Yann LeCun

Some people are right too early, and that almost always comes with a cost. Yann LeCun, a French engineer born in 1960, spent much of the 1980s and 90s convinced that a machine could "see" if it was trained on enough examples, layer by layer. The artificial intelligence field wasn't exactly on board. Neural networks had gone out of fashion, funding was flowing elsewhere, and skepticism was the safe position. LeCun kept going anyway.

In 1988 he arrived at Bell Laboratories in New Jersey with an idea: a network of artificial neurons organized into layers that could detect local patterns in images, much like the visual brain processes edges before recognizing objects. He called it a convolutional neural network, or CNN. By 1989 he already had a working system, LeNet-1, that recognized handwritten digits with 95% accuracy using zip codes from the US Postal Service. By 1998, with LeNet-5 and collaborators including Yoshua Bengio, the network was reading millions of bank checks per day for companies like NCR. This wasn't a lab paper: it was real financial infrastructure.

In 2013 he became the first director of Meta AI Research and then Chief AI Scientist at Meta, a role he holds from New York alongside his professorship in Computer Science at New York University (NYU). In 2018 he received the Turing Award, the "Nobel of computing," alongside Geoffrey Hinton and Yoshua Bengio, recognizing decades of work that now drives trillions of dollars in industry. Today he is also one of the loudest voices in the debate about AI's future: his position is that the path to general intelligence runs through self-supervised learning and reasoning, not through scaling language models indefinitely.

What fascinates me about LeCun isn't only the technical brilliance; it's the persistence without guarantees. He worked on something the world dismissed, year after year, and by the time the world changed its mind, the banking system you used to cash your check had already been running on his ideas for years. Today, every time a photo app recognizes your face, every time a car detects a stop sign, there is a CNN there, quiet and invisible, doing what LeCun proved was possible when almost no one believed him.