Unitree Technology's Wang Xingxing says real-robot alignment is the core challenge for embodied intelligence
BIBIBI
AT A GLANCE
Wang Xingxing said that the core challenge for embodied intelligence to cross the “ChatGPT moment” is AI aligning model inputs and outputs with real robots.
Article
Unitree Technology's Wang Xingxing:AI Aligning model inputs and outputs with real robots is the most critical challenge for embodied intelligence to cross the ChatGPT moment
PANews October 22 reported, 2026 World Physical AI Conference and the Unitree Partner Ecosystem Conference was held in Hangzhou today. In his keynote speech, Wang Xingxing, Chairman, General Manager, and Chief Technology Officer of Unitree Technology, said that, for leading large-model companies, they have in a sense largely achieved the self-evolution of models internally. Models can write programs and collect data themselves, while people need only provide appropriate guidance and participate. But why do people not feel this so deeply? Because it takes time for a technological breakthrough to spread to everyone for use. 5—10 years. Physical AI has now shown signs of being breached. The most critical challenge for embodied intelligence to cross the ChatGPT moment is AI aligning model inputs and outputs with real robots.
PANews reported that the 2026 World Physical AI Conference and the Unitree Partner Ecosystem Conference were held October 22 in Hangzhou.
02
Wang Xingxing, Chairman, General Manager, and Chief Technology Officer of Unitree Technology, expressed the relevant views in his keynote speech at the conference.
03
Wang Xingxing said that leading large-model companies have, in a sense, largely achieved model self-evolution internally.
04
Wang Xingxing said that models can write programs and collect data themselves, while people need only provide appropriate guidance and participate.
05
Wang Xingxing said that it takes 5—10 years for a technological breakthrough to spread to everyone for use.
06
Wang Xingxing said that Physical AI has now shown signs of being breached.
07
Wang Xingxing believes that the most critical challenge for embodied intelligence to cross the “ChatGPT moment” is AI aligning model inputs and outputs with real robots.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
Wang Xingxing means that AI even if a model's capabilities improve, the information it receives and the instructions it gives must correspond with a real robot's perception and actions in order to further advance embodied intelligence. “Model self-evolution” and “signs of being breached” in the article are both his industry judgments.
Why it matters to readers
These remarks point to a key technical area of observation for the development of embodied intelligence:AI whether models can effectively align with real robots. However, the original text did not provide specific solutions, test results, or commercialization data.
Beginners should focus on “model-robot alignment” and should not directly interpret technical views as meaning that products are already mature, have achieved mass production, or are generating revenue.
Risks and unknowns
The original article is PANews' paraphrase of the speech and does not provide a complete record of the speech.
“Largely achieved model self-evolution” and “Physical AI has shown signs of being breached” are Wang Xingxing's views; the original article did not provide supporting verification materials.
5—10 years as a diffusion cycle is a judgment made in the speech, and the original article did not explain the basis for the estimate.
The original article did not disclose specific alignment solutions, technical demonstrations, testing metrics, product progress, orders, mass production, or revenue data.
Related Developments
Loading event timeline…
Related concepts
Embodied intelligence
This term is not in the glossary yet. Browse related concepts in the glossary.
