Wang Xingxing: Embodied intelligence will usher in the "ChatGPT moment" at the earliest in 2 to 3 years. Wang Xingxing, founder, CEO and CTO of Yushu Technology, delivered a speech at the main forum of the 2026 World Robot Conference (WRC2026). He said that embodied intelligence is approaching an industry critical point similar to the "ChatGPT moment". It is expected to take 2 to 3 years at the earliest. Slower sentiment

Wang Xingxing: Embodied intelligence will usher in the "ChatGPT moment" at the earliest in 2 to 3 years. Wang Xingxing, founder, CEO and CTO of Yushu Technology, delivered a speech at the main forum of the 2026 World Robot Conference (WRC2026). He said that embodied intelligence is approaching an industry critical point similar to the "ChatGPT moment". It is expected that it will take 2 to 3 years at the fastest, and it may take 5 to 10 years to achieve a breakthrough under slower circumstances. Just the day before the speech, Yushu Technology had just landed on the Science and Technology Innovation Board of the Shanghai Stock Exchange, and the closing price on the first day of listing increased 460% from the issue price. Wang Xingxing said that the criterion for judging whether embodied intelligence has entered the industrial explosion stage is not a single scene display, but that in 80% of unfamiliar environments, robots can complete about 80% of tasks through voice or text commands. Wang Xingxing pointed out that the biggest technical bottleneck of humanoid robots currently lies in the alignment problem between the AI ​​model and the real physical world. Robots can achieve a high success rate in fully trained fixed scenes, but when items and environments change slightly, the task completion rate will drop significantly. Especially in the "last few centimeters" or even the "last few millimeters" at the end of the operation, problems such as tactile feedback and movement errors still limit the robot's generalization capabilities. Focusing on this challenge, Yushu demonstrated the "Physical AI Robot Self-Evolution V1.0" technical route that is being pre-researched. The system plans to use large AI models to automatically retrieve research results, generate control codes, and form a continuous iterative closed loop through simulation training, real machine testing, AI and manual evaluation feedback, and other links. Yushu believes that basic model capabilities, multi-source data accumulation, real robot deployment scale and skill accumulation will become important factors in promoting the self-evolution of robots. In his speech, Wang Xingxing also reviewed Yushu's development process from quadruped robots to humanoid robots in recent years. The humanoid robot G1 launched by the company has become a representative product in the industry and participated in the Spring Festival Gala "Martial BOT" and other project displays. This year, Yushu also launched manned mechas, wheeled robots As2-W and other products, and continues to explore the application of robots in factories, homes and outdoor scenes. Wang Xingxing said that the true large-scale implementation of robots still needs to solve the problems of efficiency and generalization capabilities. At present, robots can complete some simple tasks, but they often need to be retrained when facing new tasks. In the future, as AI model capabilities improve and real-world data continues to accumulate, embodied intelligence is expected to accelerate into the industrialization stage. As humanoid robots move from experimental demonstrations to commercial applications, data, models, hardware and self-evolution capabilities are becoming the core of industrial competition. The technical route proposed by Yushu also reflects that the robot industry is developing from "making robots" to "making robots continue to learn and evolve." via AI News (author: AI Base)