This investment seeks to support this aim through the creation of up to 3 cross-disciplinary hubs. The hubs will bring together researchers from across the mathematical and computational sciences to tackle the foundational problems that exist within AI.

There is a growing need for next-generation AI technologies that have the capabilities to meet the demands of real-world applications, both now and in the future. To realise the vast potential of AI, and for the UK to remain a global leader within the field, we must further develop our understanding of the theoretical foundations of AI and overcome existing methodological barriers.

This major investment will form part of EPSRC’s new strategic delivery plan and will grow investments in AI, digitisation and data along with other priority areas. This underpins the UK Research and Innovation (UKRI) strategic theme ‘building a secure and resilient world’.

This investment seeks to support this through the creation of up to 3 cross-disciplinary hubs that will, through advancing underpinning mathematical and computational concepts, develop novel approaches to methodological challenges in AI.

It is anticipated that the hubs will bring together researchers from across the mathematical and computational sciences to tackle the foundational problems that exist across a range of AI methods, fields, or capabilities.

EPSRC encourages applicants to include representation from different mathematical disciplines, as well as AI researchers, within their core research team. Evidence of co-creation, and leveraging the interface between AI and mathematics, is expected to be evident within the proposed research questions.

EPSRC is not specifying research priorities for the hubs beyond the need for them to tackle the foundational or underlying theoretical problems that exist within AI (such as the ‘how’ and ‘why’ questions of modern AI systems). This is due to the breadth of potential research, and the importance of investing in approaches that can address both current and future needs of AI technologies.

Applicants should think beyond the optimisation of current systems and are asked to propose innovative, and creative research programmes that will advance our fundamental understanding of AI and AI systems. For illustrative purposes only, this may include tackling the challenges that are associated with:

• uncertainty quantification
• integrating causality and inference into AI models
• vulnerabilities (for example, interpretability, verifiability, robustness)
• algorithm development
• algorithmic bias
• fundamentals of optimisation

Aspects of ethics, and responsible research and innovation, should be considered where appropriate.

Proposals that are based on applying current AI methods to an application area will not be accepted.

Applicants should:

• highlight why their proposed hub is nationally important
• outline how their cross-disciplinary approach will enable for novel approaches to current and future methodological challenges in AI to be developed

Eligible parties are invited to submit an Outline proposal, closing date 9 February, 4:00pm UK time.

Full details and applications