People are confused about Yann LeCun.

Every time he says that LLMs are doomed to fail, people lose their minds. Look at any of his posts on this topic and you'll find hundreds of comments accusing him of arrogance, of insulting everyone building AI products, and even undermining Meta's AI efforts, the very company that has employed him for the last 10 years.

I think most of those comments are misguided. They stem from a fundamental misunderstanding of who Yann LeCun is, what his career has been about, and what his role at Meta actually was and still is.

He's a researcher, not a product person

First and foremost, let me say the important bit: Yann LeCun is not an executive manager at Meta. He's not a product person, not a marketer, not an engineer, not an entrepreneur, not a startup founder, not an investor. He's not a Satya Nadella, an Eric Schmidt, a Peter Thiel, or a Sam Altman.

Yann LeCun has always defined himself as one thing only: a researcher. That's who he is. He's an academic researcher trying to push the science forward. This is what he truly cares about because this is his job. It's always been his job, and it will continue to be his job.

His research earned him the Turing Award, the Computer Science equivalent of a Nobel Prize. He's one of the biggest names in the entire field today. He's a big deal, the same way any Nobel Prize winner is a big deal. Along with other legendary figures like Hinton, Bengio, and many others, he laid the scientific foundation on which current LLMs are standing.

His role at Meta was never about products

When Mark Zuckerberg hired him in 2013 to create FAIR (Facebook AI Research, note the "Research"), it wasn't to manage product development. He was hired as a legendary researcher to continue doing what he had always been doing both in academia and in corporate environments like Bell Labs: pushing the science forward.

This is why it might surprise outsiders that LeCun had no involvement in creating the Llama models. His team's job at Meta was never to release products. Their job was to publish peer-reviewed papers that would be noticed by academic researchers around the world, not by consumers. That's what FAIR has always been about: releasing good scientific papers, not consumer-facing products.

The whole concept of FAIR was that this team would do fundamental research, completely detached from Facebook/Meta's business needs. If Meta wanted to use FAIR's science to build products, fine, but that wasn't FAIR's responsibility. That was the deal between LeCun and Zuckerberg in 2013.

What he means when he says LLMs are doomed

Now let me come back to his comments on LLMs. When he says things like "LLMs are doomed to fail" or "if you're young, don't work on LLMs," people confuse his statements for something they're not. He's not throwing these comments in the face of engineers, executives, and investors currently building consumer products. Many people mistakenly think he's despising their work. But no, those people aren't his target. He's simply talking to other researchers about the next decade of ML breakthroughs.

His goal as a scientist is not to shape or participate in the current engineering and business dynamics around LLMs. That's not his concern at all. I truly think he couldn't care less whether OpenAI, Anthropic, or Meta uses LLMs for their current products. That's not what he does. He's talking about science. His role is to discover the science that will lay the foundation for the next generation of AI systems, not the current one. He needs to create the scientific foundation for AI products that will exist 10 years from now, the same way he did exactly that 10 years ago for the AI products we have today. First you need the science, then you can have the products. Marie Curie needed to discover radioactivity before we could build nuclear power plants. LeCun isn't at the "building power plants" step; he's at the "creating new science" step, but for machine learning.

He's not telling OpenAI engineers to stop working on LLMs because they're ridiculous. What he means when he says "LLMs are doomed to fail" is aimed at fellow researchers. He's really saying something like: "If you're a young researcher publishing papers in machine learning, be aware that LLM technology is fundamentally limited and probably won't be the foundation for scientific breakthroughs over the next 10, 20, or 30 years." That's not an engineering statement for today, it's a scientific statement for the future. He's not looking down on current AI startups and products. That's not his subject. He's concerned with what scientific breakthroughs machine learning will produce in the coming decades.

What he doesn't care about, and why

Let's list what Yann LeCun doesn't truly care about, not because he's arrogant, but because it's not his job: Meta's corporate management drama, Meta's evolving AI strategy for consumer products, how to improve Meta's LLM technology, how to rival ChatGPT. None of this is related to his job, so none of it is his concern.

His professional goal is to publish good peer-reviewed scientific papers in machine learning, and he's always been clear that he'll continue doing exactly that. Do you think a Nobel Prize winner would want to become a corporate executive? I know I wouldn't. He'll keep working in the area where he has all the fame and experience. He'll keep doing what his life's work has been about: machine learning science.

His new company doesn't change anything

But he's starting his own company now! Doesn't that prove he's more than a researcher? No. He said himself that he won't be the CEO, precisely because, you guessed it, he's a researcher. An older one at that. He instinctively knows his value as a researcher (Nobel Prize level) will always be higher than his value as a CEO, where he has zero experience, no credentials, and probably no personal interest.

The Wang/LeCun drama

All the corporate drama around the hiring of Alexander Wang and the restructuring of Meta's AI divisions... People often interpreted it as if LeCun was an angry executive upset about being sidelined by Zuckerberg. But again, that's wrong, because in a way, he has always been sidelined. He was always isolated from Meta's corporate structure, and that was the plan from the start when he was hired to create FAIR in 2013.

From LeCun's perspective, Meta's corporate maneuvering around its new GenAI strategy isn't really his concern. He's just there to do science. Whether he reports to Wang or any other executive doesn't matter much. The company has no say in what science he produces, because in that domain, he's the authority. A Meta executive who isn't even an academic researcher won't have power over his research direction. First, because corporate executives usually have no experience in scientific research, and second, because you don't tell a Nobel Prize winner what to research. He probably knows better than anyone.

People want to see drama between LeCun, Wang, and Zuckerberg, but the reality is that LeCun has always been pretty isolated from Meta's corporate structure. That being said, maybe Zuckerberg changed his mind about FAIR. Maybe a 100% fundamental research team, essentially a gift from Big Tech to the scientific world, isn't as much of a priority for him anymore. Maybe Zuckerberg and LeCun even disagree on what the future of machine learning science will look like or how important LLMs will be in this future. But in any case, LeCun isn't at fault. He fulfilled his role. He's a researcher who kept researching, exactly as Zuckerberg expected. In or out of Meta, he's going to keep doing what he's always done: scientific research.

The real communication problem

I think that the main source of confusion is that people take what LeCun says in the very specific context of academic science and read it as a statement to the world at large, and to current AI startups and companies in particular. I believe it was never his intention.

Did he fail to anticipate how people would misunderstand him? Probably. Is he too embedded in the scientific community to realize that many people have no idea what his career and role actually are? Probably. Does he care that people misunderstand his statements? Probably not. Whether people understand that he's a researcher doesn't really impact his career. To him, it's probably just people being mean on social media, nothing serious.

But still, I think that this kind of misunderstanding is sad and, in a way, absurd. LeCun presents slides at an academic machine learning conference to his peers, and somehow people interpret it as criticism of the entire AI industry, even though that's clearly not what he's doing. The internet is messy; people will always misinterpret things in the weirdest ways.

But this misunderstanding was strong enough that I felt the need to address it.