Waiting for the moment when "GPT-3" arrives, will humanoid robots become the ultimate form of "embodied intelligence"?

Currently, China's humanoid robot industry is still in its early stages of development. However, with the continuous iteration of core technologies such as autonomous driving and the expansion of application scenarios, tech ecosystem giants are making strategic investments. The industrialization process of humanoid robots is expected to accelerate. At the same time, China has successively promulgated policies to promote the development of robot technology in recent years. For example, in December 2021, 15 departments including the Ministry of Industry and Information Technology issued the '14th Five-Year Plan' for the development of the robot industry. The plan proposes to focus on promoting the research and application of key products such as industrial robots, service robots, and special robots. By 2025, China is expected to become a global source of robot technology innovation, a high-end manufacturing hub, and a new frontier for integrated applications. Therefore, the future market development prospects for humanoid robots in China are broad and are expected to penetrate various niche vertical scenarios.

In recent years, with the successive entry of giants such as Tesla, the concept of 'humanoid robots' has become once again popular. As one of the earlier companies in China to venture into humanoid robots, UBTECH has been dedicated to the research, development, manufacturing, and sales of artificial intelligence and humanoid robots. In addition, Xiaomi also released its first full-size humanoid robot, CyberOne, last year. Although humanoid robots have not yet entered mass production and many core technologies need to be breakthroughed, it is foreseeable that humanoid robots will have broad application scenarios, higher commercial value, and opportunities for layout in all directions of the industrial chain.

"Embodied intelligence" refers to a type of machine intelligence that possesses autonomous decision-making and action capabilities. It can perceive and understand the environment like humans, and complete tasks through autonomous learning and adaptive behavior. Compared to traditional artificial intelligence systems, embodied intelligence emphasizes that the agent needs to interact with the real world and generate an understanding and transformation capability of the objective world through its own learning.

When discussing the significance of embodied intelligence for the development of AI, embodied intelligence is an important way to enhance the cognitive capabilities of current 'weak AI'. Specifically, by constructing embodied intelligent agents, the performance of reinforcement learning algorithms can be greatly improved. Therefore, embodied intelligence is very likely to become another major wave following AI.

The humanoid robot industry chain can be divided into three parts: upstream components, midstream body manufacturing and system integration, and downstream practical applications. The upstream components of humanoid robots mainly include material reducers, motors, structural parts, semiconductor chips, sensors, and materials. Among them, reducers, servo systems, and control systems are key links for technological breakthroughs, serving as the core guarantee for solving robot precision movements. They have high technical barriers and are also the main direction for localization efforts. Considering that some components of humanoid robots have high technical barriers, giants like Tesla can leverage their technological and industrial chain advantages in areas such as autonomous driving and automotive parts, integrating external suppliers of core components such as reducers and motors. Therefore, manufacturers with competitive advantages in related niche fields may be the first to benefit.

At the same time, machine vision, as one of the core technologies for humanoid robots, provides them with capabilities such as object recognition and detection, posture recognition and tracking, face recognition and emotion recognition, environmental perception, and navigation. In recent years, machine vision has continuously achieved domestic technological breakthroughs, and the localization rate is steadily increasing. Therefore, manufacturers with machine vision algorithms and mature solutions will also benefit.

The initial positioning of humanoid robots is to replace some labor in application scenarios, engaging in 'dangerous, repetitive, boring' work. Such jobs often require repetitive tasks, involve hazardous environments, or require a lot of physical labor, which may pose certain risks or be unsuitable for humans.

The application of humanoid robots can effectively solve these problems. They can undertake tasks in dangerous environments, reducing the risks to human workers. In addition, they possess precise repetitive motion capabilities and can perform the same tasks on production lines or other work scenarios, alleviating the physical burden on human workers. For those monotonous and repetitive tasks, humanoid robots can execute accurately, improving efficiency and freeing up human resources for more creative and high-value-added work.

In addition, humanoid robots can be widely applied in entertainment, education, healthcare, service industries, and more. For instance, humanoid robots can act as hotel receptionists, providing services such as check-in procedures, information inquiries, and navigation. They can communicate with guests in real-time and offer personalized recommendations and suggestions. At tourist attractions, humanoid robots can also serve as guides, providing explanations and information introductions to visitors.

The indicators for evaluating the maturity of the humanoid robot industry can include the following aspects:

The technical maturity of humanoid robots is an important indicator for measuring their development level. This includes the maturity of key technologies such as robot environmental perception, motion control, and human-machine interaction. Mature technologies can achieve a high level of humanoid simulation, intelligent interaction, and precise action execution, meeting the needs of various complex application scenarios.

Secondly, the commercialization level of humanoid robots is also an important indicator for evaluating their maturity. It reflects the maturity of various aspects such as technical feasibility, market demand, business model, user acceptance, and ecosystem construction. For example, commercialization means that humanoid robot technology has been verified in practical applications. In real commercial environments, humanoid robots need to face various challenges and complex situations, such as interaction with real users, environmental adaptability, reliability, and safety. Only when humanoid robots are successfully applied in commercial scenarios and recognized by users can the technical feasibility be truly proven.

There are many difficulties in the process from industrial applications to personal service applications. In industry, humanoid robots usually operate in relatively controlled environments, such as production lines or warehouses, and the 'command-execution' working mode can somewhat avoid the technical difficulty of real-time perception in the real world. However, in personal service applications, humanoid robots need to adapt to complex and uncertain environments, such as homes, public places, or offices. This involves enhancing the capabilities of humanoid robots such as navigation, obstacle avoidance, object recognition, and spatial perception.

In addition, personal service-side applications require humanoid robots to be able to interact naturally and effectively with humans. This includes the development of technologies such as voice interaction, gesture recognition, emotion recognition, and face recognition. Humanoid robots need to possess high levels of comprehension and communication skills, capable of accurately identifying users' intentions and responding accordingly. Therefore, only when difficulties in core technologies, costs, security, and other aspects are resolved can the smooth transition and widespread application of humanoid robots from industrial to personal service-side applications be achieved.