Zhejiang Robot Industry Association
Zhejiang University Robotics Institute
Golden Conference & Exhibition Group
Shanghai Top Sequoia Exhibition Co., Ltd.
Shanghai Berrick Exhibition Co., Ltd.
From quadruped robot dogs clearing garbage on Mount Tai, to the humanoid robot Fuxi performing traditional Yangko dance on the Spring Festival Gala stage, and diverse humanoid robots racing side-by-side along the track of the Beijing Yizhuang Half Marathon — embodied artificial intelligence, or Embodied AI (EAI), has quietly stepped into public view.
Artificial intelligence brings unprecedented development opportunities, alongside unprecedented risks and challenges. During the 20th group study session of the Political Bureau of the CPC Central Committee, Xi Jinping emphasized: “We must uphold self-reliance and self-strengthening, prioritize application-oriented development, and advance the healthy and orderly growth of artificial intelligence.” During his recent inspection tour of Shanghai, he pointed out: “China boasts abundant data resources, a complete industrial system and massive market space, delivering broad prospects for AI development. We shall ramp up policy support and talent cultivation to develop more safe, reliable and high-quality AI products.”
How will humanoid robots reshape the operation of human society in the future? How can we foster the sound, orderly development of humanoid robots and other AI technologies? This issue invites three distinguished experts to discuss the technological evolution, application landscape and ethical implications of embodied AI.
Liu Huaping, Professor, Department of Computer Science and Technology, Tsinghua University
Zhang Linghan, Member of the UN High-Level Advisory Body on AI; Professor, Institute of Data Law, China University of Political Science and Law
Li Zhenjing, Director & Research Fellow, Institute of Economic System and Management, National Development and Reform Commission
Journalist: Against the backdrop of embodied AI, why have humanoid robots become a core industrial development track?
Liu Huaping: Humanoid robots are not a new concept. Scholars launched relevant research as early as the 1970s, and China saw a domestic research boom in the early 2000s. Notably, the latest wave of humanoid robot research is driven by embodied AI, with the two fields reinforcing and advancing each other mutually. Therefore, current research on humanoid robots places greater focus on how much genuine intelligence these machines integrate.
The biggest advantage of humanoid robots lies in their anthropomorphic form: they can operate with human-equivalent efficiency in most workspaces and drastically boost productivity. Undoubtedly, humanoid robots integrate structural engineering, materials science, motion control, perception and many other disciplines, serving as an all-in-one carrier for numerous cutting-edge technologies. Advancing humanoid robots not only caters to demand within the robotics industry, but also drives breakthroughs across multiple core technologies.
Furthermore, R&D of humanoid robots breaks bottlenecks constraining robotics development. It also facilitates research on bionic mechanisms of coordinated morphology and behavior, laying critical foundations for uncovering the origins of embodied intelligence.
Zhang Linghan: At its core, a humanoid robot combines generative artificial intelligence with a physical body. This fusion enhances the robot’s capacity to perceive and comprehend the world, helping it overcome core bottlenecks in generalized locomotion, build motion models for complex scenarios, and achieve all-round optimization of its "brain", "cerebellum" and mechanical limbs.
Equipped with both an intelligent "mind" and an anthropomorphic "body", humanoid robots have evolved from sci-fi fantasy into tangible consumer technology. In recent years, they have played pivotal roles in industrial manufacturing, while demonstrating immense potential across service, medical and education sectors.
Li Zhenjing: As a core carrier of embodied AI, humanoid robots will exert far-reaching impacts on economy and society through technological integration, market demand expansion, industrial upgrading and job creation. The humanoid form delivers superior flexibility and environmental adaptability, enabling seamless integration with existing infrastructure and workplace layouts.
From an economic perspective, humanoid robots will drive coordinated growth across the full industrial chain. Socially, though they may displace certain traditional jobs in the short run, in the long term they will free humans from repetitive manual labor, accelerate industrial upgrading, improve socioeconomic operational efficiency, generate new occupations and ignite innovation vitality.
China is accelerating independent innovation breakthroughs in humanoid robotics, with expanding manufacturing and market scale advantages. Backed by a comprehensive industrial system, a suite of targeted policy initiatives has guided the domestic robotics industry from quantity to quality growth, yielding breakthroughs in core technology R&D and "large model + vertical scenario" deployment.
