Unitree Technology's Wang Xingxing: Aligning AI model inputs and outputs with real robots is the core challenge for embodied intelligence to pass the ChatGPT moment
English text is being prepared.
Wang Xingxing, Chairman and CTO of Unitree Technology, said at the 2026 World Physical AI Conference that the core challenge for embodied intelligence to pass the ChatGPT moment is achieving alignment between the inputs and outputs of AI models and real robots. He believes that top large-model companies have basically achieved model self-evolution internally, with models able to write programs and collect data on their own while humans only need to provide appropriate guidance; however, it will take 5–10 years for technological breakthroughs to spread to everyone's use, and physical AI has already shown signs of being broken through.
Body
据证券时报，10月22日，2026世界物理AI大会暨宇树合作伙伴生态大会在杭州举行，宇树科技董事长、总经理兼首席技术官王兴兴在大会主旨演讲中表示，对于顶尖大模型公司，某种意义上，他们内部已经基本实现了模型的自进化。模型自己写程序、采集数据，人只需要适当引导参与。但大家为什么感受不是那么深刻？因为从技术突破到扩散至每个人使用，需要5—10年的时间。物理AI目前已经看到被击穿的苗头。具身智能迈过ChatGPT时刻最核心挑战是AI模型的输入输出和真实机器人的对齐。