Yann LeCun

Yann LeCun's contribution to AI is primarily the convolutional neural network (CNN) — an architecture that, from the late 1980s at Bell Labs, he applied to handwritten digit recognition. His 1998 LeNet demonstrated that CNNs could reliably read handwritten zip codes for the US postal service. At the time this was a practical, deployed system, not a research curiosity. The same architectural principles — local connectivity, weight sharing, pooling — became the foundation of all modern computer vision when the field returned to neural networks after 2012.

LeCun has been Chief AI Scientist at Meta since 2013 and continues to hold a professorship at NYU. He was awarded the 2018 Turing Award alongside Hinton and Bengio.

His public positions diverge sharply from Hinton's on existential risk. LeCun argues that current large language models are fundamentally limited — they lack the causal reasoning, world models, and grounding that biological intelligence relies on — and that the path to human-level AI, if it is achievable at all, requires architectures that do not yet exist. He is therefore sceptical of both claims that AGI is imminent and warnings of near-term existential risk. He has been vocal in criticising what he characterises as "AGI doom" narratives.

Meta's release of the Llama model family as open-weight models is partly a product of LeCun's influence: he has publicly argued that open AI development is safer than closed development because it enables scrutiny and reduces concentration of power.