make strategic choices at all – AI is too consequential to be met with a collection of worthwhile initiatives or an accumulation of announcements.
I agree very much with the central proposition here: strategy requires choices. We can aspire to do almost anything, but we cannot do everything.
Two other things I read this week made me wonder whether we need to push that argument further.
The first was the New York Times piece about leading Google DeepMind researchers leaving to establish a new AI research company, The Discovery Loop. One of them, Oriol Vinyals, explained the attraction of creating their own organisation: “Having extreme focus is very powerful.” He went on to say that the ultimate form of extreme focus comes when the objective is the only thing that matters and aligns completely with the mission of the company.
He was talking about companies, not countries, but I think the principle travels. In something moving as rapidly and competitively as AI, strategy without extreme focus is unlikely to be effective.
That leaves the harder question: where do we focus?
There is an old ice-hockey maxim, usually attributed to Wayne Gretzky, that the trick is to skate to where the puck is going to be. Comparative advantage cannot simply be an audit of what Britain is good at today. We have to make some bets about where technology is going.
Which brings me to the other Google story that caught my attention. Google engineers have demonstrated real-time speech translation, entirely offline, using a Gemma model running on a Raspberry Pi 5.
The translation isn't the interesting bit. The interesting bit is where it is happening.
We talk about AI in terms of enormous models, hyperscale data centres, huge energy requirements and billions of dollars of capital. Britain is unlikely to beat the Americans or Chinese at that game.
But another puck may be moving in the opposite direction: small, inexpensive, specialised AI running locally.
Put sufficient intelligence into a cheap embedded computer and the possibilities become enormous. Almost every manufacturing process could benefit from embedded intelligence observing and optimising it. So could agricultural machinery, laboratory instruments, medical equipment, energy systems and infrastructure. Most don't need the world's most intelligent model. They need enough intelligence, specialised for the job, cheaply and reliably where the work happens.
And I can't resist noticing what Google chose to run its demonstration on: a Raspberry Pi.
There are echoes here of a remarkable Cambridge computing lineage: Acorn, the BBC Micro, ARM and Raspberry Pi. I owe my own career in IT in no small part to the BBC computer.
That's not an argument for industrial policy based on nostalgia. It is evidence of an engineering tradition around doing remarkable things with constrained, inexpensive hardware. If that is where part of the puck is heading, we should at least ask whether it is somewhere Britain could skate.
There is another comparative advantage running through much of the Serious Strategy: Britain's reputation for engineering, standards and trusted governance.
Our engineering institutions, universities, professional and chartered bodies, regulators and common-law tradition have accumulated international credibility over generations. In a world where AI will increasingly need assurance, certification, professional standards and governance, that reputation has economic value.
We should treat it as national infrastructure. Government should recognise it, celebrate it and build upon it. You can buy GPUs and build a data centre. You cannot buy two centuries of accumulated institutional trust.
None of this means Whitehall should decide that embedded AI is the answer and commission a five-year plan to build it.
Successful strategy needs to be simultaneously top-down and bottom-up.
Government should set direction, concentrate resources on a small number of serious bets, fund the underlying science and infrastructure, remove obstacles and, as you suggest, become an intelligent first customer.
Then let universities, entrepreneurs, engineers and companies discover what actually works.
And evaluate the results ruthlessly. Fund experiments, measure them, stop doing things that don't work and put substantially more behind the things that do. We need something closer to the discipline of venture capital than the governmental temptation to keep every initiative alive because somebody has acquired an interest in it.
Extreme focus must not become extreme stubbornness.
Perhaps that gives us three useful tests for a serious technology strategy: focused direction, distributed experimentation and ruthless evaluation.
Britain can aspire to do almost anything. We cannot do everything. We need to choose a small number of places where our inherited advantages intersect with where we think the puck is going, and pursue them with something approaching Vinyals's “extreme focus”.
Otherwise we risk repeating a familiar British story: contributing disproportionately to creating the future, while capturing rather too little of the capability, industry and economic value it creates.
I wonder whether a strategy should have 10 points or 3-5. With 10 points, I think that the focus is already lost. An important aspect concerns resources. What is the main objective of the strategy, and what are the deployable resources? These have to be identified and separated. For instance, the NHS is (to me) a resource which has large data sets, a huge budget, many competent staff, and can be used both as a testbed and demonstrator and as a market. However, its budget and staff are highly constrained by the NHS' primary mission.