Meet the Team Q&A: Tom Coates
I was blown away by how powerful AI tools can be for theorem discovery. It’s impossible to spend a lot of time using these tools, and their more powerful LLM cousins, without becoming intensely curious about how they work and how to reason about them.
In this edition of our Meet the Team Q&A we sat down with Tom Coates. Tom is a Professor of Pure Mathematics at Imperial College London. He is also an Erlangen Hub Co-Investigator and Theme D Deputy Lead focusing on the development of autonomous, self-adaptive AI that uses formal methods to understand its limits, and act safely and ethically.
Can you share a bit about your background and your current research focus?
For the past decade or more, I have been involved in a large-scale collaboration to find and classify algebraic varieties called Fano manifolds — one can think of this as building a “Periodic Table” for shapes. This collaboration involves a lot of large-scale computation, data mining, and machine learning, which is how I got interested in AI as a tool for scientific discovery. More recently I have also been on part-time secondment to the Office of the Chief Scientific Adviser, which has led to my current focus on the interface between AI and policy.
What inspired you to pursue this area?
Originally I was attracted by a purely theoretical, purely mathematical question: how to find the “basic pieces” from which more complex geometries are made. But as we began to deploy machine learning and AI-powered pipelines to tackle this question, I was blown away by how powerful AI tools can be for theorem discovery. And it is not possible to spend a lot of time using these tools, and their more powerful LLM cousins, without becoming intensely curious about how they work and how to reason about them.
Which themes are you connected to within the Erlangen AI Hub and how does your work within the hub intersect with your research background?
I lead on the Government Engagement and Theme D.
What attracted you to the Erlangen AI Hub and what do you hope to see it achieve?
There is such a need for more mathematicians with the skills to work at the interface between AI and policy. The Erlangen Hub is a perfect opportunity to pilot interventions in this space, and to support and grow more early career researchers who are “bilingual” — communicating fluently with both policymakers and scientists in AI and adjacent fields.
What’s been the most surprising or exciting finding in your work so far?
I’ve been enormously proud of the work that Sara Veneziale and I have done developing and delivering training courses on the fundamentals of AI to civil servants. More than 250 civil servants in at least 15 departments to date! This is having a demonstrable impact on policy formulation and delivery, in an area that is critical to both security and economic growth.
What challenges have you faced in your research, and how did you overcome them?
I have been very fortunate to work with an incredible group of collaborators across my career. There have been challenges here and there, but also amazingly creative colleagues to help overcome them.
What advice would you give to someone just starting out in your field?
Get stuck into the practicalities of AI: training large models, fighting the data-cleaning pipeline, wrangling the GPUs. There is no substitute for practical experience — and it helps you to choose and formulate the correct research questions too.