Skip to content

Search the site

Keir Starmer wants to “unleash” AI on UK Plc. What does that mean and should we be delighted?

"We've heard similar rhetoric before"

Image credit: https://unsplash.com/@luna_and_boo

Gilt yields are soaring, hiring is slumping, the chancellor is seeking investment from China – an authoritarian nation intent on maliciously attacking critical national infrastructure and stealing intellectual property – and Prime Minister Keir Starmer is unceremoniously hitching UK Plc’s cart to a riproaring AI bandwagon in the hope that it leads to sunlit economic uplands; or at least the dopamine hit of rare positive press.

May you live in interesting times. The Stack’s inbox suggested he’d certainly achieved a sugar rush of goodwill from the majority of the technology sector after agreeing to a new “AI Opportunities Action Plan.” It overflowed with platitudinal* press releases after HMG said on Monday that AI “will be unleashed… to deliver a decade** of national renewal.”

“The UK's AI Opportunity Action Plan is a clear-sighted and ambitious policy initiative to establish the UK as a global AI leader. Delivering these commitments can boost economic prosperity, enhance public services and foster the growth of a thriving startup ecosystem" – NVIDIA in a emailed comment so deeply vacuous that it was probably AI-generated itself.

AI “unleashed”

Mental images of electricity-thirsty, hallucinating, probabilistic algorithms strutting the nation like bully breeds that have slipped the lead and want to mark your kids’ homework (more on that below) aside, what’s new?

Good things: Starmer touted a chunky £14 billion in private sector investment. A whopping 85% of that was attributed to one company, co-location provider Vantage Data Centres; which did not respond to requests for comment from The Stack seeking detail on this investment allocation.

Exclusive: Shell's Chief Data Officer on building the foundations for AI

"I am determined the UK becomes the best place to start and scale an AI business. That will be the centrepiece of our Industrial Strategy" – The PM.

(Vantage in 2023 entered London with its 48MW £500 million “LHR1” campus and swiftly followed that with plans for a £250 million LHR2. On those numbers, the £12 billion Starmer says it will invest should get it another 16 more co-location facilities; Wales anticipates some of them...)

Lights, Cameras, AI Opportunities Action Plan...

The government’s “AI Opportunities Action Plan” also promises fresh investment in a supercomputer with “enough AI power to play itself at chess half a million times a second”; a strong gambit, but with no hard numbers behind it, nor a timescale, nor a location, unless we missed it?

This comes after Labour scrapped the previous government’s promise of £1 billion+ for the UK's first exascale supercomputer, saying it was unfunded. Streamlined planning for these DCs and the delivery of a mysteriously nebulous “National Data Library…” are also on the cards. (More details on the National Data Library are promised “in due course.”)

The plan accepts all 50 recommendations of Matt Clifford – the founder of incubator Entrepreneur First and chair of the government’s Advanced Research and Innovation Agency – who in a recently commissioned report also suggested positively that the “NDL should run open calls to receive proposals from researchers and industry to propose new data sets.) 

There’s much ambition here, in short; but strikingly little detail. 

See also: UK nails the coffin shut of the “New Payments Architecture”. Now, burn it?

Clifford’s report proposed investing to ensure that HMG has the equivalent of 100,000 GPUs in government-owned “sovereign” compute capacity by 2030. The government’s response today does not commit to this specific number but promises investment in improved sovereign compute capacity.

As Tom Sheridan, VP, Investment at tech VC firm RTP Global noted however, this is “equivalent to the compute used to train Meta’s Llama foundational AI model today. We are extremely far behind.”

He added: “The UK initiatives announced today to support the AI sector are welcome, but we need to be clear-eyed about their impact…

“Access for academia and public services to dramatically more compute power will accelerate the speed of AI innovation in our domestic health, energy and transport sectors. Ssupport for AI research without attached commercial interests that will act as a draw to the UK for world-class AI talent who want to use their expertise to solve society’s big problems."

Sheridan added: What this does to place AI at the heart of efforts to grow the UK economy, on the other hand, is questionable. Tech start-ups, even those developing ‘sensitive’ AI applications, seemingly won’t have access to this expansion in sovereign compute. This means they’re going to continue looking elsewhere for compute and U.S. cloud companies will be happy to serve them..."

Dr Marc Warner, CEO and co-founder, AI firm Faculty, said: “We’ve heard similar rhetoric before. Bold policy must now follow today’s encouraging soundbites if the UK is truly to have AI ‘mainlined into its veins’.” 

Exclusive: Shell's Chief Data Officer on building the foundations for AI

Recommendation 13 of Clifford’s report, with which HMG says it “partially” agrees meanwhile also calls for “a copyright cleared British media asset training data set, which can be licenced internationally at scale.” 

Given that everyone and their dog is merrily scraping this data anyway, quite how much value this will have by the time it reaches fruition is an open question. As Jason Raeburn, Partner and Head of Intellectual Property & Technology Litigation, at Paul Hastings, told The Stack: “The published commitment from the UK government in this regard only goes as far as ‘engaging’ with partner organisations to consider the ‘potential role’ of government – it would seem unlikely that a data set of this nature will gain much traction in the near-term.”

Further downstream, meanwhile, there’s the promise/threat of “AI tools to help with marking and generating detailed, tailored feedback for individual students… so teachers can focus on delivering brilliant lessons.”

If late stage communism was a case of “they pretend to pay us; we pretend to work”, education in the late stage capitalism AI era is apparently set to be a case of “we pretend to mark their homework; they pretend to do it.” 

See also: UK gov't reveals “£1 billion” semiconductor strategy, moans that industry ignored it

*What do you mean ‘that’s not a word!’?

**Why a decade? Two parliamentary terms? Two commodity server refresh cycles? A period chosen at random? Answers on a postcard

Latest