Advanced Machine Intelligence Labs (AMI Labs) has secured £760 million in funding to develop a new class of artificial intelligence designed to operate in real-world environments.
Announced at VivaTech in Paris, the initiative is led by founder Yann LeCun, who argued that to-day’s generative AI systems lack the intelligence required for autonomous, real-world performance.
Backed by major investors including NVIDIA and Jeff Bezos, the funding will support the develop-ment of Joint Embedding Predictive Architecture (JEPA) - a so-called “world model” designed to move AI beyond statistical pattern recognition.
Moving beyond generative AI
Generative AI has dominated recent innovation cycles, powering everything from content creation to coding assistants. However, these systems rely on probability rather than an understanding of how the physical world works.
According to LeCun, this limitation makes them inherently reactive. While they can generate highly convincing outputs, they struggle to reliably predict real-world outcomes - particularly in environments governed by physics and causality.
Scaling large language models alone, he argues, will not deliver the level of intelligence needed for advanced robotics or autonomous systems.
JEPA and the rise of world models
AMI Labs’ answer lies in world models. AI systems designed to simulate how environments behave. JEPA aims to build an internal representation of the world, enabling AI to:
- Understand cause-and-effect relationships
- Filter out irrelevant data
- Predict what will happen next before acting
Rather than generating outputs based purely on statistical likelihood, JEPA focuses on modelling reality itself.
For example, instead of simply describing what happens when a glass is dropped, a JEPA-based system would anticipate the outcome based on physical reasoning - from gravity to surface impact.
Bridging the gap between reaction and prediction
This approach represents a significant shift in AI development.
Where generative models are reactive, JEPA introduces a predictive layer - allowing systems to simu-late different scenarios and choose the most appropriate response.
By focusing only on meaningful variables, the architecture is designed to improve both efficiency and decision-making accuracy in complex environments.
What JEPA means for the future of AI
World models such as JEPA could play a critical role in advancing AI beyond digital tasks into physi-cal, real-world applications.
Potential use cases include:
- Autonomous robots operating independently
- Real-time decision-making systems
- More natural and responsive human-machine interaction
If successful, AMI Labs’ approach could mark a turning point in the evolution of AI - shifting the in-dustry from content generation to true environmental intelligence.
Today’s news was brought to you by TD SYNNEX – the UK’s number one solutions distributor.
Destination AI
Destination AI from TD SYNNEX is our comprehensive end-to-end programme designed to help you and your customers get ahead of the AI curve.
