Review of Digital Research Infrastructure Requirements for AI

The National AI Strategy set out a ten-year vision to make the UK a global AI superpower and acknowledged access to people, data and compute as key drivers of progress and strategic advantage in AI. Digital Research Infrastructure (DRI) plays an integral role in the wider compute for AI ecosystem. Ensuring the DRI ecosystem meets the current and future needs of the research and innovation community developing and using AI will be essential to meeting the ambitions set out in the National AI Strategy.

On behalf of UKRI, The Alan Turing Institute in conjunction with Technopolis has conducted a review to better understand the UK’s current and future DRI needs for AI. This exercise focused on consulting with AI communities, AI researchers and researchers who use AI to solve problems, plus wider stakeholders to understand their needs across three main elements (compute, data access and people/ skills), currently and in five to ten years’ time. This summary report sets out the views of this collective community, as communicated during the review.

Key findings

• The UK needs to scale-up and then continuously invest in DRI for AI if it seeks to become a global AI superpower
• Compute capacity for AI needs to be increased while ensuring it is easily accessible, configurable, adjustable, and promote collaboration to enable major scientific advances
• Any investment in hardware/compute for AI needs to be matched by investment in training and support to maximise uptake, efficiency and generated scientific outputs
• Unified data management standards and sharing policies are needed

5-10 year outlook

If implemented in full, the recommendations put forward by the community, as identified in this review, could amount to an integrated and holistic programme of support for compute capacity, data access, and people and skills. This would likely have an important impact on the UK’s ambitions to be world-leading in AI research and innovation over the next 5 to 10 years.

The key benefits envisaged include more straightforward and equitable access to significantly enhanced compute capability for AI research and innovation, supporting a wider diversity of research communities, organisations, and geographic locations. The enhanced AI capability would incorporate cloud native technology where appropriate, and be complemented by a breadth of high-quality AI-ready open and public data sources. Improved arrangements would also be put
into place for access to public sector data, restricted data and commercially licensed data.

In parallel, adoption of AI tools and techniques would be supported across research disciplines and in industrial R&D by developing and nurturing a highly skilled cadre of Research Technology Professionals and upskilling the wider research community. This would enable AI researchers to exploit DRI for AI to its fullest potential, through continued professional development, training opportunities and embedded support at an institutional level.