Sir Keir Starmer has launched a plan to make use of artificial intelligence across the private and non-private sectors that he hopes will cut costs and boost Britain's economic growth.
The AI Opportunities Action Plan, written by enterprise capitalist Matt Clifford, comprises 50 recommendations. They were all approved by Starmer to make sure the UK is “considered one of the most important AI superpowers” and an “AI maker” reasonably than an “AI taker”.
But experts warn that major obstacles lie ahead, from access to energy and computing power to deep-rooted public concerns about learn how to govern rapidly evolving technology and the wasteful use of personal data.
What plans does the federal government have with health data?
Starmer has outlined plans to provide researchers and AI corporations access to public datasets, including anonymised NHS patient data, in a bid to make Britain a horny location for business and improve productivity within the struggling health service.
Ministers intend to incorporate NHS data in a brand new “national data library” and five public datasets are expected to form the primary tranche when the project is up and running.
The Prime Minister didn’t confirm on Monday whether these datasets can be sold for business purposes and whether work is underway to cost the info.
However, he said that while it was vital for the federal government to take care of “control” of health records, “I don't think we must always take a defensive stance that forestalls the form of breakthroughs we want.”
Nell Thornton, a fellow on the Health Foundation, said the think tank's survey showed that 75 per cent of the general public can be blissful to share not less than a few of their data to develop AI systems throughout the NHS.
Still, health experts and privacy activists are skeptical that the federal government will make sure the needed safeguards are in place to guard private health information and the service itself.
Saif Abed, a former NHS doctor and cyber security and public health expert, said he “fears that the wanton use of (AI) without strong safeguards will leave the NHS vulnerable”.
“Who will perform the governance, audit and oversight function for all these suppliers on behalf of the NHS? “How will suppliers be punished for bad behavior?” he added.
What's planned for supercomputers?
The government confirmed that an £800 million “exascale” supercomputer wouldn’t be built on the University of Edinburgh. Instead, it is going to spend money on a brand new supercomputer that focuses more specifically on AI.
It is unclear how advanced Starmer's recent project can be, how much it is going to cost and when it is going to be launched, but he has committed to increasing Britain's government computing power by 20-fold.
Analysts argue that “sovereign” computing capability – that’s, data processed and stored within the UK – is crucial to making sure that UK corporations and researchers have access to computing power at a time when access to computing power is becoming increasingly difficult reliable and fast computing power at reasonable costs have a geopolitical battlefield.
Exascale supercomputers – defined as the power to perform a billion operations per second – are being developed by many major economies all over the world, including the United States and China.
AI data center company Nscale announced plans on Monday to speculate £2 billion within the UK over the following three years to construct sovereign AI computing facilities to scale back British corporations' dependence on US cloud providers. Karl Havard, Nscale's chief operating officer, said the federal government's renewed deal with AI was a “catalyst to make this move much faster.”
Clifford's proposals call for a rise in computing power that may very well be made available to academics and AI researchers – the equivalent of 100,000 GPUs in government capability by 2030. But the goal falls wanting the capability already available to Big Tech corporations .
Elon Musk's xAI recently opened a knowledge center in Memphis to coach the following generation of its AI model, which has 100,000 GPUs, and plans to extend that number to greater than 1 million specialized AI chips in the approaching years.
Do now we have enough energy to power our recent AI data centers?
Data centers require electricity to operate and stay cool. How much is determined by their size, with requirements for average locations being relatively modest.
However, in accordance with the International Energy Agency, a large “hyperscale” center can use as much electricity annually as 350,000 to 400,000 electric cars.
In a November report, the UK's National Energy System Operator, the agency answerable for operating and planning the UK's electricity system, predicted that demand for electricity from data centers within the UK would quadruple by 2030.
However, the UK electricity system is under pressure because of long waits for connection to the grid and high electricity prices.
“There are countries which have taken the lead (on data centers) and we aren’t considered one of them,” said Kate Mulvany, principal consultant at Cornwall Insight, an energy consulting firm. “If we wish to deliver excellence, energy availability and value matter,” she added.
The government said today it could create special “AI growth zones” where projects would have “higher access to the energy grid.” The first will happen in Culham, Oxfordshire.
Luke Alvarez, managing partner at Hiro Capital, a London-based technology investor, said the UK is “quickly needing lots more energy” to deliver on its AI guarantees.
How long will it take for AI to spice up economic growth?
The prime minister is under intense pressure to persuade markets that his policies will boost economic growth within the medium term as he tries to maintain his government's budget plans heading in the right direction. A sell-off in bonds sent Britain's borrowing costs to a 16-year high on Monday.
Starmer said: “I don't think it is going to take five or 10 years for productivity to double, not for a moment,” adding he was “absolutely confident that the timeframes we're talking about are lots, are much shorter”.
But a lot of the data centers that store and process the info that goes into AI algorithms take years to construct, and the federal government's plans for a national data library are still of their infancy.
