The politics of the AI Action Plan

The Cabinet has split on a central economic question and the losers face years of irrelevance

Section 3 of the AI Opportunities Action Plan, launched by Keir Starmer on Monday, is a break with 40 years of failure in the UK.

The plan acknowledges that AI may not be world-shattering, but argues it is an asymmetric bet where the cost of missing out if it does live up to the hype is astronomical. To pursue this opportunity, it articulates policies of three kinds. Section 1 provides for computing and data infrastructure, talent and regulation. Section 2 promotes adoption in the public and private sectors. So far, so vanilla. In the four decades during which governments in the UK have limited themselves to this kind of pre-competitive, far-from-market policy work, Google, Meta, Nvidia, OpenAI and a host of other US giants have started up. But Section 3 is something else.

Section 3 says, “If this is to benefit the UK we must be an AI maker, not just an AI taker: we need companies at the frontier that will be our UK national champions.”

“Generating national champions will require a more activist approach and something more akin to Japan’s MITI or Singapore’s Economic Development Board in the 1960s, not the ‘invisible hand’.”

“Our goal should be a thriving domestic AI ecosystem, with serious players at multiple layers of the “AI stack” and widespread use of AI products and services across the economy.”

The vehicle for curating this ecosystem will be a new unit, Sovereign AI. “It will support the private and academic sectors in doing what they do best, with the ability to collaborate internationally, create joint ventures, as well as invest in, incubate and spin out AI companies – refining its strategy and approach as the technology matures.”

The detail of how all this will be achieved is understandable, so far as it goes. There’s no budget for Sovereign AI, nor is its institutional form resolved. A lot of questions will have to be resolved in the spending review and industrial strategy that are coming through in the spring.

The detail however is less important than the principle. “National champions” – with those words, Starmer’s plan plunges a sword through the heart of the consensus that has dominated industrial strategy in the UK since Margaret Thatcher.

For decades, neoliberals in both the Conservatives and Labour have insisted that the era of national champions is over. The globalists implemented a laissez-faire approach to takeovers that has allowed what were the UK’s national champions to be bought by overseas rivals, leaving us with a chronic lack of industrial scale and huge holes in our strategic capabilities. They have presided over two waves of de-industrialisation, first under Thatcher and then under New Labour. They have limited the role of the state to far-from-market research that by definition has little impact on the commercial prospects of firms. One result has been that good ideas bubbling out of our universities have nowhere to go and routinely end up delivering most of their jobs and tax revenues overseas. Another has been the division of the country into two halves, one prosperous, the other left behind. It has been failure upon failure.

In the plan, the national champions are not an end in themselves. Rather, they are imagined as part of a larger ecosystem covering all levels of the ‘technology stack’, the dynamism and growth of which is the key goal. The strategy therefore can be understood as the adoption by a country of the kind of ecosystem approach that has allowed Big Tech to dominate profitable niches, as Google dominates the adtech stack.

Template

Rachel Reeves said on launch, “This action plan is the government’s modern industrial strategy in action.” Thus the ‘eco-stack’ approach in the AI plan provides a template for other subsectors in the coming industrial strategy. It could, for example, be applied to engineering biology, which is similar to AI in that it is emerging from long development in academia into commercial relevance. On the day the AI plan was launched, the House of Lords science and technology committee also published a report with the provocative title, “Don’t fail to scale: seizing the opportunity of engineering biology”.

This approach could also be applied to the defence sector, which shares with AI the strategic dimension. Indeed, it could be applied to any industrial subsector in which the UK has a comparative advantage or strategic imperative. This is one way the UK could try to make sense of the strengths it has got and develop the kind of tech economy that drives economic growth for the country as a whole. Indeed, why invoke Japan’s MITI and Singapore’s Economic Development Board if there is no intention to echo their economy-wide activism? Why get us thinking of the economics of Mariana Mazzucato if you have no intention of following through?

This approach means Sovereign AI is positioning itself in AI as a rival to Big Tech (and other countries with sovereign AI ambitions). One aspect of that rivalry is that it will be competing for the interest of AI companies, as collaborator, investor and potentially long-term strategic shareholder. And for this, the government needs to move quickly to flesh out the bones of the plan and make itself credible in the biggest gold rush in history.

Here's what an exited founder of one of the UK’s successful AI start-ups told me after the plan came out. “There’s a lot of skepticism out there. In the WhatsApp founders’ groups, it’s about 50:50. Half are willing to give it a chance. Half are, ‘Fuck this, I’m off to Dubai’.”

Although the increase in employer’s National Insurance Contributions hasn’t helped, the sourness is primarily the legacy of years of sloganizing prevarication by Rishi Sunak. The author of the AI plan, entrepreneur Matt Clifford, has been at the centre of the government’s policymaking on AI for almost two years. He is a realist rather than an ideologue. The problem in developing a plan hasn’t been a lack of ideas or understanding – it’s been the lack of support at the political level.

