Yann LeCun

One of the founders of deep learning, best known for convolutional networks. He shared the 2018 Turing Award with Geoffrey Hinton and Yoshua Bengio, but unlike them he is the most prominent expert skeptic of AI risk. He was Meta’s chief AI scientist for over a decade, then left in 2025 to found AMI Labs, which works on AI that learns from the physical world.

His two main claims:

• Language models are not the route to human-level AI. They lack a model of the physical world, persistent memory, reasoning and planning. Getting there needs new architectures.
• The risk is overblown. Capability will arrive gradually, and safety will be engineered along the way, as with aircraft. Many serious people say AGI is dangerous quotes him calling a sudden takeover “the preposterous scenario”. See Can we learn from AI failures before they are fatal.

Timelines

• 2024-10-16, Hudson Forum: human-level AI is “going to take years before we can get everything here to work, if not a decade.”
• 2026: after founding AMI Labs, he repeated that language models are not a route to human-level intelligence.

See AGI timelines.