Profile

Yann LeCun, born near Paris in 1960, is a French-American computer scientist. He trained as an engineer at ESIEE in Paris and earned a PhD in computer science at Pierre and Marie Curie University. He then did a postdoc with Geoffrey Hinton at the University of Toronto.

In 1988 he joined AT&T Bell Labs, where he did his best-known work. He has been a professor at New York University since 2003. In 2013 he was invited to found Facebook AI Research (FAIR) and lead it, and he later served for many years as Meta’s chief AI scientist. He left Meta in November 2025 and went on to co-found AMI Labs, a company focused on world models.

Key contributions

LeCun’s central contribution is the convolutional neural network (CNN). He used backpropagation to train networks with convolutional structure, letting models learn features directly from pixels instead of relying on hand-designed ones. That idea became the foundation for image recognition, object detection and almost every other vision task.

At Bell Labs he and colleagues also worked on the DjVu image compression technology and assembled and popularised the MNIST handwritten digit dataset, for years one of the most widely used starter benchmarks in machine learning.

More recently he has focused on self-supervised learning and proposed ideas such as the Joint Embedding Predictive Architecture (JEPA), arguing that AI should learn to predict abstract representations of the world through observation rather than just predicting the next word.

Notable work

The LeNet family of networks he developed in the 1990s recognised handwritten digits and was deployed in systems that read bank cheques automatically, one of the earliest large-scale real-world uses of convolutional networks. A 1998 paper with collaborators on gradient-based learning for document recognition summarised the approach and is still heavily cited.

At Meta, the FAIR lab under his leadership released a large body of open research and tools, and he has been a vocal supporter of open models.

Honours

  • 2018 Turing Award, shared with Geoffrey Hinton and Yoshua Bengio.
  • 2022 Princess of Asturias Award for Technical and Scientific Research.
  • 2025 Queen Elizabeth Prize for Engineering.

Influence

LeCun, Hinton and Bengio are often called the “godfathers of deep learning.” Convolutional networks let computers interpret images with practical accuracy for the first time, and their influence is visible in face recognition, medical image analysis and the perception systems of self-driving cars.

He is also one of the most active public voices in AI. He maintains that today’s large language models are not enough to reach human-level intelligence and that new architectures able to understand the physical world and plan are needed. He is openly sceptical of claims that AI poses an existential threat and is a strong advocate of open source. His positions often spark debate, and they keep the field thinking about an alternative path beyond simply scaling up.