When AI "refines" human form, what can it do for humans?

On July 4th, at the 2024 World Artificial Intelligence Conference, the "Eighteen King Kong Crowd Creation Future - Humanoid Robot Pioneer Array" was filmed by Wang Xiang
On August 21-25, at the 2024 World Robot Conference held in Beijing, 27 domestically produced humanoid robots attracted widespread attention. They "played the yangqin, splashed ink and waved their hair, competed in ball games, debated, and danced" Subject Three "at the conference, earning enough traffic.
On July 4th, at the 2024 World Artificial Intelligence Conference (WAIC) held in Shanghai, the "Eighteen Golden Kings" array consisting of 15 humanoid robots made neat movements to the audience, which was particularly eye-catching. Domestic embodied robot products such as the "Qinglong" of the National and Local Collaborative Humanoid Robot Innovation Center, the Expedition A2 of Zhiyuan Robot, the Unitree H1 of Yushu Technology, the GR-1 of Fu Liye, and the bipedal XR4 of Dati have all made their debut.

On July 6th, at the 2024 World Artificial Intelligence Conference, people watched the universal humanoid robot Unitree H1 from Yushu Technology perform the "Subject Three" dance (Photo by Wang Xiang)
"Embodied intelligence is the next wave of AI development," said Dai Qionghai, an academician of the CAE Member, dean of the School of Information Technology of Tsinghua University, director of the National Research Center for Information Science and Technology in Beijing, and chairman of the Chinese Association of Artificial Intelligence.
The so-called "embodied intelligence" refers to adding a "body" to the AI "mind" that can understand, reason, and interact with the physical world. The "body" can be a mobile phone, car, machine tool, robotic arm, etc. The complete and realistic "humanoid" body, due to its unique affinity for humans and the potential to achieve complex functions, is regarded as an ideal carrier of AI, even the "ultimate form". In the past two years, driven by the rapid development of large models, the humanoid robot industry has ushered in a trend and become one of the hottest tracks in the global technology industry.
In August 2024, the American embodied intelligence company Figure AI released the second-generation humanoid robot Figure 02. Previously, Figure 01 had already started working at the BMW factory. Domestically, Uber's humanoid robot Walker S has successively entered automobile factories such as BYD, NIO, Dongfeng Liuqi, and Jike to "work". Zhiyuan Robot's humanoid robot Expedition A1 has also completed tasks such as chassis assembly and appearance quality inspection in automobile factories
AI big models will empower humanoid robots to move towards active interaction and become disruptive products after computers, smartphones, and new energy vehicles, "said Xu Xiaolan, a member of the 14th National Committee of the Chinese People's Political Consultative Conference, former deputy minister of the Ministry of Industry and Information Technology, and chairman of the China Electronics Society, at the press conference of the 2024 World Robot Conference on August 13th.
More and more 'like' people
In the National and Local Co built Humanoid Robot Innovation Center (hereinafter referred to as the "National Center") located in Zhangjiang Science City, Pudong, Shanghai, the humanoid robot "Qinglong", which is 185 centimeters tall, weighs 80 kilograms, and has up to 43 active degrees of freedom, gently extends its arms and uses five steel fingers to pick up bread.

The world's first full-size universal humanoid robot open source public version machine - "Qinglong" robot is on display (Photo by Wan Honglei)
The 'Qinglong' robot platform has five sensory abilities: it can see, smell, hear, touch, and posture sensors. Its hands are very flexible, and it can accurately grasp small objects up to two centimeters long. It can use tools to pick out sesame seeds in Xiaomi, "said Xu Bin, General Manager of Guodi Center, to the Oriental Outlook Weekly." The core technology of the 'Qinglong' robot platform has been accumulated for more than ten years in the early stage. From the establishment of Guodi Center to the recent appearance of the whole machine at the World Artificial Intelligence Conference, our team has made breakthroughs from scratch in just six months
Carrying water and grasping sounds simple, but it is complex to do, involving multiple technologies from perception, decision-making to control, "said Jiang Lei, chief scientist of the National Center for Geosciences." The 'Qinglong' has 7 degrees of freedom on one arm, 6 degrees of freedom on the end finger, and tactile sensors on the fingertips, so it can complete autonomous judgment of grasping
For humanoid robots, the brain corresponds to intelligent systems, the cerebellum corresponds to control systems, and the facial features correspond to the robot's perception and execution through joints, "said Zhou Meijie, Vice President of Shanghai Robot Industry Technology Research Institute.
