Unitree CEO predicts GPT-7 could advance robot intelligence by 2027

Unitree CEO Wang Xingxing predicts that the development of GPT-7 could lead to a decisive shift in robotic intelligence by 2027. This advancement aims to solve challenges in embodied intelligence, which is the ability of an AI system to perceive, make decisions, and physically interact with its environment.

According to Wang, a significant turning point will occur when a robot can successfully perform 80% of requested tasks in 80% of unknown environments using only voice or text commands. This potential leap is supported by recent progress seen in models like GPT-6 Astra, which has already demonstrated capabilities in testing environments, such as a Unitree machine performing cleaning tasks in an unfamiliar kitchen through a collaboration with Stanford University’s Movement Lab.