Politics

A tactful way of making the same point is to say, “Ministers said the previous government had been focused on AI safety alone, rather than the opportunities presented by the technology.” Which is what PA reported at the launch of the plan. But why be diplomatic? Why not put the boot in?

The way to understand the politics of it, I think, is that the Blairites are in retreat and struggling to hold together their alliance with Conservatives such as William Hague, Greg Clark and David Willetts while simultaneously remaining relevant to Starmer’s direction of travel. Meanwhile, true to the original shape of Starmerism, the neo-Bevinite social democrats that are in fact driving the policy are bending over backwards not to embarrass Blair by exposing his irrelevance.

To substantiate those claims, start with the pre-existing Blair-Hague position. Here is an extract from an article I wrote six weeks ago criticising the way the Tony Blair Institute framed sovereign AI capabilities in its report Governing in the Age of AI:

[The TBI] recommends:

“Build sovereign AI capability in collaboration with trusted private-sector partners: government will need to create its own models, fine-tune existing ones or build tools on top of off-the-shelf LLMs, deciding on the appropriate approach and the best foundation model for each use case, to ensure that it can: first, train its own custom LLM for national-security purposes (which for the purposes of this report we name CrownIntel) on open and official data, and fine-tune it on sensitive and confidential information in a secure environment; second, create a ChatGB “legal advisor” tool for government by fine-tuning an off-the-shelf GPT-4 class model on legislation and parliamentary records; third, work with individual departments’ digital teams to build or procure AI tools using commercial foundation models. These tools for ministers, officials, caseworkers and analysts would be bespoke to each department’s individual use cases and could number in the hundreds or thousands."

This is an approach that builds in a dependency on Big Tech as the gateways to the main Large Language Models. Presumably because the security services have insisted on it, under this plan the UK will in fact produce its own LLM, CrownIntel. But this effort is to be siloed in such a way that it can have zero commercial implications. No effort is made to envision how a £10 billion effort by the state in a rapidly moving technology in which the UK has strong position in the basic science is to be leveraged into British companies and British jobs.

As I said then, a better approach would be along the lines that have now emerged in the AI plan, to aim for “a monopolistic ecosystem in which the technology stack is curated by the state rather than Silicon Valley”.

So, Blair and Hague can only endorse Starmer’s plan by repudiating their own approach. Faced with such an unappealing choice, they have done the thing that drives the rest of us crazy with frustration and fudged it. In a joint article in the Times, they welcomed the plan but managed to avoid mentioning any of the provisions in Section 3. Indeed, there is no mention of industrial strategy. It’s all about AI in public services. But then, at the end, the very last line says, “nothing less than a relentless focus on delivering Matt Clifford’s plan will do”.

After decades in which they scorned state intervention in the market, with profound consequences for the country, one line. Is that it? Has Blair truly become an interventionist? Does he now repudiate the approach that New Labour took to the economy? What about Hague? Is he going to defend Section 3 in the House of Lords while the likes of Peter Lilley and David Frost accuse him of being a socialist?

Rather than a change of heart, it seems to me that, in order to claim the position of AI Ultras, Blair and Hague want to avoid arguing against Clifford’s plan. At the same time, they cannot bring themselves to endorse the provisions of Section 3. And to square the circle, they have opted for a kind of insincerity.

Kyle’s statements, including his introduction to the plan, are only slightly less grudging. Again, you will not find any mention. of industrial strategy, let alone explicit support for the provisions in Section 3. Thus it seems likely that Section 3 has been imposed on Kyle. By, it seems, the combination of Clifford and Reeves.

Sovereign AI is an argument that Blair and Hague have lost and I have won. I have won not because I’m guru to Clifford or Reeves but because it is the blindingly obvious thing for the UK to do – unless, that is, you’re wedded to an outdated, neoliberal, Thatcherite ideological dogma that insists the government stay out of supporting research that is close to the market.

The need to manage the factional tensions has created a fog of confusion. In communicating the new policy, Kyle avoids talking about Section 3 and softens the impact on his allies with milquetoast criticisms, further diluted through being made unattributable. At the same time, Blair and Hague obscure their defeat with insincerity. And Starmer diplomatically goes along with Kyle’s soporific language and tactful focus on public services. The overall effect is to give the impression that little is changing at exactly the moment when there is in fact a substantial change of direction. For many, the policy has come off as either routine government support for techy things or just another round of AI boosterism.

Starmer has made his choice. The status quo of four decades has been broken. National champions and MITI are back on the menu, albeit in a novel way. In this way, he has given the lie to the idea that he is a mere cipher. It turns out that, even when he has ceded a department to the Blairites, he is ready to impose his own economic outlook.

Blair’s attempts to hold the line have failed and there is no way back to relevance for him. In industrial strategy, he is now become like the hollow men of T.S. Elliot’s poem, dead inside:

…

Between the desire

And the spasm

Between the potency

And the existence

Between the essence

And the descent

Falls the Shadow

                              For Thine is the Kingdom

For Thine is

Life is

For Thine is the

This is the way the world ends

This is the way the world ends

This is the way the world ends

Not with a bang but with a whimper.