A coalition of leading British universities has established the Science of Fundamental AI Research (SOFAIR) Lab, a major new initiative aimed at developing next-generation open-source AI technologies capable of running on widely accessible hardware, independent of the data centre infrastructure currently controlled by a small number of large technology companies.
The lab brings together researchers from Oxford, Cambridge, Edinburgh, and University College London, including Oxford professors Yarin Gal, Mark van der Wilk, Michael Bronstein, and Michael Wooldridge. David Barber, appointed as SOFAIR Lab Director, described the initiative as the missing piece in the UK AI ecosystem, uniting the country’s leading groups across natural language processing, probabilistic inference, agentic AI, and neuroscience.
Rather than building on existing foundation models, SOFAIR is pursuing fundamental research into the architectures, training methods, and distributed systems that underpin AI itself. The lab is also developing an in-house open-source multimodal frontier foundation model as a testbed, with an interdisciplinary structure spanning computer science, mathematics, statistics, and neuroscience, drawing in part on insights from how the human brain integrates different types of reasoning.
The initiative has received recognition from UK AI Minister Kanishka Narayan, who highlighted its potential to secure Britain’s position as a global AI leader by ensuring transformative research advances domestically. Fully funded four-year PhD studentships are now open, based at the UCL AI Centre in London with extended visits to partner institutions.
