AI industry focus on physical and embodied intelligence

The artificial intelligence industry is increasingly focusing on the integration of AI with the physical world, often referred to as physical or embodied intelligence.

In Europe, industry leaders have highlighted the importance of capital-efficient physical AI and the value of in-person collaboration for startup momentum. For instance, Minimal aims to deploy autonomous vehicles for urban goods movement in the UK, while founders at events in London and Turin emphasized that “the real trust, I think, happens in person” to foster connections.

This focus has expanded to include discussions on world models and predictive architectures. In South Korea and China, experts have advocated for AI that moves beyond language learning to understand physical environments through “world models.” Industry leaders suggest that advancements in large models could lead to a tipping point for large-scale robotic applications within one to two years, provided challenges in generalization and efficiency are addressed.

Recent developments reinforce the importance of these models. AI pioneer Yann LeCun, founder of AMI Labs, asserts that the next breakthrough will stem from “world models” capable of understanding and predicting the physical world, rather than large language models alone. LeCun noted that while LLMs excel at predictable sequences, they struggle with the complexities of physical reality. AMI Labs is utilizing Joint Embedding Predictive Architecture (JEPA) to develop systems for autonomous planning and interaction, recently securing $1.03 billion in seed funding from investors including Nvidia and Samsung.

In the robotics sector, expectations are recalibrating. Unitree Robotics founder Wang Xingxing noted that embodied intelligence could arrive within one to two years, while DeepSeek is reportedly exploring the market. At the Hong Kong Autumn Electronics Fair, companies showcased practical applications, including Acceleration Evolution’s T4 humanoid robot and SenseRobot’s AI-integrated chess hardware.