Amid the AI hype around AGI and superintelligence, AMI Labs founder Alexandre LeBrun dismisses both labels as vague and unhelpful, focusing instead on concrete world‑model research.

Key Takeaways

  • AMI Labs refuses to use the terms ‘AGI’ or ‘superintelligence’.
  • World models aim to bring real‑world context to robotics and other physical‑domain applications.
  • South Korea’s advanced hardware ecosystem draws AMI Labs’ early‑stage Asia strategy.

While most AI firms scramble to brand their breakthroughs as “AGI” or “superintelligence,” Alexandre LeBrun, CEO of Yann LeCun’s world‑model startup AMI Labs, has deliberately avoided the buzzwords. In a TechCrunch interview conducted in Seoul during the International Conference on Machine Learning, LeBrun explained, “We never used the word AGI, and now nobody even says ‘superintelligence’—they’ll switch to something else next.” He argues that the lack of a solid definition makes the terms more noise than guidance.

What a World Model Means

A world model integrates physics to predict the next state of the environment, moving beyond the token‑by‑token prediction of large language models (LLMs). LeBrun points out that robotics stands to gain the most: today’s robots follow fixed, “completely static” routines, and AI remains “really dumb in the physical world.” Adding contextual awareness could prevent absurd mishaps—such as a dancing robot accidentally kicking a child at a public event.

Read Also:Russia-Aligned UAC-0099 Deploys 'GuardBreaker' Malware to Sabotage AI Security Analysis

LLMs vs. World Models

LeBrun does not claim that world models replace LLMs; rather, they are complementary. “LLMs are the most efficient tools for processing language, while world models provide context and real‑world understanding,” he says, drawing a parallel to the human brain’s distinct language and reasoning functions.

Why South Korea?

AMI Labs is still pre‑product, but the company is courting robotics, manufacturing, and electronics partners. South Korea’s deep industrial base—spanning robotics, semiconductors, and heavy manufacturing—offers the real‑world data needed to train a world model outside a lab. LeBrun adds that Korea’s rapid adoption of new technologies, exemplified by its early internet uptake, makes it an “ideal launchpad” for AI hardware and data‑center investments, including the government’s $880 billion plan for chips and physical AI.

Future Outlook

Despite raising $1.03 billion in March at a $3.5 billion pre‑money valuation, AMI Labs has no product to market yet. LeBrun remains confident, promising a surprise when the technology is ready. His refusal to cling to vague labels suggests that the next wave of AI will be judged on tangible capabilities rather than lofty terminology.