From the Wall Street Journal, November 14:
Yann LeCun invented many fundamental components of modern AI. Now he’s convinced most in his field have been led astray by the siren song of large language models
As a graduate student in the 1980s, Yann LeCun had trouble finding an adviser for his Ph.D. thesis on machine learning—because no one else was studying the topic, he recalled later.
More recently, he’s become the odd man out at Meta. Despite worldwide renown as one of the godfathers of artificial intelligence, he has been increasingly sidelined as the company’s approach diverged from his views on the technology’s future. On Tuesday, news broke that he may soon be leaving Meta to pursue a startup focused on so-called world models, technology that LeCun thinks is more likely to advance the state of AI than Meta’s current language models.
Meta Chief Executive Mark Zuckerberg has been pouring countless billions into the pursuit of what he calls “superintelligence,” hiring an army of top researchers tasked with developing its large language model, Llama, into something that can outperform ChatGPT and Google’s Gemini.
LeCun, by his choice, has taken a different direction. He has been telling anyone who asks that he thinks large language models, or LLMs, are a dead end in the pursuit of computers that can truly outthink humans. He’s fond of comparing the current state-of-the-art models to the mind of a cat—and he believes the cat to be smarter. Several years ago, he stepped back from managing his AI division at Meta, called FAIR, in favor of a role as an individual contributor doing long-term research.
“I’ve been not making friends in various corners of Silicon Valley, including at Meta, saying that within three to five years, this [world models, not LLMs] will be the dominant model for AI architectures, and nobody in their right mind would use LLMs of the type that we have today,” the 65-year-old said last month at a symposium at the Massachusetts Institute of Technology.
LeCun has been talking to associates about creating a startup focused on world models, recruiting colleagues and speaking to investors, The Wall Street Journal previously reported. A world model learns about the world around it by taking in visual information, much like a baby animal or young child does, versus LLMs, which are predictive models based on vast databases of text.
LeCun didn’t respond to requests for comment, and Meta declined to comment.
Early innovations
LeCun was born in Paris, raised in the city’s suburbs and attended what’s now known as the Sorbonne University in France in the 1980s. While getting his Ph.D., he married his wife, Isabelle, and they had the first of their three sons. A woodwind musician, he played traditional Breton music for a Renaissance dance troupe.Always ahead of the curve, LeCun studied machine learning before it was en vogue. He worked in Nobel Prize winner Geoffrey Hinton’s AI lab in Toronto before Hinton became an AI legend, and spent much of his early professional career in New Jersey at Bell Labs, the institute famous for the sheer number of inventions that came out of it.
“The thing that excites me the most is working with people who are smarter than me, because it amplifies your own abilities,” LeCun told Wired magazine in 2023.
At Bell, LeCun helped develop handwriting-recognition technology that became widely used by banks to read checks automatically. He also worked on a project to digitize and distribute paper documents over the internet.
LeCun, who’s said he’s always been interested in physics, mostly worked with physicists at Bell and read a number of physics textbooks.
“I learned a lot by reading things that are not apparently connected with AI or computer science (my undergraduate degree is in electrical engineering, and my formal CS training is pretty small),” he said during a Reddit ask-me-anything session 12 years ago.
In 2003, LeCun started teaching computer science at New York University, and later he became the founding director of NYU’s Center for Data Science. When he’s in New York, he has been known to frequent the city’s jazz clubs.
In 2013, Zuckerberg personally recruited him to head up a new AI division at what was then called Facebook. LeCun oversaw the lab for four years, stepping down in 2018 to become an individual contributor and Facebook’s chief AI scientist.
He won the 2018 A.M. Turing Award, the highest prize in computer science, along with Hinton and Yoshua Bengio. The award honored their foundational work on neural networks, multilayered systems that underlie many powerful AI systems, from OpenAI’s chatbots to self-driving cars.
