Why AMI Labs’ Alexandre LeBrun Won’t Call His AI ‘AGI’ or ‘Superintelligence’

2026-07-20 · Source: AI on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

Alexandre LeBrun, CEO of AMI Labs, a world model startup co-founded by Yann LeCun, explicitly rejects industry labels like "AGI" and "superintelligence," deeming them undefined and unhelpful. AMI Labs secured \$1.03 billion in March at a \$3.5 billion pre-money valuation from investors including Samsung Electronics, SBVA, Jeff Bezos' Bezos Expeditions, and Nvidia, despite having no product. The company is developing a Joint Embedding Predictive Architecture (JEPA), a "world model" designed to predict the next "state" or "action" to address AI hallucination and provide real-world understanding, complementing large language models for physical AI applications. LeBrun highlights the critical need for world models in robotics operating in open environments and in healthcare, areas where current LLMs are insufficient. AMI Labs is strategically focusing on Asia, particularly South Korea, to access advanced industrial bases in robotics and manufacturing, and leverage the region's rapid AI adoption for real-world training partnerships.

Key takeaway

For AI Directors or investors evaluating next-generation AI architectures, recognize that "world models" like AMI Labs' JEPA represent a significant, complementary shift beyond LLMs for physical AI. Your investment strategies or development roadmaps should consider integrating these context-aware systems, especially for robotics in unstructured environments or real-world healthcare applications. Prioritize partnerships offering access to diverse physical environments for effective model training, as this approach is crucial for addressing current AI safety and hallucination limitations.

Key insights

World models, specifically JEPA, offer a complementary approach to LLMs for achieving real-world understanding in AI systems.

Principles

• World models predict the next "state" or "action" to mitigate AI hallucination.
• LLMs and world models are complementary, not replaceable, for comprehensive AI.
• Physical AI requires real-world context and understanding for safe operation.

Method

AMI Labs is building a Joint Embedding Predictive Architecture (JEPA) as a "world model" to move AI beyond statistical pattern recognition, requiring real-world environments and partners for training.

In practice

• Integrate world models into robotics for safe operation in open, unstructured environments.
• Apply context-aware AI to healthcare beyond language processing for real-world experience.
• Seek partnerships in regions with advanced industrial bases for real-world AI training.

Topics

• AMI Labs
• World Models
• Joint Embedding Predictive Architecture
• Physical AI
• Robotics
• AI Investment
• South Korea

Best for: Research Scientist, AI Scientist, Director of AI/ML, Investor

Related on AIssential

• Why AMI Labs’ Alexandre LeBrun won’t call his AI ‘AGI’ or ‘superintelligence’ — World models, distinct from LLMs, provide physical intuition for context-aware robotics, addressing real-world safety and understanding.
• A Tutorial on World Models and Physical AI — World models unify prediction, planning, and generalization by internalizing environment dynamics for efficient decision-making.
• What is Advanced Machine Intelligence or AMI Labs? — A new wave of AI research is shifting from AGI to Superhuman Adaptable Intelligence and Physical AI.
• Simulating everything, sort of: The promise and limits of world models — World models, simulating physical reality beyond language, are the next frontier in AI, attracting significant investment.
• Yann LeCun’s AMI Labs Launches With $1.03 Billion to Build AI That Understands the Real World — AMI Labs is pioneering "world models" to enable AI to understand physical reality, moving beyond LLM limitations.
• The Sequence Knowledge #842: Everything You Need to Know About World Models — World models represent the shift from text-based AI to physical simulation, enabling AI to understand and operate within dynamic reality.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI on Medium.