Skip to main content
Menu

Large ARIA grant will help Oxford engineers revolutionise the architecture of AI systems

A team of academics, led by Professor Noa Zilberman, has received a large grant under the Advanced Research + Invention Agency (ARIA) Scaling Compute programme, which aims to unlock AI hardware at 1/1000th the cost

L to R: Professors Amro Awad, Dominic O’Brien, Noa Zilberman, Nick McKeown, Martin Booth, and Dr Patrick Salter

Professor Noa Zilberman and her colleagues have received a large prestigious grant under the Advanced Research + Invention Agency (ARIA) Scaling Compute programme. On 25 October ARIA announced the award of nearly £50m for 12 projects to unlock AI hardware at 1/1000th the cost, with the goal of breakthrough R&D for the UK and beyond.

Led by Programme Director Suraj Bramhavar, ARIA’s Scaling Compute programme challenges key pillars of our computing paradigm. If successful, this programme will unlock a new technological lever for next-generation AI hardware, alleviate dependence on leading-edge chip manufacturing and open up new avenues to scale AI hardware – an industry which is expected to be worth trillions of pounds.

The Scalable AI systems project will look to drop the cost of AI hardware by >1000x, and provide an entirely new vector for improved computing performance. The vision of the Oxford team is to create AI systems that are not only affordable and accessible, but also sustainable and resilient to failures. 

Current mechanisms for training AI systems utilise a narrow set of algorithms and hardware building blocks, which require significant capital to develop and produce, resulting in a mismatch in supply and demand. Natural processing systems innately process complex information more efficiently – on several orders of magnitude. Through its coordinated R&D effort, ARIA seeks to uncover new technological pathways that can bridge the gap.

In the Scalable AI systems project, the Oxford team are aiming to introduce a new interconnect for scalable AI systems that solve the communication bottleneck. Using an optical interconnect through glass, and fitting hundreds to thousands of accelerators within a single computer, the project aims to revolutionise the design of AI systems.  

The ARIA R&D Creators will be made up of 16 researchers led by Professor Noa Zilberman, Professor Amro Awad, Professor Martin Booth, Professor Nick McKeown, Professor Dominic O’Brien, and Dr Patrick Salter.

Eight new posts are being recruited for including postgraduate students, postdoctoral researchers and a Project Manager. For further information see the project website.

An important part of the project is engagement with industry and end-users, reflected by an industry affiliates program. If you are interested in keeping track of updates and helping steer direction, please get in touch with Professor Noa Zilberman.

This project is funded by the Advanced Research + Invention Agency (ARIA).