Turing Award–winning AI researcher Yann LeCun has spent years arguing that large language models, at least in their current form, will not lead to truly intelligent machines. Now he has secured more than $1 billion to try a different approach.
Advanced Machine Intelligence (AMI), the startup founded by the former chief AI scientist at Meta, announced Tuesday that it raised $1.03 billion in seed funding. The round values the company at $3.5 billion before the investment and ranks among the largest early-stage finance rounds for an AI research startup.
The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, with participation from a long list of venture firms and technology investors. The funding will support the company’s development of AI systems built around reasoning, planning and so-called “world models,” an AI research direction LeCun has promoted as an alternative to today’s dominant LLM approach.
AMI released this image with its announcement, featuring the Veil Nebula and taken by LeCun from his backyard (Credit: AMI)
“We are enabling the next AI revolution by building a new breed of AI systems that (1) understand the real world, (2) have persistent memory, (3) can reason and plan, (4) are controllable and safe,” AMI states on its website.
LeCun, who left Meta in November 2025 after more than a decade leading Facebook AI Research, has said that predictive models will not lead to artificial general intelligence. While language models have demonstrated strong performance in areas like coding and summarization, LeCun has been critical of their lack of ability to understand the physical world and plan actions within it.
AMI is looking to change that. The company is developing systems designed to learn structured representations of real-world environments and predict how those environments will evolve over time. In theory, these models could allow machines to reason about the consequences of their actions and plan sequences of behavior, capabilities that researchers say are necessary for robotics and other complex applications.
So far, the startup has a small team, but it has already attracted a group of researchers and executives with experience at major AI labs. Alex LeBrun, a former Meta engineer and founder of the healthcare AI company Nabla, is serving as CEO. LeCun is chairman and remains a professor at New York University.
Yann LeCun
“It is time to move beyond shortcuts and work on a foundational solution. World models learn abstract representations of real-world data, ignoring unpredictable details, and make predictions in representation space,” LeBrun wrote in a post announcing the seed round. “Action-conditioned world models allow agentic systems to predict the consequences of their actions, and to plan action sequences to accomplish a task, subject to safety guardrails.”
LeBrun told the New York Times the company will initially operate much like a research laboratory while exploring possible applications of its technology. The approach is based in part on Joint Embedding Predictive Architecture, or JEPA, a framework LeCun proposed in 2022 (before the explosion of LLMs) that works by learning abstract representations of data rather than directly generating text or images.
The company’s target customers include organizations that operate complex physical systems, such as manufacturers, aerospace companies, biomedical firms and pharmaceutical groups. These industries often require AI systems that can operate reliably in dynamic environments where errors can have significant consequences.
“AMI will advance AI research and develop applications where reliability, controllability, and safety really matter, especially for industrial process control, automation, wearable devices, robotics, healthcare, and beyond,” LeBrun said.
(sopa phetcharat/Shutterstock)
Robotics is currently a hot area of interest for AI research. Researchers in the field have had difficulty building systems that can adapt to unfamiliar situations outside of controlled environments. LeCun and his colleagues believe that models trained to understand the structure of the physical world could help robots work more effectively in settings like homes, factories or hospitals. AMI’s first disclosed partner is LeBrun’s company Nabla, which plans to explore how the new models could complement language-based AI systems used in healthcare applications.
The funding round also reflects continued investor enthusiasm for AI startups led by prominent researchers, even as some analysts still warn of a potential AI investment bubble. Several other startups founded by former researchers from large technology firms have recently raised large sums, often before releasing commercial products.
For LeCun, AMI represents an attempt to push AI research in a direction he believes has been overshadowed by the mania surrounding generative models. Whether that approach produces practical systems remains to be seen. But with more than $1 billion in backing, AMI now has significant resources to advance its research agenda.
The post Yann LeCun’s AMI Secures $1B Seed to Develop AI World Models appeared first on AIwire.
