Meta's Chief AI Scientist Yann LeCun is reportedly leaving the tech giant to launch his own AI startup after declaring large language models (LLMs) a "dead end" for reaching human-like AI. This departure comes as Meta shifts its focus towards advanced AI systems and superintelligence research, with LeCun's long-term work at Meta's Fundamental AI Research (FAIR) lab being put on the backburner.
LeCun's exit marks another major shake-up at Meta, which is doubling down on its research into world models. According to sources familiar with the matter, LeCun's new venture will focus on training AI systems to understand the physical world rather than generating language alone. This move isn't surprising given his skepticism about LLMs, which he has previously called a "dead end" for achieving human-like intelligence.
LeCun believes that world models are key to developing AI that can reason, plan complex actions, and make predictions. He argues that training on text data will never be enough to achieve human-level intelligence. This perspective is echoed by other researchers and companies, including Stanford's Fei-Fei Li, who has raised over $230 million for her startup World Labs.
Google DeepMind has also explored world models through its Genie releases, while Nvidia is pushing into physical AI with products like its Cosmos world models. LeCun's new venture could potentially rival these efforts, as he brings his expertise and knowledge of machine learning to the table.
The French-American computer scientist joined Meta in 2013 and played a key role in launching the company's FAIR team. He became Meta's chief AI scientist in 2018 and was awarded the Turing Award for his contributions to neural networks. However, following the company's recent restructuring, LeCun now reports to Alexandr Wang, head of Meta's superintelligence division.
LeCun's departure from Meta marks a significant shift in the company's focus towards world models. The FAIR lab, which contributed to early versions of Meta's Llama model, has suffered setbacks in recent months, including the loss of its leader Joelle Pineau.
LeCun's exit marks another major shake-up at Meta, which is doubling down on its research into world models. According to sources familiar with the matter, LeCun's new venture will focus on training AI systems to understand the physical world rather than generating language alone. This move isn't surprising given his skepticism about LLMs, which he has previously called a "dead end" for achieving human-like intelligence.
LeCun believes that world models are key to developing AI that can reason, plan complex actions, and make predictions. He argues that training on text data will never be enough to achieve human-level intelligence. This perspective is echoed by other researchers and companies, including Stanford's Fei-Fei Li, who has raised over $230 million for her startup World Labs.
Google DeepMind has also explored world models through its Genie releases, while Nvidia is pushing into physical AI with products like its Cosmos world models. LeCun's new venture could potentially rival these efforts, as he brings his expertise and knowledge of machine learning to the table.
The French-American computer scientist joined Meta in 2013 and played a key role in launching the company's FAIR team. He became Meta's chief AI scientist in 2018 and was awarded the Turing Award for his contributions to neural networks. However, following the company's recent restructuring, LeCun now reports to Alexandr Wang, head of Meta's superintelligence division.
LeCun's departure from Meta marks a significant shift in the company's focus towards world models. The FAIR lab, which contributed to early versions of Meta's Llama model, has suffered setbacks in recent months, including the loss of its leader Joelle Pineau.
Can you believe Yann LeCun is leaving Meta? I mean, who needs a social media platform when you can just invent your own AI to take over the world? 
Just kidding, kinda. Anyway, it seems like he's onto something with these "world models" and I'm low-key excited about it 
. It's like he said, generating language is so last season. Who needs human-like intelligence when you can just build a robot that can understand the physical world? 
That's some next-level thinking right there! 
. i mean, anyone who thinks training ai solely on text data is gonna get us to true intelligence needs to reevaluate their life choices 
. it's like, can't we all just get along... and have more diverse AI goals? 
He's all about world models now, which makes sense considering he thinks LLMs are a dead end 
I mean, who wouldn't wanna tackle that kinda challenge? Fei-Fei Li and Stanford's World Labs are already on it, so this isn't like anyone is starting from scratch 
Nvidia's Cosmos world models are pretty cool too... LeCun's expertise is gonna be a game-changer 
. And have you seen Google DeepMind's Genie releases? That stuff is some next-gen world modeling 
. So, if LeCun's new venture takes off, it's gonna be a major player in the AI game 
.
. It feels like a bit of deja vu, but with LeCun's departure, we're getting a fresh perspective on what AI research should look like 
.
i mean, LLMs are cool and all but we need more than just language understanding for human-like AI 
he was super key in developing those large language models at meta but now he's all about world models 
this is huge news! Yann LeCun's departure from Meta and his new AI startup sounds like a game-changer 
. let's see what lecun's new venture will bring to the table! 
. I mean, he's been saying that for ages and Meta was like "yeah we got this" 
. But now they're shifting their focus to world models which is actually kinda cool 
. And who knows, maybe his departure will give Meta some space to figure out where they really want to go with their AI research 
... LeCun is like a legend in the AI space and he just left Meta for his own startup. He thinks LLMs are dead ends? That's pretty bold 
. I'm curious to see what he'll come up with next. Maybe we'll finally get some real progress on AI development 
.
AI researcher Yann LeCun is leaving Meta after saying large language models are a "dead end" for creating human-like AI 
. This means his expertise and knowledge will now be focused on world models, which try to understand the physical world rather than just generate text 
he invested so much time and effort into meta and fair lab but now it seems like his work isn't valued anymore 
just not gonna cut it. World models, on the other hand, are like building a skyscraper - whoa, that's ambition! 