Chief AI Scientist At Mark Zuckerberg's Meta Says 'No Way' Scaling ChatGPT-Like Models Is Going To Lead To Human-Level AI
Meta's chief AI scientist, Yann LeCun, asserts that scaling large language models will not lead to human-level AI, dismissing optimistic timelines from peers as unrealistic. He emphasizes that current chatbots lack true reasoning capabilities, functioning instead through pattern matching. LeCun advocates for a collaborative approach, urging governments to share anonymized data for open-source models. He has expressed skepticism about OpenAI's chances of achieving artificial general intelligence and continues to engage in a public feud with Elon Musk over AI advancements. Meta shares rose 3.16% to $687.95 following these comments.
Meta Platforms, Inc.'s META chief AI scientist, Yann LeCun, says the tech industry won't close the gap to human-level intelligence by scaling today's large language models and piling on more parameters.
What Happened: "We are not going to get to human-level AI by just scaling up LLMs. This is just not going to happen. There's no way — absolutely no way," LeCun told host Alex Kantrowitz on the Big Technology podcast in March. He dismissed bullish two-year timelines from "more adventurous colleagues" as "complete BS."
In a clip of the podcast which was resurfaced on YouTube last week, LeCun likened current chatbots to "a system with a gigantic memory and retrieval ability, not a system that can invent solutions to new problems," adding that even if the models can answer most routine questions, "it's not a Ph.D. you have next to you." Instead of reasoning, he said, today's systems "pattern-match" the next word.
LeCun contends the best path forward is collaborative. According to a report by Business Insider, at the AI Action Summit in Paris, which took place in February, he urged governments to contribute anonymized data to a larger open-source model.
See also: Jeff Bezos Let Customers Cancel Their Own Amazon Orders And It Paid Off Big: ‘If You Give People More Control…Maybe They’ll Order More’
Why It Matters: LeCun has long doubted that OpenAI will win the race to artificial general intelligence (AGI), a stance he first voiced in December 2023. Last week, he pointed Elon Musk toward a new FAIR study on "Contextual Positional Encoding," telling the xAI founder it could boost Grok and then amplified the paper by sharing Meta researcher Jason Weston's explanatory thread on X.
The exchange unfolded amid LeCun's running feud with Musk. After Musk posted xAI job openings on Monday, LeCun quipped that applicants should expect a boss who insists their project "will be solved next year." He later applauded Musk's engineering triumphs in cars, rockets, and satellites while slamming the billionaire's politics, conspiracy theories, and habitual hype.
Price Action: Meta shares closed 3.16% higher at $687.95 on Wednesday, according to Benzinga Pro.
Benzinga Edge Stock Rankings indicate Meta has a Momentum in the 88th percentile and a strong Growth score in the 93rd percentile. Check to see how other Magnificent Seven stocks rank in comparison.
Photo Courtesy: Tapati Rinchumrus on Shutterstock.com
- Jim Cramer Says The Chinese ‘Could Really Hurt Us’ As Beijing Tightens Grip On Rare Earth Magnets: Warns Of Threat To ‘Everything’ From Autos To Data Centers
