Unitree's Wang Xingxing: Biggest Past Mistake Was Not Trusting AI Enough

At the opening ceremony of the 26th China International Industry Expo on Oct. 12, Wang Xingxing, founder of Unitree Technology, said during a deep-dialogue session that the recent release of GPT-6 had a major impact on the embodied intelligence industry's thinking. His team also deployed the model on robots to carry out task tests, and the results were fairly good, he said.

Wang said the tipping point for embodied intelligence is likely to arrive relatively quickly, possibly within one or two years. To enable large-scale robot applications, the industry still needs to solve problems of generalization and efficiency, but he expects those issues to be resolved in the coming years.

Wang added that his own belief in AI has only grown over time. He described his biggest mistake over the past decade as not trusting AI enough. He also observed that many people have still not fully used or embraced AI, suggesting that adoption remains limited even as model capabilities advance.

The remarks, reported by Yicai, highlight how foundation models are increasingly shaping robotics research and product planning. Unitree's tests with GPT-6 indicate that leading robotics firms are evaluating cutting-edge AI systems for real-world tasks, while Wang's timeline places the sector's critical turning point much closer than some forecasts. If generalization and efficiency improve as expected, robots could move from specialized demonstrations to broader deployment, though reliable hardware, data and safety will remain key conditions.