On the R&D front, technological milestones including dexterous hands with multi-dimensional tactile perception, and general intelligent agents capable of complex motion planning and fluid human-machine interaction have emerged. On the application front, embodied AI has been deployed in diverse real-world scenarios: data-driven scientific research assistance, intelligent manufacturing, commercial services and special hazardous operations. As of early 2025, China has maintained its position as the world’s largest industrial robot market for 11 consecutive years, accounting for over 50% of global industrial robot installations.
Journalist: Powered by large language models, will humanoid robots truly enter every household anytime soon?
Liu Huaping: Large language models have greatly accelerated the advancement of humanoid robots. Many research prototypes and demo systems simply combine large models with robotic hardware, while other studies leverage large models to optimize robot structural design, motion control and environmental perception. Overall, the synergy between large language models and humanoid robots is still in exploratory stages.
Humanoid robots represent typical embodied AI systems, while large language models are disembodied AI systems lacking physical form. Their integration will undoubtedly fuel technological progress, yet the exact pathways of mutual empowerment remain unclear. Even so, humanoid robots act as a traction technology driving cross-field innovation. Once individual component technologies mature, mass household adoption will naturally follow.
Robotics research constantly ventures into uncharted territory. Robots that gain wide household popularity are usually those whose core functions reach staged maturity during technological iteration. Robotic vacuum cleaners have achieved partial household penetration, yet their functionality remains limited with massive room for improvement. Therefore, household humanoid robots stand as a long-term vision that propels continuous technological advancement.
Zhang Linghan: I believe mass household adoption of humanoid robots will take time, as daily deployment hinges on three critical factors: AI ethical compliance, technical hurdles and cost barriers.
First, ethical regulatory challenges are prominent. Household humanoid robots need to recognize family members and learn daily living habits to deliver personalized, intelligent services — a process reliant on processing massive volumes of personal biometric and lifestyle data, which raises severe privacy risks.
Second, humanoid robots represent deep integration of generative AI and physical machinery, demanding far more sophisticated technical capabilities than conventional software-only AI. Widespread commercialization for domestic use hinges on resolving fundamental technical bottlenecks and achieving technological popularization.
Third, cost is the decisive factor for household penetration. Affordability determines public acceptance of any new technology; excessive pricing will block mass-market entry. With sustained technological iteration, I am confident humanoid robots will eventually become affordable household companions in the near future.
Li Zhenjing: China’s humanoid robotics industry has delivered remarkable results with promising prospects. Smart factories have launched coordinated training programs for multiple humanoid robots performing diverse tasks, advancing the shift of new industrialization toward fully flexible intelligent manufacturing.
In the future, humanoid robots will evolve into two major categories: advanced labor collaborators and mass consumer home appliances. They will serve as auxiliary labor units, supporting human work and decision-making as humanity’s new partners. Once challenges around battery endurance and production costs are resolved, alongside improved public recognition and trust, humanoid robots hold great potential as daily consumer goods, catering to household chores, educational companionship, elderly care and other diversified domestic demands.
Journalist: What key technological breakthroughs will drive humanoid robot development forward?
Liu Huaping: Dramatic progress has been achieved in humanoid robot motion control in recent years, reigniting industry optimism for commercialization. Moving forward, motion control systems must fully leverage advances in hardware and machine learning to boost coordinated locomotion across diverse working environments, shifting from mere demonstration performance to efficient multi-task operation.
On top of dexterous movement, R&D must strengthen robotic perception and autonomous learning capabilities to enable adaptation to variable environments and shifting task requirements. Humanoid robots feature abundant degrees of freedom, flexible bipedal walking and precise manipulation, and should not merely act as passive hardware executing outputs from large model inference. In high-challenge industrial and emergency scenarios, their unique mechanical structure can significantly elevate perception and learning performance — an advantage unavailable to robots of other forms. Only tight integration of physical body control and perceptual learning can unlock the full potential of humanoid robots, representing the core breakthrough direction for the sector.
Zhang Linghan: Humanoid robots are a strategic fusion of artificial intelligence, high-end manufacturing, new materials and other disciplines; progress across all these fields will collectively drive robotics innovation.
