As the AI industry continues to pursue the concepts of "AGI" and "superintelligence," Alexandre LeBrun, CEO of AMI Labs, has chosen to distance himself from such terminology. He states that these terms lack clear definitions and cannot accurately describe the world model approach the company is advancing.
Do not follow the AGI narrative
LeBrun told TechCrunch that AMI Labs never used the term "AGI" and does not endorse the concept of "superintelligence." In his view, this is more a shift in industry narrative than a clearly defined technological boundary.
AMI Labs, co-founded by Turing Award laureate Yann LeCun, focuses on the "world model" approach. These models aim not to predict the next word, but to predict the next state in the real world, enabling AI to better understand how the physical world evolves.
Focus on robot deployment
LeBrun believes that one of the most important applications of world models is robotics. Many current robots still rely on fixed procedures and lack sufficient ability to understand their surroundings and safely perform tasks when entering open environments such as homes or streets.
He said that hardware has made rapid progress in recent months, but the robot still lacks a sufficiently mature "understanding layer." If AI had contextual awareness, misjudgments and risks by robots in public settings could be reduced.
He also emphasized that world models are not meant to replace large language models; rather, the two are more complementary—large language models handle language processing, while world models provide environmental context and judgments about the real world.
Cooperation in Asia continues to advance.
AMI Labs has not yet launched a product but has begun engaging with partners in robotics, manufacturing, and electronics. Last week, during the International Conference on Machine Learning in Seoul, LeBrun also sought local industrial partners, multinational corporations, and researchers.
He stated that world models cannot be trained solely in laboratories; they must be validated in real-world environments, so the company aims to leverage partners to obtain real-world scenario data. South Korea is attractive due to its strong foundation in robotics, semiconductors, and manufacturing, as well as its rapid policy implementation.
JP Lee, CEO of SBVA, AMI’s Asian investor, also noted that the South Korean government has consistently invested in AI in recent years, prioritizing semiconductors, AI data centers, and physical AI.
Funding has been secured, but the product has not yet been announced.
Despite significant external attention, AMI Labs currently has no official product and has not provided a clear release timeline. LeBrun stated only that the team will share more information once they are ready.
The company raised $1.03 billion in March this year, with a pre-money valuation of $3.5 billion. For an AI startup that has not yet launched a product, this funding round highlights the capital market’s strong interest in world models and physical AI.