The reason why 'Qinglong' is intelligent lies in the deep integration of robots and large models. The 'brain' responsible for human-computer interaction and decision-making (Zhuque embodied brain model) and the 'cerebellum' responsible for stable walking (Xuanwu cerebellum model) jointly form the technical foundation of humanoid robots. Jiang Lei introduced that Zhuque embodied brain is like an intelligent command and dispatch center, with a multi-modal large model as the core, capable of receiving various inputs such as text information, image information, etc., and outputting accurate voice interaction and wise task decision-making information; The Xuanwu cerebellar model focuses on task execution, and its trajectory planning module adopts an end-to-end imitation learning driven approach.
With the support of big data, big models, and big computing power, humanoid robots are becoming increasingly intelligent and gradually possessing the ability to independently complete tasks, "said Xu Xiaolan.
The robotics industry has been developing for many years, and in the past, robots generally only focused on one "skill point", "writing a piece of code to do one thing". For example, some industrial robots, mainly focused on developing motion control technology compared to "muscles"; The usability of a robotic vacuum cleaner depends on its ability to avoid obstacles and clean the house, with a focus on developing navigation and sensing technology. To have the comprehensive ability of "integrating knowledge and action" like humans, humanoid robots need to achieve breakthroughs in many cutting-edge technologies and comprehensive applications such as chips, sensors, software, artificial intelligence, communication, machinery, materials, etc.
Before this round of development of large-scale models, there was almost no technological breakthrough at the level of robot 'brain', and the models launched by various companies could only achieve simple tasks. In the past two years, related technologies have reached a turning point of qualitative change. "Xu Bin said," In the future, with the rapid 'growth' of humanoid robots, their minds may soon break away from immaturity and move towards the track of 'adults'
Deep learning enables artificial intelligence to perceive the world, while big models give AI the ability to autonomously generate decisions, "said Zhang Chi, Chairman and founding partner of Beijing Xinding Rongsheng Capital Management Co., Ltd." In the near future, humanoid robots will move towards universality and commercialization.
Multi city layout
Humanoid robots are becoming a new highland of technological competition and a new track for future industries, with huge market space and development opportunities, "Xu Xiaolan said at the press conference of the 2024 World Robot Conference.
As early as 2021, the outline of the 14th Five Year Plan placed the robotics industry in an important position. The "14th Five Year Plan" for the development of the robotics industry, issued by the Ministry of Industry and Information Technology, the National Development and Reform Commission, and 15 other departments, proposes that by 2025, China aims to become a global hub for innovation in robotics technology, a gathering place for high-end manufacturing, and a new highland for integrated applications.
In 2023, with the accelerated iteration of AI big model technology, "humanoid robots" will frequently appear as a buzzword in national and relevant departmental documents. In October of that year, the Guiding Opinions on Humanoid Robot Innovation and Development issued by the Ministry of Industry and Information Technology proposed that by 2025, China's humanoid robot innovation system will be initially established; By 2027, the innovation capability of humanoid robot technology will significantly improve, and its comprehensive strength will reach the world's advanced level.
Guided by policies, various provinces and cities are accelerating the layout of the humanoid robot industry.
In April 2023, Hangzhou proposed to strive to achieve an industrial output value of over 50 billion yuan for the entire intelligent robot industry chain by 2026.
In May 2023, Shanghai issued the "Three Year Action Plan for Promoting High Quality Development of Manufacturing Industry in Shanghai (2023-2025)", which specifically mentioned accelerating the innovative development of humanoid robots. At the same time, Shanghai released the 2024 "Invest in Shanghai" policy package at the Global Investment Promotion Conference, with the most eye-catching being the billion dollar parent fund. The parent fund is established under the promotion of the Shanghai State owned Assets Supervision and Administration Commission, focusing on the three leading industries of integrated circuits, biomedicine, and artificial intelligence, with a focus on investing in areas such as integrated circuits, innovative drugs, and intelligent robots.
At the same time, Shenzhen issued the "Action Plan for Accelerating the High Quality Development and High Level Application of Artificial Intelligence in Shenzhen (2023-2024)", which clearly proposes the development of humanoid robots.
In June 2023, Beijing issued the "Beijing Robot Industry Innovation and Development Action Plan (2023-2025)", which provides policy guidance for the humanoid robot industry and proposes to accelerate the layout of humanoid robots.
