Advanced Machine Intelligence Labs (AMI Labs) is a Paris-based artificial intelligence research laboratory co-founded by Yann LeCun, the former Chief AI Scientist at Meta Platforms, in December 2025. The organization focuses on developing world model AI systems that can understand the physical world, maintain long-term memory, and strategize complicated tasks, while ensuring they remain secure and easily manageable. AMI Labs raised $1.03 billion in a seed funding round from investors including Bezos Expeditions, Nvidia, and Samsung Electronics, and operates locations in Paris, New York City, Montreal, and Singapore.
AMI Labs was established against the backdrop of LeCun's departure from Meta in November 2025. LeCun argued that Meta and other leading AI companies would eventually reach a dead end with their single-minded focus on large language models (LLMs) and set out to build a new company centered on developing a more comprehensive world model for intelligent machines. The lab is intended to remain a research organization, not expected to produce a saleable product for around five years.
Background and Founding
Yann LeCun, a prominent figure in the field of Artificial intelligence and a recipient of the Turing Award, left his position at Meta (AI research) in November 2025. His departure was motivated by a belief that the prevailing approach to AI, dominated by Large language models, was fundamentally limited. LeCun has long advocated for a shift toward world models, which are AI systems that can learn internal representations of the physical world, enabling them to predict outcomes and plan actions. This vision became the foundation of AMI Labs, which was officially launched in December 2025.
The seed funding round, which raised $1.03 billion, was one of the largest ever for an AI startup at that stage. Investors included Bezos Expeditions, the personal investment vehicle of Amazon founder Jeff Bezos; NVIDIA, a leading manufacturer of AI hardware; and Samsung Electronics, a global technology conglomerate. The substantial funding underscored investor confidence in LeCun's vision and the potential of world models to transform AI capabilities.
AMI Labs has established a global presence with offices in Paris, New York City, Montreal, and Singapore. Paris serves as the headquarters, reflecting LeCun's French origins and the city's growing status as a hub for AI research. The Montreal office is particularly significant, as the city is home to a vibrant AI research community, including the Canadian Institute for Advanced Research (CIFAR) AI and the MILA (alternative).
Research Focus: World Models
The core research focus of AMI Labs is the development of world models. Unlike traditional AI systems that rely on pattern recognition in text, world models aim to create a comprehensive understanding of the physical world. These systems are designed to:
- Understand the physical world: They learn how objects interact, how time progresses, and how actions lead to consequences.
- Maintain long-term memory: They can store and retrieve information over extended periods, enabling them to build on past experiences.
- Strategize complicated tasks: They can plan multi-step actions to achieve specific goals, similar to human reasoning.
LeCun has described world models as a necessary step toward achieving human-level AI, which he believes cannot be attained through Deep learning on text alone. He has often criticized the limitations of Transformer (architecture)-based models, which are the basis of most modern LLMs, for their lack of common sense and their inability to understand cause and effect.
AMI Labs' approach is distinct from that of companies like OpenAI and Google DeepMind, which have focused on scaling up LLMs. Instead, AMI Labs aims to integrate insights from Robotics, Neural network research, and cognitive science to build systems that can interact with and reason about the world.
Initial Partnerships and Applications
AMI Labs initially plans to develop its world models together with companies from data-intensive industries, such as manufacturing, biomedicine, and robotics. This collaborative approach is intended to ground the research in real-world applications and ensure that the models are useful in practical settings.
Its first official partner is the French startup Nabla, also led by Alexandre LeBrun, which develops AI software for doctors that automatically documents patient conversations and creates medical notes. This partnership highlights the potential of world models to handle complex, real-world data beyond text, such as audio, video, and sensor inputs.
In the biomedical field, world models could be used to simulate biological processes, aiding in drug discovery and personalized medicine. In manufacturing, they could optimize production lines by predicting equipment failures and planning maintenance. In robotics, they could enable robots to navigate and manipulate objects in unstructured environments, a task that remains challenging for current AI systems.
Competition and Landscape
AMI Labs is not the only research laboratory working on world models. The US company World Labs, founded by AI researcher Fei-Fei Li, is also focused on this approach. Additionally, Google DeepMind, OpenAI, and Meta (AI research) have been working on world models for many years, though often as a complement to their LLM-based systems.
Google DeepMind has explored world models in the context of reinforcement learning and simulation, while OpenAI has investigated them for planning and decision-making. Meta, despite LeCun's departure, continues to fund research in this area. The competition is intense, but AMI Labs' dedicated focus and substantial funding position it as a major player in this emerging field.
Other notable organizations in the broader AI landscape include Anthropic, known for its safety-focused approach, and Together, a platform for open-source AI. However, AMI Labs' unique emphasis on world models distinguishes it from these peers.
Funding and Growth
The $1.03 billion seed round was a landmark event in the AI industry. It provided AMI Labs with the resources to attract top talent, build state-of-the-art computing infrastructure, and support research over the long term. The involvement of Nvidia is particularly strategic, as the company's GPUs are essential for training large AI models, and a partnership could ensure access to cutting-edge hardware.
Samsung Electronics' investment signals interest in applying world models to consumer electronics and robotics, areas where Samsung has a strong presence. Bezos Expeditions' participation reflects Jeff Bezos' long-standing interest in AI and his belief in the transformative potential of world models.
As of early 2026, AMI Labs is in the process of expanding its team and establishing its research agenda. The company has not announced any products, consistent with its stated intention to focus on research for the first five years.
Leadership and Team
Yann LeCun serves as the co-founder and chief scientist of AMI Labs. His leadership is a major draw for researchers in the field. LeCun is known for his contributions to Deep learning, including the development of convolutional neural networks, and for his role in shaping the modern AI landscape.
The company has not publicly disclosed its full leadership team, but it is expected to include researchers from leading institutions such as Meta (AI research), Google (AI), and OpenAI. The presence of offices in multiple countries suggests a global recruitment strategy, aiming to attract talent from diverse AI communities.
Future Directions
AMI Labs' long-term goal is to create AI systems that can understand and interact with the world in a human-like manner. This includes developing models that can learn from limited data, generalize across tasks, and reason about causality. The company plans to publish its research, contributing to the broader scientific community, while also building proprietary technologies.
The focus on safety and manageability is a key differentiator. LeCun has emphasized the importance of designing AI systems that are secure and controllable, addressing concerns about the potential risks of advanced AI. This aligns with the work of organizations like the Artificial intelligence safety institute, though AMI Labs approaches safety from a technical rather than policy perspective.
In the coming years, AMI Labs is expected to release research papers and possibly open-source some of its models, following the tradition of other AI labs. The company's success will depend on its ability to make breakthroughs in world model research, a challenge that has eluded many researchers for decades.
See Also
References
- Source facts provided from Wikipedia (CC BY-SA).