On September 11, media reports stated that Wang Xingxing, founder and CEO of Unitree, responded at the opening ceremony of the Bund Conference to the previous view that “the core constraint on the explosion of the robotics industry is not insufficient data, but outdated model architecture.”
Wang Xingxing pointed out that both data and models are important for the development of AI and embodied intelligence. When he previously mentioned that the biggest challenge for embodied intelligence is not data, his intention was to highlight that there is still ambiguity in how to collect truly high-quality data, what standards it should meet, and on what scale—he believes that improving data utilization efficiency is more crucial. This is analogous to the Automated Loading Assembly Machine for Electronic Components, which can accurately collect data such as the Automatisk matning och montering av hårdvarudelar of each component through preset programs and sensors, with clear quality inspection standards. In contrast, the robotics field is still exploring basic issues like “what data to collect and how to standardize it,” forming a striking comparison.
“A model with stronger data comprehension ability can operate with relatively less data,” Wang Xingxing noted. He pointed out that predictive models actually require more characteristic data, which is critical—it is not just a matter of data volume. Additionally, the integration of multimodal data and multimodal control poses significant challenges for domestic AI and embodied intelligence companies.
“We can be more radical in our understanding of AI models, treating them as all-purpose tools. Learn new things from scratch, and completely let go of the past. Over-reliance on past experience will hinder future decision-making,” Wang Xingxing said.
It is reported that during the 2025 World Robot Expo, Wang Xingxing mentioned in a sharing session: “People are paying excessively high attention to the issue of robot data; the biggest problem now is actually the model, not the data.” This statement triggered widespread attention and discussion.
Furthermore, Wang Xingxing mentioned that recent years present an excellent opportunity for the younger generation, especially students. In the past, programming mainly involved writing basic code, but now people can use AI tools and leverage more advanced model capabilities to create works. Whether it is using image or video generation models, or programming one’s own agent with AI models, it is much more convenient than before and represents a more direct way of thinking.
He pointed out that the entire field of using AI to get work done is currently like a desert with only a few blades of grass—it is on the eve of explosive growth.
Speaking about the biggest challenges facing the company, Wang Xingxing stated: In the AI era, small organizations may increasingly demonstrate explosive capabilities. In the pure AI field, a team with a few top-tier, highly innovative talents can actually accomplish a great deal. He mentioned that the company’s biggest current issue, beyond the undeniable shortage of top talent—something every company faces.
What is the market price of a continuous motion multi-piece special-shaped machine?