A new role has emerged in the strategic landscape of developing humanoid robots in multiple regions - the innovation center for humanoid robots. The innovation center essentially bears the responsibility of mobilizing local resources and promoting the development of the local humanoid robot industry, "said Xu Bin.
The Guodi Center, unveiled in Shanghai in May 2023, is the first national level public platform in the field of humanoid robots. At that time, in order to select the location for the national center, we inspected several cities, "Xu Bin recalled." Many of the technological foundations of humanoid robots are common to the industrial robot and new energy vehicle industry chains, and Shanghai has a strong industrial foundation in these fields. Due to comprehensive considerations such as industrial foundation and policy environment, settling in Shanghai is the preferred choice
One of the core tasks of the Guodi Center is to accelerate the landing of universal robot scenarios. "On the whiteboard across from Xu Bin's desk, his thoughts on the organizational structure of the Innovation Center are densely handwritten. In 2024, we plan to build a training ground in Shanghai that can accommodate 100 humanoid robots. We will first choose relatively simple and easy to implement scenarios to land, promote technological innovation in the humanoid robot industry, and drive the landing of universal scenarios
The so-called 'training ground' refers to the site for data production and data scaling. A real-time calibration system, control system, data acquisition system, and data feedback system will be built here to generate synthetic data through digital twins, creating real and generative data for the humanoid robot industry.
As a physical industry with a wide range of technological fields and complex production processes, humanoid robots involve thousands or even tens of thousands of segmented industries with components. The development of humanoid robots is of great significance for directly creating output value, driving employment, and leading the development of the industrial chain. This is also one of the reasons why many regions are actively building innovation centers for humanoid robots and vigorously promoting the development of the humanoid robot industry, "said Xu Bin.
At present, cities such as Beijing, Shanghai, Ningbo, Hefei, and Chengdu have all established innovation centers for humanoid robots, taking the lead in the field of humanoid robots.
Industrialization is imminent
According to data from Qichacha, currently, robot related enterprises in China are mainly concentrated in Shenzhen, Guangzhou, Shanghai, and Suzhou, with 49400, 44100, 36100, and 35200 related enterprises in these four cities respectively. The top suppliers and supply chain enterprises of the most critical core components of humanoid robots are mainly distributed in the Yangtze River Delta and Pearl River Delta.
The development of the industry cannot do without capital investment. With humanoid robots taking off in the face of the wind, VC (venture capital) and PE (private equity investment) are rushing into this science fiction track one after another.
On August 18, 2024, Peng Zhihui, co-founder of Zhiyuan Robotics, known as the "genius youth", released five new commercial humanoid robots in one go, namely the "Expedition" and "Lingxi" series. At the same time, Peng Zhihui announced that Zhiyuan Robotics has launched a series of open source projects, including millions of real machine datasets and millions of simulation datasets required for training embodied intelligent robots.
We hope to open source data and provide components to enable more people to quickly master humanoid robot technology using Zhiyuan's components, "said Jiang Qingsong, Partner and Vice President of Marketing Services at Zhiyuan Robotics.
One year ago in August 2023, Zhiyuan Robotics released its first generation universal embodied intelligent robot - Expedition A1, and defined 2024 as the "first year of commercial use" for Zhiyuan Robotics. The company predicts a shipment volume of 300 units in 2024, including 200 bipedal robots and 100 wheeled robots, mainly used for delivering service and industrial scenarios.
As a star enterprise in the industry, Zhiyuan Robotics has raised funds rapidly. At present, the company has completed multiple rounds of financing, including angel round, A round, A1 round, A1+round, A2 round, and up to 6 rounds of A-series financing.
In February 2024, Yushu Technology announced the completion of a B++round of financing, with investors including Meituan, Jinshi Investment, and Source Code Capital, as well as old shareholders Shenzhen Venture Capital, China NetEase Investment, Rongyi, Dunhong, and Mida Junshi, raising nearly 1 billion yuan.
By elongating the timeline, it can be seen that investment institutions are crowded behind start-up companies such as Dati Robotics, Fu Liye Intelligent, and Le Sen Robotics. According to statistics, the total amount of related financing in the first half of 2024 exceeded 7 billion yuan.
In terms of market prospects, according to international investment bank Goldman Sachs, the market size of humanoid robots is expected to reach $154 billion (approximately RMB 1103.7 billion) by 2035.
