"Don't Do LLM": Yann LeCun Founded AMI with $1.03 Billion Bet on World Model, Saining Xie Joins

According to 1M AI News, AMI (Advanced Machine Intelligence), the world modeling company founded by Turing Award winner and former Chief AI Scientist at Meta Yann LeCun, has completed a $1.03 billion financing round, with a pre-money valuation of $3.5 billion. The company, initially planning to raise about €500 million in December last year, ultimately raised around €890 million. AMI, headquartered in Paris, will establish offices in New York, Montreal, and Singapore.

Yann LeCun serves as Chairman, with the CEO role taken by French serial entrepreneur Alexandre LeBrun, whose previous AI healthcare assistant company Nabla will become AMI's first partner. Assistant Professor at New York University and co-originator of the Diffusion Transformer (DiT) architecture Saining Xie has been appointed Chief Scientific Officer, with Michael Rabbat, head of the V-JEPA series, as Vice President of World Modeling, Pascale Fung, Chair Professor at The Hong Kong University of Science and Technology, as Chief Research and Innovation Officer, and Laurent Solly, former Vice President of Meta Europe, as Chief Operating Officer. Four of the six core founders are direct from Meta AI Research lab FAIR. This round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, with participation from Nvidia, Samsung, Temasek, Toyota Ventures, Yang Shi Group, and individual investors including former Google CEO Eric Schmidt, World Wide Web inventor Tim Berners-Lee, and prominent investor Marc Andreessen.

AMI follows the technical roadmap of JEPA (Joint Embedding Prediction Architecture) proposed by Yann LeCun during his time at Meta, aiming to enable AI to learn abstract representations of the world rather than relying on textual statistical regularities like large language models. CEO LeBrun admits that AMI is not a typical applied AI startup, and the transition from fundamental research to commercialization may take several years, but the company is committed to publishing papers and open-sourcing substantial amounts of code. Yann LeCun has stated, "Don't do LLM, it's meaningless, you can't catch up with the industry. Breakthroughs won't come from scaling up LLM."