CNBC AI News, August 9 – At the 2025 World Robot Conference, Wang Xingxing, CEO of Unitree Robotics, addressed the current state and future trajectory of the robotics industry.
Wang emphasized a potential imbalance in the industry’s focus, stating that there’s too much emphasis on data and not enough on the underlying models.
The core bottleneck, according to Wang, lies in model architecture: “The biggest problem right now is the model itself. Current robotic model architectures are insufficiently robust and lack unification. Even with substantial, high-quality data for training, it remains largely unusable.”
He acknowledged the impressive growth the robotics industry has witnessed this year but maintained that the most critical challenge, both presently and in the future, lies in developing robust “embodied intelligence robot foundation models,” or large AI models specifically designed for robots operating in the real world.
“However, model development remains relatively slow, and architectures lack standardization,” Wang stated. “We haven’t seen a significant breakthrough yet. To draw a parallel with the advancement of Large Language Models (LLMs), we’re approximately 1-3 years before the ‘ChatGPT moment’ for robotics.” In other words, the robotic industry is in its nascent stage for unified models.
Wang also expressed Unitree’s ambition to develop general-purpose humanoid robots capable of performing diverse tasks. The aim is to create robots versatile enough for applications ranging from factory work and entertainment to home assistance, instead of being limited to a single, specialized role.
He mentioned that a significant portion of Unitree’s workforce, including himself, is dedicated to training robots to perform tasks across various settings.
Looking ahead, Wang highlighted key priorities for intelligent robotics technology over the next 2-5 years: unified, end-to-end intelligent robot foundation models, lower-cost hardware with extended lifespans, ultra-high-volume manufacturing, and access to low-cost, large-scale computational power.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/6761.html
