University College London

What it is

University College London, usually called UCL, is a public research university in Bloomsbury, central London. Its place in AI history rests less on one famous project than on a few people and one research unit. That unit is the Gatsby Computational Neuroscience Unit, which studies the brain with computational models, meaning mathematical descriptions that can be run on a computer. Computational neuroscience is the field that uses such models to explain how neurons and brain circuits process information, and it is the bridge between biology and machine learning that matters for UCL's role.

Origins

UCL was founded on 11 February 1826.^[1] Its connection to AI came much later and runs through two people. The first is Geoffrey Hinton, who was the founding director of the Gatsby Unit from 1998 to 2001, before he returned to the University of Toronto.^[2] The second is Demis Hassabis, who earned his doctorate at UCL and did postdoctoral research at the Gatsby Unit before co-founding DeepMind.^[3]

Known for

Hinton is one of the central figures in deep learning, the family of neural network methods that learn from large amounts of data. His spell at UCL was short, and he later returned to Toronto, but it is the best-documented link between the university and the field.^[2]

Demis Hassabis earned a PhD in cognitive neuroscience at the UCL Queen Square Institute of Neurology in 2009. His thesis, “Neural Processes Underpinning Episodic Memory”, was supervised by Eleanor Maguire.^[3] Episodic memory is the memory of specific personal events, such as what happened on a particular day. His first academic paper, “Patients with hippocampal amnesia cannot imagine new experiences”, appeared in PNAS in 2007 and was listed by Science among its Top 10 Breakthroughs of 2007.^[3] The title states the finding directly: people with hippocampal amnesia, a memory disorder that affects the hippocampus, cannot imagine new experiences.

After his PhD, Hassabis held a Henry Wellcome postdoctoral fellowship at the Gatsby Unit, working with Peter Dayan. He met Shane Legg there, when both were postdocs.^[3] In 2010 Hassabis co-founded DeepMind with Legg and Mustafa Suleyman.^[3]

In 2024 Hassabis shared the Nobel Prize in Chemistry with John Jumper of Google DeepMind, for work on protein structure prediction.^[4] Protein structure prediction means working out the three-dimensional shape of a protein from its chain of amino acids. A protein's shape largely determines what it does, so the problem matters both to biology and to drug design.

Place in AI history

UCL's contribution to AI is best read as a meeting point. Neuroscience posed questions about how learning works in the brain, computational models turned those questions into forms that could be tested, and people trained in both moved on to machine learning. Hinton's Gatsby years and Hassabis's postdoctoral work are two examples of that path. Neither was a single breakthrough made at UCL. Each was a step in a career that later went on to work far beyond the university, and the Nobel Prize in Chemistry that Hassabis shared in 2024 shows how far the line from neuroscience to AI has run.