Yann LeCun

If you have heard his name attached to a talk or a quote, it is almost certainly in connection with convolutional networks and self-supervision — the two ideas that his reputation now indexes. He is a scientist and engineer, trained in software and electrical engineering, and his work sits in the part of the field where research ambitions have to survive contact with real systems.

Born in Paris on 8 July 1960, he built a working method out of small, compounding improvements rather than single breakthroughs, and treated scale as something to be designed for from the start rather than hoped for later. Early on he showed a knack for turning technical possibility into something operational. Laboratories and markets both acted as places where the ideas were tested. Mentors, collaborators and opponents supplied the constraints that sharpened his judgement.

In mid-career he led teams that took prototypes through to platforms, making decisions about architecture, culture and where capital went. Failures were written up as experiments with something to teach rather than buried. His public output — memos, interviews, keynotes — worked as a way of teaching the method itself, and regulatory and ethical questions were folded into design rather than handled afterwards. Later, the emphasis shifted towards mentorship and building institutions; the archives of talks and design notes show the rigour that sits underneath what can look like intuition. His site is at yann.lecun.com.