Alibaba’s Qwen AI models have demonstrated strong capabilities in understanding the physical world through code and screenshots but lack the ability to act within it. The company is addressing this gap with new models: Qwen-RobotNav, Qwen-RobotManip, and Qwen-RobotWorld. These models aim to close the bottleneck in embodied intelligence, focusing on translating perception and reasoning into physical actions.

• Qwen AI models can perceive and reason about the physical world but cannot physically interact with it.
• Alibaba’s Tongyi Lab identified the gap between understanding and action as a bottleneck for embodied intelligence.
• Three new models – Qwen-RobotNav, Qwen-RobotManip, and Qwen-RobotWorld – are being introduced to address this.
• The challenge lies in the translation layer between seeing a need and executing the required physical movements.