Boao Asia Forum Discusses Humanoid Robots: "ChatGPT Moment" Gradually Approaching, Staying Human-Centered Is the Key to Long-Term Success

Securities Times Reporter Wu Shaolong

On March 25, at the Boao Asia Forum 2026 Annual Conference, humanoid robots “Embodied Tian Gong 3.0” and “Q5” took turns giving intelligent self-introductions. Baidu Smart Cloud Digital Human ViviDora participated in the discussion in holographic form. The high-tech opening kicked off the sub-forum “Advancement and Breakthrough of Humanoid Robots.”

During the forum, attendees engaged in in-depth discussions on breakthroughs in humanoid robot technology, industry implementation, regulation and governance, and human-machine relationships. They outlined the path from “laboratory” experiments to “everyday life” and addressed core challenges in development.

Farewell to the Show-Off Era

The trend of technological innovation and industrial transformation toward practicality has become the core consensus among forum participants.

Xiong Youjun, CEO of Beijing Humanoid Robot Innovation Center, said that from a technical perspective, robot hardware technology is gradually converging. With the development of large models, motion control capabilities are rapidly improving. The robot’s “brain” shows significant co-evolution of the cerebrum and cerebellum, with diverse technological paradigms including layered control of language behavior large models, multimodal large models with perception technologies, and integration of world models with multifunctional large models. The trend of cerebrum and cerebellum co-evolution is prominent.

More importantly, the emergence of AI agents like OpenClaw has endowed humanoid robots with self-evolution capabilities. Wang Xiaogang, co-founder, executive director of SenseTime, and chairman of Dash Robotics, noted that in recent years, the development of related models has been largely limited by data. Previously, data collection relied on manual control, which was highly inefficient. AI agents like OpenClaw support autonomous invocation, self-reflection, and memory, enabling robots to shift from single-machine embodied intelligence to group collaboration, continuously evolving in environment exploration and task execution. This technology is expected to push robots from digital employees toward physical implementation, becoming a major industry trend.

Alongside technological breakthroughs, humanoid robots are gradually moving from show-off performances to practical scenarios. Xiong Youjun pointed out that the application of humanoid robots is accelerating into broader industrial fields, covering automotive, home appliance manufacturing, as well as handling, sorting, and logistics, transitioning from specialized to general-purpose.

“ChatGPT Moment”

Gradual Arrival

Behind the booming industry, for some time, the industry has been eager to predict when humanoid robots will experience a disruptive leap similar to the “ChatGPT moment.” Although forum guests had differing predictions on the timing of the “ChatGPT moment” for humanoid robots, they agreed on a “gradual breakthrough.”

“Currently, humanoid robots still face significant challenges in stability, durability, and dexterity,” said Shen Dou, Executive Vice President of Baidu Group and President of Baidu Smart Cloud. “While large models have begun to provide a unified technical foundation, the ‘brain’ and ‘cerebellum’ of robots, as well as overall technical solutions, have not yet formed a unified system. The industry is far from reaching the stage of a ‘ChatGPT moment.’”

Referring to key breakthrough nodes like ChatGPT, top scholars from China and the US generally agree that the “big probability” of such a breakthrough will occur in about five years, according to Chen Jianyue, founder of Xingdong Era. He said that even if the “ChatGPT moment” does not arrive immediately, it does not prevent humanoid robots from being applied in industrial and other scenarios first, because these environments are highly verticalized and standardized. Robot manufacturers can leverage existing models, invest some data and computing power, and quickly implement applications in key roles.

Xiong Youjun said that the “ChatGPT moment” for humanoid robots will not happen overnight but will be coupled with technological maturity, commercial value, and social demand. Currently, in some structured and simple repetitive scenarios, applications based on existing large models have already been implemented. Data shows that last year, nearly 20,000 humanoid robots were shipped in China, with significant growth expected this year and next, possibly faster than anticipated.

Centering on Human Needs

While technological advances accelerate, building governance systems and defining human-machine boundaries are urgent. Global cooperation and a human-centered approach have become the core bottom line for industry development.

Helen Clark, former Prime Minister of New Zealand, said that as AI and humanoid robot technologies develop rapidly, questions about responsibility for risks and bridging digital divides need collective answers. “We are the first generation to coexist with AI, and perhaps the last to set boundaries between AI and humans.” She believes this transformation is enormous, and China remains a global leader in humanoid robots.

Sam Daws, senior advisor at the Oxford Martin School’s AI Governance Initiative and director of multilateral AI, said that robot regulation must consider social, human, and ethical dimensions, and avoid cutting off human connections. Currently, establishing open, mutually recognized global technical standards is key. The EU, Singapore, and others have already introduced ethics-based robot regulation laws.

“AI capabilities will continue to grow, so we must clarify human-machine boundaries and take the initiative,” Chen Jianyue emphasized. First, goal-setting authority: robots’ tasks should be decided by humans, with robots only providing suggestions; they should not autonomously set goals. Incidents like errors in AI applications such as “Lobster” are often due to over-delegation. Second, rule definition authority: human beings must set constraints and boundaries for robot behavior. Third, ultimate responsibility: humans, not machines, must bear responsibility. “If robots malfunction, humans must take responsibility—whether individuals or organizations.”

Laying a Strong Foundation for Long-Term Development

Regarding the most concerned issue for consumers—home application—guests provided rational predictions and addressed the controversy over “humanoid robot traps,” pointing the way forward for industry development.

“Embodied intelligence entering home use should be the final step in development,” Shen Dou said. “Home scenarios impose upper limits on robot capabilities. Compared to hazardous industrial environments, which are more standardized, home environments are the greatest challenge.” He judged that it is unlikely for robots to systematically enter homes within two years, and whether it can be achieved in ten years remains uncertain.

Responding to doubts that overly anthropomorphized humanoid robots are a “trap,” Shao Hao, chief scientist at vivo Robotics Lab, said that this mainly stems from the perception that current hardware and software capabilities are limited, and that robots like vacuum cleaners or lawnmowers outperform humanoids in single tasks. However, scene dictates form. Long-term, within ten years, humanoid robots will have applications far beyond specialized robots.

At the end of the forum, the humanoid robot “Embodied Tian Gong 3.0” asked Helen Clark, “As robots, we want to serve humans. What can we do to earn the trust of ordinary people?” Clark replied, “Be stable, adaptable to environments, responsible, and maintain boundaries without overstepping.” This simple exchange encapsulates the essence of industry development: only by adhering to a human-centered core, allowing technology to innovate within norms and improve through implementation, can humanoid robots truly move from concept to reality and become a vital force for social progress.

(Editors: Zhang Xiaobo)