Yann LeCun is a French-American computer scientist best known for developing the Convolutional neural network (CNN), an architecture that became the standard tool for Computer vision and a foundational building block of modern Deep learning.
Career
Born in 1960 near Paris, LeCun earned a PhD from Université Pierre et Marie Curie in 1987, working on early versions of Backpropagation applied to neural networks. He then joined Bell Labs in the United States, where in the late 1980s and 1990s he built LeNet, a CNN trained to recognize handwritten digits that was deployed commercially to read checks, one of the first large-scale industrial applications of neural networks. The architecture's use of convolutional filters and pooling, inspired partly by earlier biologically motivated models, let networks learn visual features directly from pixels rather than relying on hand-engineered ones.
After periods at NEC Research and as a professor at New York University, LeCun became the founding director of Facebook's AI Research lab (FAIR) in 2013, later renamed Meta AI, and served as the company's chief AI scientist. Under his direction Meta pursued an Open-weights models strategy for its models, including the Llama family, in contrast to the closed approach of labs like OpenAI, a choice LeCun has defended as better for research, safety scrutiny and competition.
Views on scaling and world models
LeCun shared the 2018 Turing Award with Geoffrey Hinton and Yoshua Bengio for foundational deep learning work, but has become one of the field's most prominent skeptics of the idea that simply scaling Large language model systems will lead to human-level or general intelligence. He argues that autoregressive text prediction is a poor substrate for reasoning and planning, and has instead promoted research into "world models," self-supervised systems that learn predictive internal representations of physical reality, an approach closely tied to the broader World model concept. He has also been publicly dismissive of near-term Existential risk from AI concerns, putting him at odds with Hinton and Bengio on that question despite their shared technical lineage.
In late 2025, LeCun announced plans to leave Meta to pursue this world-model research independently, reflecting a broader divergence between the "scale is enough" camp associated with frontier large language model labs and researchers who believe new architectures are needed to reach Artificial general intelligence.