It wasn’t easy getting to Culham — there is no direct service from London. Instead, a train from Paddington takes you to Didcot, where you can catch an hourly, two-car diesel service to the village.
As the train gently chugged into Culham station, belching fumes, I looked around to see a ticket office, a row of dilapidated houses and a small cafe. The rest was fields. The ticket office was shut — along with the adjacent red brick platform, it had been sold off and was now used as a print works, with a new concrete platform hastily erected alongside it — and the cafe didn’t open till 2pm. It was deserted.
It did not feel like the most auspicious surroundings for a place that had just become the centrepiece of the government’s new AI Opportunities Action Plan, a fifty-point vision to make the UK one of the leading global centres for the advance of artificial intelligence. The Plan proposed the creation of “AI Growth Zones”, clusters of AI expertise in which the normal planning rules would be suspended to build things quicker and cheaper. The only growth zone identified in the plan was Culham.
My doubts aside, I ventured off in my search for the white heat of Britain’s next technological revolution. My phone map took me to a small business park near the station, surrounded by a wire fence with an entrance guarded by a security hut. I wondered if I’d be let in, or if the tech being built here was of such importance, it would be kept well away from the prying eyes of journalists. I needn’t have worried.
As I approached the security hut, I noticed its windows had been smashed in. No one was inside, but I did find a microwave, an old PC and a traffic cone. With some trepidation, I ventured on.
My experience went from bad to worse once I was inside. Rows of decaying pre-fabs and aircraft hangars were dotted about the place. The site had once been home to HMS Hornbill, a Royal Navy airbase used to train reservists in the 1940s. Many of the buildings had not seen a fresh lick of paint since World War Two.
Random rusting vehicles interspersed the prefabs. One of the buildings, numbered 163A, was flanked on one side by a battered old London bus (destination: Lewisham) and on the other by a mouldy speedboat, with one of the windows missing.
I stopped there. Were the government’s hopes for frontier AI development really pinned on this place? Not exactly, as it would turn out. Though the site is a crucial piece in the puzzle of why it chose Culham.
In July, a planning application was submitted to turn the business park, inventively named ‘Culham No 1 Site’, into a huge new commercial centre, set to feature 115,000 sq m of office space, a 100-room hotel, rows of shops, restaurants, a gym and a nursery.
The scheme itself is part of a much larger plan to turn the tiny village of Culham into a huge new complex known as Culham Science Village, which will see the construction of at least 3,500 new homes.
Though little more than a proposal at this stage, if it goes ahead the notion of Culham as a major AI hub becomes a lot more plausible. But the location of Britain’s new AI Growth Zone was in fact the other side of the fence, next door at a site known as the Culham Science Centre. It was there that I headed next.
Mercifully the security hut here didn’t have smashed windows, and two friendly receptionists sat behind a desk to greet me. From there I was escorted to a newly-opened building to meet Tim Bestwick, chief technology officer of the UK Atomic Energy Authority, or UKAEA.
The UKAEA was formed in the early 1950s as a means of continuing the atomic research that the UK had previously conducted jointly with the US during what was called the Manhattan Project.
“There was a real race on to develop our own domestic capability because it was seen as important,” says Bestwick.
“Simultaneously and tied up with that, the new understanding of atomic processes led to huge enthusiasm for nuclear power in the form of fusion power.
“So there were big incentives and I think at the end of the second world war, the zeitgeist was a country that really wanted to find new purpose and do new meaningful things and put the country back on its feet again.”
The culmination of that enthusiasm was the Joint European Torus JET, a magnetically confined plasma physics experiment to explore the possibility of using nuclear fusion to generate power (nuclear fission, a different process, is what existing nuclear power stations used).
The project started construction in 1979, it became experimentally operational in July 1983 and it was expected to run into the 1990s. But then plans changed.
“I have a memo somewhere that says it will run into the 90s but by any event JET will close by 1999, we turned it off last year after a very productive 40-year campaign that clearly lasted much longer that anyone thought it was going to, but it’s now in decommissioning,” Bestwick says.
Though the decommissioning work for JET means Culham’s primary work on fusion is now in the past, the event has opened up opportunities. For a start, taking apart a nuclear fusion reactor is so complex that there is much to be learned from the process — and novel techniques of handling the components are being developed in the new RACE (Remote Applications in Challenging Environments) building onsite.
The second is the huge power consumption of JET — it has its own 400MV connection to the grid — is now largely not in use. And it was this free capacity, combined with the high power consumption of data centres, that made Culham the government’s top pick for the creation of a new AI growth zone.
“I genuinely don’t know what being an AI growth zone means for planning and whether that’s an issue or not, but for planning on this site we’ve always had a constructive relationship [with the planning authority] anyway, says Bestwick.
“Regardless of being an AI growth zone we have already identified long ago that having access to major compute capability and the latest techniques in machines is important to our future. In fact the compute requirements and simulation requirements of fusion are quite daunting, it’s a very sophisticated challenge, and we already knew we needed it, and we have a team in our advanced computing who work on the very big machines and do the very big simulations anyway.
“But the machines aren’t here –we have a very modest compute capability of our own — so the idea of having one that we have enhanced access to is very exciting.”
All of the details have to be worked out, Bestwick says, but they have already begun work on identifying commercial partners for the construction of data centres.
But it’s clear that, with the right coming together of planning, investment and government support, Culham could prove one of the most vibrant hubs for AI in the coming years.
The post Welcome to Culham: Inside the weird village home to the UK’s first-ever AI growth zone appeared first on UKTN.
