Research Outline

What I cannot create, I do not understand’ (Feynman, 1988). We develop machines that learn and think like people, leveraging our most fundamental models of cognition and behaviour. We work on artificial intelligence from first principles, developing fundamental descriptions of intelligence in natural systems and advancing toward scalable, aligned artificial agents. We believe this research program will help us unlock the next generation of AI systems, which are robust and adaptive, safe and trustworthy, data-efficient and environmentally friendly, with transformative advantages for society and scientific discovery.