Among general-purpose core technologies, end-to-end models have unlocked transformative opportunities for humanoid robots. Abandoning the traditional serialized robotic architecture, end-to-end systems employ a unified model to process raw sensory input and directly output locomotion commands, eliminating siloed operation between separate functional modules. Riding the boom of large foundation models, the development of a unified universal large model that covers the full pipeline from sensor data input to motor execution is highly anticipated.
Beyond shared foundational technologies, different application scenarios impose distinct technical priorities: industrial robots prioritize durability, motion precision and task completion accuracy; household robots prioritize safety, ease of operation and low-cost hardware; customer service robots center on voice recognition, emotional interaction and decision-support logic.
Li Zhenjing: Humanoid robots have made notable strides in motion control and emotional interaction technologies, yet we must acknowledge persistent challenges amid optimistic long-term outlooks. Several core components of domestic humanoid robots still rely on imports, particularly servo control systems and high-precision sensors — areas requiring urgent indigenous breakthroughs. Cost-performance balance remains another critical bottleneck preventing large-scale commercial rollout and mass adoption.
Journalist: As humanoid robots integrate deeper into human society, a series of ethical and social challenges will inevitably emerge. How should we address these risks?
Liu Huaping: Every stage of robotics development brings corresponding ethical and social dilemmas, just as the internet and smartphones triggered new regulatory and ethical debates upon mass adoption. When existing moral norms and legal frameworks lag behind technological evolution, society updates rules and standards to adapt to new realities. Similarly, ethical and social challenges posed by humanoid robots can be resolved through continuous social progress. We should encourage broad public participation in cross-stakeholder dialogue, integrate diverse perspectives to fully assess societal impacts, and collectively guide healthy development of humanoid robotics.
Zhang Linghan: The development of humanoid robots must uphold a people-centered philosophy. Throughout R&D and deployment, human beings must retain primary agency. We need to build an accident liability framework with flexible regulatory elasticity, and strictly enforce fairness and non-discrimination principles to ensure all populations can equally access, develop and benefit from this new technology.
We must strike a balanced approach between industrial advancement and risk prevention. As a critical frontier in global technological competition, humanoid robotics is vital to boosting national competitiveness; governance should prioritize development while refining ethical review mechanisms to standardize technology R&D. This ensures humanoid AI evolves for benevolent purposes, guiding innovation toward progress benefiting human civilization.
Three key measures will mitigate risks brought by humanoid robot applications:
Enhance transparency and explainability of artificial intelligence, building auditable, monitorable, traceable and trustworthy AI systems.
Establish a comprehensive risk assessment and response framework, iterating through real-world deployment to improve robotic safety, reliability and controllability.
Develop a shared social ethical governance mechanism, engaging the public in formulating ethical norms and standards for humanoid robotics.
Li Zhenjing: Rapid advancement of humanoid robots fosters new business models, creates new jobs, drastically lifts labor productivity and improves quality of life, yet we must guard against potential societal hazards.
Many privately deployed large models suffer inadequate server cybersecurity protections. Likewise, humanoid robots collect massive volumes of user interaction data during daily operation, creating urgent risks of data leakage and misuse that demand resolution.
As humanoid robots grow more intelligent, debates over liability and autonomous agency grow prominent: when a robot’s autonomous decisions or movements cause property damage or personal injury, should liability rest with developers, operators or the robot itself? This requires clear legal definitions, with timely and dynamically updated relevant legislation recommended.
We must firmly reject the narrative of "human redundancy". While widespread humanoid robot deployment will reshape employment structures, it will simultaneously generate new occupations including robot maintenance, fleet management and AI supervision — opening emerging industrial tracks and creating new growth drivers and competitive strengths.
In the short term, humanoid robots will relieve humans from repetitive manual labor, freeing time and space for creative exploration, lifting productivity and fostering fairer competitive landscapes. Long-term, humanoid robotics will profoundly reshape global economic structures, production relations and distribution systems.
Only by effectively mitigating adverse technological impacts and bridging the global digital divide between developed and developing nations can humanoid robots become reliable assistants and intelligent advisors for humanity, ultimately realizing the ideal vision of harmonious human-robot coexistence.
Reporter: Fang Quyun