👁️

Yann LeCun

Original nameYann LeCun

French-American Computer Scientist, Inventor of the Convolutional Neural Network

Those Who Conceived That "a Machine Could Think" · Pioneers

The Convolutional Neural Network (CNN)The "LeNet" Handwritten-Digit Recognition SystemThe 2018 Turing Award

Who they are

Yann LeCun (1960– ) is a French-American computer scientist who, while working at Bell Labs, developed the "LeNet" convolutional neural network and applied it to recognizing handwritten check digits, one of the earliest actual commercial cases of deep learning. The convolutional neural network drew on the hierarchical processing of the biological visual cortex, greatly reducing, through local receptive fields and weight sharing, the parameters and training difficulty of image tasks, and thereafter became the mainstream architectural basis of computer vision for nearly thirty years, from face recognition to autonomous-driving perception systems all built on this idea. LeCun thereafter long served as a professor at New York University, in 2013 becoming chief scientist of Meta’s (formerly Facebook’s) AI research institute, and is also one of the most active public debaters in deep learning, long waging public debates on social media with critics over the capability limits of large language models; he holds always that the current text-prediction-centered large-language-model path has a fundamental limit and cannot lead to true general AI, and advocates that a "world model" is a more promising research direction. He won the Turing Award in 2018 with Hinton and Bengio.

Primary sourcesLeCun et al., "Gradient-based learning applied to document recognition" (1998)The 2018 Turing Award citation

Key stories

Teaching a Machine to "See," Starting From Recognizing Check Digits

The LeNet system LeCun developed at Bell Labs had as its first real application the automatic recognition of handwritten dollar amounts on bank checks, a rather plain commercial need that let the convolutional neural network for the first time prove its practical value in a real industrial scene. This architecture designed on the hierarchical processing of the biological visual cortex was thereafter proven highly universal, its original design idea extended from face recognition to medical-image diagnosis to the road perception of autonomous cars.

Relationships

Disciple

🦙Mark Zuckerberg　LeCun has been chief scientist of Meta’s AI research institute since 2013, the core science adviser Zuckerberg long relied on for the AI technical path

Ally

🧠Geoffrey Hinton 🧭Yoshua Bengio

Nemesis

🤨Gary Marcus　The two have long waged public debates on social media over the capability limits and limitations of large language models

Echoes today

Below are how modern works borrow or reinterpret this name or story — not the original material. The two differ, so keep them apart.

The public debate on social mediaThe public debate between LeCun and critics such as Marcus over the capability boundary of large language models is long one of the most watched public discussion topics in AI on tech media and social platforms.

Appears in

The Deep-Learning Revival and the Shock of ImageNetPersistence Through the "Neural-Network Winter" and the Shock of AlexNet

← Back to Figures