Since then, LeCun, who speaks with a light French accent and is known for wearing black Ray-Ban glasses and collared shirts, has largely become a figurehead for the company. He wasn’t part of the team that helped create Meta’s first open-source large language model, called Llama, and he hasn’t been involved in the day-to-day operations of their development since.
LeCun works on his own projects and travels to conferences, talking about Meta’s AI glasses and his own views on the path to AI advancement, among other things, people who have worked with him said.
Léon Bottou, a longtime friend of LeCun’s, previously told The Wall Street Journal that he’s “stubborn in a good way,” meaning he is willing to listen to others’ views, but has strong convictions of his own.
He also holds strong opinions on a variety of other topics. “I am everything the religious right despises,” he wrote on his website: “a scientist, an atheist, a leftist (by American standards at least), a university professor, and a Frenchman.”
Breaking away
Most of his recent takes have been knocks on the LLMs at the center of Zuckerberg’s ambitions–and also of nearly every other major tech company’s.“We are not going to get to human-level AI just by scaling LLMs,” he said on Alex Kantrowitz’s Big Technology podcast this spring. “There’s no way, absolutely no way, and whatever you can hear from some of my more adventurous colleagues, it’s not going to happen within the next two years. There’s absolutely no way in hell to–pardon my French.”
This summer, as part of a major restructuring, Zuckerberg named 28-year-old Alexandr Wang as Meta’s new chief AI officer–LeCun’s new boss–and ChatGPT co-creator Shengjia Zhao as Meta’s new chief scientist....
....MUCH MORE
Regarding the LLM question, December 2024:"What is an AI agent? A computer scientist explains the next wave of artificial intelligence tools"
We've been saying it (sometimes literally*) for quite a while, chatbots are not the be-all and end-all of artificial intelligence....
AI: Chatbots Are Sooo 2023; Here Comes Interactive AI
"ChatBots Are Not The Be-All And End-All Of Artificial Intelligence":
Far from it.
And all the focus on ChatBots and LLMs are more than just a distraction, they are a perverse representation of what AI is doing and will do and could potentially cost you money or opportunity or both....ChatBots Are For Children: "What’s Ahead for OpenAI? Project Strawberry, Orion, and GPT Next"
IEEE Spectrum - "What Are AI Agents?"
"First impressions of ChatGPT o1: An AI designed to overthink it"CoinTelegraph has developed an artisanal, homebrew AI specialty. Here's one of our previous visits:
AI Use Case: Biological Immortality By 2030This would be a pretty good answer to the question "What is the use case for AI?"But I don't buy it. AI will be like the nanotech revolution that never was, never that is, in the sense of a nanotech industry. Instead, as with nanotech, AI will be embedded in the processes and protocols of every facet of human existence and we won't even notice it.
"AI agents are the 'next frontier' and will change our working lives forever"
Former Google CEO Schmidt On The Ever-Increasing Tempo Of AI
Also:
Where Is Artificial Intelligence Going From Here: One Of The Gurus Speaks
And on LeCun:
Tech leaders respond to the rapid rise of DeepSeek
....Yann LeCun, the Chief AI Scientist for Meta’s Fundamental AI Research (FAIR) division, posted on his LinkedIn account:
“To people who see the performance of DeepSeek and think:
‘China is surpassing the US in AI.’
You are reading this wrong.
The correct reading is:
‘Open source models are surpassing proprietary ones.’
DeepSeek has profited from open research and open source (e.g. PyTorch and Llama from Meta)
They came up with new ideas and built them on top of other people’s work.
Because their work is published and open source, everyone can profit from it.
That is the power of open research and open source.”....
Previously on the LeCun channel:
November 2024 - Chief AI Scientist at Meta, Yann LeCun: "I don't wanna say "I told you so", but I told you so."
February 2024 - "Meta’s A.I. Chief Yann LeCun Explains Why a House Cat Is Smarter Than The Best A.I."
And many more.
