Cambridge-1 and the future of medicine
Medical researchers finally get the compute power they need. Nicole Kobie reveals how Nvidia’s first public supercomputer could change healthcare in the UK and the world
Medical researchers finally get the compute power they need. Nicole Kobie reveals how Nvidia’s Cambridge-1 could change healthcare in the UK and the world.
The future of AI in medicine is in Cambridge – and it’s superfast. Nvidia has unveiled Cambridge-1, a practical name for the most powerful supercomputer in the UK and among the top 50 globally, installed during the pandemic at a cost of $100 million (£73 million) to the GPU giant.
The aim isn’t just to cozy up to the UK government at a time when Nvidia is hoping its acquisition of British chipmaker ARM will be approved, but to show off its DGX SuperPOD scalable computing setup for AI – and help power the future of medicine. Cambridge-1’s first projects will be with pharmaceutical giants GSK and AstraZeneca, medical startup Oxford Nanopore, and researchers at Guy’s and St Thomas’ NHS Foundation Trust and King’s College London.
“Cambridge-1 will empower world-leading researchers in business and academia with the ability to perform their life’s work on the
UK’s most powerful supercomputer, unlocking clues to disease and treatments at a scale and speed previously impossible in the UK,” said Jensen Huang, founder and CEO of Nvidia, at the launch of the new supercomputer. “The discoveries developed on Cambridge-1 will take shape in the UK, but the impact will be global, driving groundbreaking research that has the potential to benefit millions around the world.
“To do this work, scientists need a powerful rocket for their journey,” said Huang. Cambridge-1 was named in honour of the city where genomics and computing got its start, but it’s also where Nvidia hopes to base its UK HQ — if the controversial ARM deal is approved by regulators. “Cambridge-1, and our purchase of ARM, is our giant investment in the UK, giving us a platform to work with the amazing universities, companies, and over 1,000 AI startups in this community,” said Huang.
Scalable supercomputing
At eight petaflops of Linpack performance, the industry standard for measuring supercomputers, Cambridge-1 is now the UK’s fastest supercomputer, just pipping the current leader run by the Met Office.
So how has Nvidia, famous for its gaming GPUs, achieved this feat? The answer lies two decades in the past, when researchers struggling with machine learning shifted from standard CPUs to Nvidia’s hardware for their AI projects, sparking a boom in AI that continues today. Nvidia has
reacted by developing hardware specific to AI tasks, and that includes the Ampere processor at the core of Cambridge-1.
The Cambridge-1 is built around Nvidia’s SuperPOD system, which in turn is designed around Nvidia’s AI-system-in-a-box, the DGX. That features eight A100 Ampere GPUs as AI accelerators, alongside a 128-core AMD CPU, and comes complete with built-in networking and storage as well as AI software and models. The idea is that an organisation can scale up its AI efforts by plugging together multiple DGX modules, while a SuperPOD stacks 80 of those together in one package, for a supercomputer in a box.
All of that is impressive, but what’s intriguing about Cambridge-1 is that it’s Nvidia’s first supercomputer that people outside the company can have a go on. The chipmaker builds supercomputers for its own uses, in particular developing its tech, but the Cambridge-1 is the first created just for third-party partners — and medical ones at that. “Cambridge-1 is our first big bet on the digital biology revolution,” said Huang.
The Cambridge-1 isn’t actually in Cambridge, however. In order for the supercomputer to be powered by renewable energy, it’s sited at green hosting provider Kao Data’s location in Harlow, Essex.
Future of medicine
AI is increasingly applied to medicine in myriad ways, from scanning images to spot symptoms of cancer to sifting through data to spot patterns for research on diseases such as dementia. It also develops models to help create new drugs. While we’ve been hearing about healthcare AI for years, it remains in the early stages, as ethics and safety concerns mean development must be done carefully.
Right now, there are “two core problems”, said Sebastien Ourselin, head of the School of Biomedical Engineering and Imaging Sciences at King’s College London: access to data and the computing power to process growing models.
On the first problem, industry and the NHS need to work closely together to increase accessibility to relevant data, an area that’s been fraught in the UK because of privacy concerns.
Ourselin added that it’s key to ensure that AI is developed using unbiased datasets that actually represent the demographics of the local population. Otherwise, such systems won’t work as well for everyone.
And there’s more to data than access, notes Rosemary Sinclair Dokos, vice president of product and programme management at Oxford Nanopore: the data needs to be processed to be used too. “There’s one challenge, which is generating, cataloguing and curating data, but you’re also performing discovery on that data,” she said. “So you have the challenge of having to go backwards and forwards and checking and testing your assumptions all the time – and having a limit to your compute doesn’t help with those challenges.”
On the second point, AI is getting more complex to help drive accuracy. The more complicated a machine learning model is, the more data points that can be pulled in and manipulated – ideally improving accuracy. But that also requires more data, and the combination of more complicated models and massive datasets means training a model and running it takes more computing power than simpler systems. “If you scale those things up, the data and the size and complexity of your model, then the power of your computer limits you,” said Lindsay Edwards, vice president of data science and AI, respiratory and immunology, biopharmaceuticals R&D at AstraZeneca. “If it takes you three weeks to train a model, your ability to hone that model is really difficult – you need to be able to iterate quickly.”
“Some of the first projects are using AI to sequence genomics, discover novel drugs, and to unlock the mysteries of dementia”
Hence the excitement from medical organisations to have a go on Cambridge-1. Access to this supercomputer could spark a leap forward in more detailed, complex models, in turn finally allowing faster breakthroughs in drugs, genomics and understanding of diseases – though first researchers need to learn how to use the machine and develop their AI to work on the Nvidia system.
The first projects to access Cambridge-1 include drug discovery and medical imaging with AstraZeneca, genetics and health predictions with GSK and genomic sequencing with Oxford Nanopore Technologies. It will also generate synthetic brain images for King’s College London and Guy’s and St Thomas’ NHS Foundation Trust to power research into dementia, strokes, cancer and multiple sclerosis. “Some of the first projects slated for the supercomputer are using AI to sequence genomics, discover novel drugs, and to unlock the mysteries of dementia by studying MRI scans,” said Huang.
AstraZeneca and drug discovery
AstraZeneca is now well-known thanks to its production of the Oxford-developed Covid vaccine, which was being jabbed into arms less than a year after the pandemic hit – and speedy drug development is what AstraZeneca is hoping to create on Cambridge-1. The pharmaceutical giant is using a new transformer-based neural network architecture. A neural network is a type of