Technology and Science16 July, 20262 min read
AMI Labs CEO Alexandre LeBrun Rejects AGI Label, Pursues World Model For Robotics
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
LeBrun avoids labeling AMI's AI as AGI or superintelligence. AMI Labs pursues world models rather than AGI, distancing from hype.
World models over labels
AMI Labs CEO Alexandre LeBrun says his company avoids AGI and superintelligence terminology while pursuing a world model aimed at bridging simulation and the physical world.
LeBrun told TechCrunch, “We never used the word AGI. And I just noticed that nobody is using it anymore; they switched to superintelligence,” and he added, “There’s no good definition. What is superintelligence? I don’t know. It’s not a very useful word.”
TechCrunch reported that LeBrun met in Seoul last week for The International Conference on Machine Learning while scouting for local industrial partners, global companies, and researchers.
In LeBrun’s account, AMI Labs is still pre-product but is courting robotics, manufacturing, and electronics players, and it argues that a world model needs to prove itself outside the lab.
He frames the technical goal as complementing large language models by using physics to predict and work with the real world, while saying robots are “completely static” and AI in the physical world remains “really dumb in the physical world.”
Robots, safety, and context
LeBrun argues that world models are meant to provide context and real-world understanding that fixed routines cannot, describing robots as running “completely static” schemes. In TechCrunch’s reporting, he said a context-aware system could have prevented a robot “dancing and doing kung fu at a public event” from approaching and kicking a child. He also told TechCrunch that “Robots are not safe right now,” and he said “There’s no solution for that today.”
The Tech Buzz similarly quotes LeBrun’s refusal to chase labels, saying, “We’re not chasing labels,” as AMI Labs builds world models that predict and simulate how the physical world works rather than processing text. Across both accounts, LeBrun positions world models as predicting the next state of the world—contrasting with LLMs that predict the next word or text—while insisting the two are “complementary, not replaceable” for physical-world AI.
Asia push and funding stakes
LeBrun told TechCrunch that AMI needs “access to the real world,” and he said it is “easier for us to do that with partners,” which he links to his pull toward Asia. In TechCrunch’s account, he said Korea’s appeal rests on advanced industries in robotics, semiconductors, and manufacturing, plus speed, pointing to Korea’s track record as “the fastest adopter of the internet 25 years ago.” JP Lee, CEO of SBVA and one of AMI’s backers in Asia, told TechCrunch that he has been “telling Alex and the team to come to Korea,” and he said the government has done “a tremendous job” funding local sovereign LLM models.
The Tech Buzz adds that AMI Labs was co-founded with Yann LeCun, described as Meta’s chief AI scientist and a Turing Award winner, and it frames the company’s approach as a contrarian bet against an AGI marketing playbook. In the broader reporting, LeBrun’s stance is tied to what AMI can build next—world models for physical AI—while the sources emphasize that AMI Labs remains pre-product despite raising over $1.03 billion with an implied valuation of around $3.5 billion.