Unitree’s Wang Xingxing: The “ChatGPT Moment” for Embodied Intelligence Requires Alignment Between AI Models and Real Robots

Bot Telegraph, August 20th – At the 2026 World Robot Conference, Wang Xingxing, CEO of Unitree Robotics, proposed that the “ChatGPT moment” for embodied intelligence has arrived, requiring a specific standard: robots must be able to complete approximately 80% of tasks in 80% of unfamiliar scenarios through voice or text commands. He pointed out that the biggest bottleneck currently preventing embodied intelligence from crossing this threshold lies in the deviation between the instructions output by AI models and the actual actions executed by real robots—namely, the alignment issue between input/output and the physical world. To tackle this challenge, Unitree Robotics is developing a solution called “Physical AI Robot Self-Evolution V1.0.” This approach enables AI models to autonomously write robot control code, which is then trained and validated in a simulated environment before being deployed on real robots for feedback analysis. Wang Xingxing emphasized that the effectiveness of this self-evolution process depends on three key factors: the strength of the foundational model, the scale of multi-source data, and the deployment scale of real robots. The more real robots deployed, the faster data accumulation and capability iteration become, thereby driving continuous skill accumulation and expansion, ultimately forming a large-scale skill ecosystem.

Source: Bot Telegraph China