Inside the world's largest AI computer chip unveiled at prestigious Scottish university

NHS and other public bodies could benefit from ‘very exciting technology’

The largest artificial intelligence (AI) chip ever built in the world has been used in a new system unveiled by Scottish computing experts.

The development at Edinburgh University’s supercomputing centre will keep the nation at the “leading edge” of the technology, while bringing a “uniquely powerful AI resource” to the region, potentially driving efficiencies in public sector bodies such as the NHS.

Having previously used the first and second generation systems built by Californian company Cerebras, the new model at the Edinburgh centre features the latest technology developed by the firm - the world’s largest AI chip.

Cerebras CS-3 system cluster installed at EPCC | EPCC and University of Edinburgh

Professor Mark Parsons is director of the supercomputing centre, which is the biggest of its kind in the UK, and is known as the EPCC (Edinburgh Parallel Computing Centre).

“Since 2018 we have been beneficiaries of the Edinburgh City Region Deal and what we do as part of that is run a big computing infrastructure to support the use of data science and AI by companies, by the public sector, the NHS and universities in the region, basically to understand the value of data analytics, AI, in their day-to-day businesses,” he told The Scotsman.

“As part of that we have gradually been building quite a large AI capability, and this is just the next step along that road. It’s significant because this is unusual technology, it’s very exciting technology, and it’s also the largest deployment to date in Europe of this technology.

“It is the biggest AI chip in the world. It is 21cm by 21cm - so the best part of a page of A4. And each of these chips has got 900,000 little processors on it.

“If you think your laptop has got four processors, or cores as we call them, and this has got 900,000 cores. So across the four systems we have bought that are connected together, we’ve got 3.6 million cores.

“The systems have been available for about six months, these big ones, and we’re the first and largest deployment in Europe of them.”

Leading political figures have highlighted the potential for using AI to drive efficiencies in the public sector in recent months.

Prof Parsons said the challenge was not with the technology, but making sure people could train AI models and understand how to use them. He used the example of a major hospital in the US that is already using the technology to help administrators write letters to patients, cutting bureaucracy and freeing up staff time.

Such advances are emerging amid ongoing concern among some at the potential long-term impact of AI, however.

“I think people are right to be wary, actually,” said Prof Parsons.

“Some of the applications of AI that aren’t these large language models are really very well understood. So when you are training AI to look at medical images, for example.

“Some of the things that we see. When you or I are having a conversation with Chat GPT or whatever, where it imagines things ... well you certainly don’t want it doing that in terms of working with health data.

“So the wariness, absolutely correct. But not all AI is a large language model. There are many applications of AI which are more mundane that I think are very, very powerful. I don’t think we’ve really got to the benefits of using those types of systems yet at all.”