You’d be forgiven for thinking the government’s new AI Opportunities Action Plan lacks ambition. While frontier AI businesses in the US and China are developing dazzlingly powerful AI tools to cure diseases and solve mind-bending equations in physics, Sir Keir Starmer today promised that artificial intelligence would ‘spot potholes more quickly’.
Leaving this Starmerism aside, there is lots of detail in the plan to get tech companies excited. The creation of so-called ‘AI growth zones’ to build tech infrastructure faster and attract clusters of AI expertise in targeted locations brings huge promise. The history of technological advances, from social media to semiconductors, is often a story of bright, like-minded individuals firing off ideas in close proximity.
The AI Action Plan shows that Starmer is a man in a hurry
The creation of a new national data library to support AI development is also welcome; while a new dedicated AI energy council shows a degree of seriousness in the making this work happen. But to deliver on this plan there will be plenty of bumps in the road for Labour to that even AI won’t detect.
The Action Plan is one of the first meaty policy announcements the new Starmer government has made, offering a valuable glimpse into Labour’s broader economic strategy. It shows that the traditional Labour refrain of spending to grow the economy is not on the cards. There are no major spending commitments in this plan: there are a few general pledges to increase the government’s AI work and computing power, but no one puts a figure on these. That is in stark contrast to Biden’s CHIPS act in the US, for example, which pledged $50 billion (£41 billion) in subsidies to build semiconductor foundries.
It’s quite clear that, in the wake of rising gilt yields and the emergence of bond vigilantes, no one wants to rock the boat by betting growth on greater borrowing. That has focused minds on other policy levers, such as slashing red tape and removing the planning frictions which stop businesses growing. For a lot of tech firms, that is no bad thing.
It’s clear that much stronger ties with the EU is also off the table for the foreseeable. In a not especially subtle nod to this today, Starmer has said that ‘we don’t need to walk down a US or an EU path on AI regulation – we can go our own way, taking a distinctively British approach’. That is also a signal that the UK wants to steer clear of emulating the EU’s onerous AI Act. This, too, is good news for UK tech. A lot of the big US AI firms have been pretty vocal about the problems with the Act (even Europhile Nick Clegg has had his gripes), and UK AI firms tell me the EU’s rules have effectively removed the bloc from the AI race altogether.
Starmer is hoping that the economic gains from technological advances can be spread widely across the UK’s regions, rather than stay clustered in London as they have been traditionally. There was not a single mention of ‘London’ in the nearly 3,000-word press release the government sent to journalists on its plan, despite the fact the capital is host to the offices of OpenAI, Google DeepMind and Elon Musk’s XAI, among others.
AI is, Starmer is betting, his silver bullet to improve all corners of the UK. And in some respects his optimism is warranted: building an AI business does not come with strong geographic preconditions – the computer power is all accessed remotely in the cloud. And we’re already seeing lots of AI firms popping up around the UK. Our research conducted at UKTN has found that regions outside London are now seeing a faster rate of AI startup creation than London, while the West Midlands alone has built an AI cluster of over 300 specialist AI businesses, employing over 11,000 people.
The AI Action Plan shows that Starmer is a man in a hurry. The PM knows he needs to deliver tangible improvements to the economy well before the end of the decade if he is to cling on to his big majority. And herein lies the biggest problem with his plan.
AI businesses require huge numbers of computers to do their work – lots of big data centres which consume an awful lot of electricity. And data centre companies tell me it takes five to seven years between submitting a planning application and getting the thing built. To add to that, as one government official put it to me, ‘even if you get planning permission, National Grid will say you won’t be hooked up to the network till 2038’.
In other words, Starmer doesn’t have a hope in hell of seriously getting his action plan into action before the next election. It would require a huge overhaul of the planning system, a total transformation of the energy network, a civil service that can move fast and break through bureaucratic barriers (good luck with that) and a host of local authorities ready to wave through huge construction projects in the name of economic growth.
Each of these elements independently poses huge challenges – but doing them all at once? So many ducks have to get in a row to make it work. I hope this government plan turbocharges Britain’s AI sector. But the benefits of technological advances typically take a decade or more to be realised, and there is a reason for that.
Comments