In addition, according to the Human Shape Robot Industry Research Report released at the first China Human Shape Robot Industry Conference in April 2024, it is predicted that the market size of Chinese humanoid robots will reach about 2.7 billion yuan in 2024, 10.4 billion yuan in 2026, and 75 billion yuan in 2029, accounting for 32.7% of the world's total at that time, ranking first in the world, and is expected to reach 300 billion yuan by 2035. By achieving key breakthroughs in deep learning of the "smart brain", precise control of the "agile cerebellum", and innovative design of the "robust limbs", the humanoid robot industry will accelerate its transition from "prototype technology" to commercial applications.
"The humanoid robot is the carrier of the deep integration of artificial intelligence technology and manufacturing scene, and China's humanoid robot industry is on the eve of a blowout." Xu Xiaolan made this judgment.
Open source new ecosystem
At the 2024 World Artificial Intelligence Conference, Guodi Center officially released "Qinglong" as the world's first full-size universal humanoid robot open-source public version machine.
Every year, our team updates a version of the public version of the robot. With our design references, everyone can innovate in many aspects of humanoid robots with targeted ideas. "Xu Bin introduced that as the world's first full-size universal humanoid robot," Qinglong "integrates 43 active degrees of freedom, achieving full-size design of the head, hands, arms, legs, waist, and ankles. Drawing on the experience and practices of the scientific and technological communities in technology sharing and open source innovation, after evaluation, we have made the design drawings, software code, and dataset of 'Qinglong' open source, aiming to attract developers, researchers, and entrepreneurs from around the world to participate in the research and application of humanoid robots, accelerate the popularization and industrialization process of humanoid robot technology
"We have a common idea to build the base technology of Chinese humanoid robots." Jiang Lei said that open source can not only bring together global wisdom, accelerate technological innovation, but also win more voice for Chinese science and technology on the international stage.
For a long time, open source projects in the global technology field have been mostly led by foreign countries. We urgently need to break this situation in the field of humanoid robots and establish our own full-scale open source community. "Jiang Lei said that through open source, we can open up related technologies to the whole society, lower the entry barrier, and attract more resources and developers to enter this industry, which is conducive to promoting the development of the entire industry. In 3 to 5 years, humanoid robots can basically be used on simple production lines such as logistics sorting. In 5 to 10 years, they can basically be used in high demand and open environments such as homes and services
In response to the common problem of "lack of high-quality training data for large models" in the current artificial intelligence industry, Guodi Center will provide standardized data collection tools and rapid data secondary evaluation based on digital twin systems. In addition, Guodi Center will also build open and open-source annotation tools, encourage ecological alliance units to accelerate dataset construction together, and empower applications through open-source data scenarios and industry-specific data.
At present, the OpenLoong open-source community website built by Guodi Center has already opened sourced the hardware structure and parameters of the robot, and will soon open source the embodied intelligent software package. Overall, the scale of the humanoid robot race is still relatively small, and the establishment of an open source ecosystem is expected to attract more talents and enterprises to join and jointly break through the common key technologies of humanoid robots.
In addition to the "Qinglong" series of robot projects, the National Center for Robotics is currently accelerating the construction of training centers to promote the application and landing of humanoid robots in more real-life scenarios. According to the plan, the first phase of the center can accommodate 100 humanoid robots for training, and it is planned to accommodate 1000 humanoid robots for simultaneous training by 2027.
As one of the "Eighteen Golden Kings" at the entrance of the 2024 World Artificial Intelligence Conference, the Fu Liye Universal Humanoid Robot GR-1 was appointed as the "Lobby Manager" at the Shanghai Pudong Branch of China Construction Bank in July.
At the Bank Lobby Manager Scene Training Base of China Construction Bank, GR-1 is responsible for guiding customers, diverting call numbers, and providing simple anti fraud training to customers Currently, GR-1 focuses more on language interaction with customers. We hope to develop more physical functions in the next step, allowing robots to complete tasks such as welcome reception, patrol, text binding, and even deposit and withdrawal, "said Gu Jie, CEO of Fu Liye Intelligent
The public is eagerly anticipating the use of humanoid robots to provide home services such as taking care of the elderly and children. However, it remains to be seen how far away humanoid robots are from achieving large-scale commercial use and even "flying into ordinary people's homes" at an affordable price?
The future is promising, investment is hot, but the hot track requires a calm approach, "said Xu Bin.
The current humanoid robots have limited functions and are expensive. Before the emergence of ChatGPT, people also felt that AI was not intelligent enough, which is somewhat similar to the current state of humanoid robots, "said Wang Xingxing, founder and CEO of Yushu Technology.
"Rich scenes are one of our greatest advantages. Scenes are expected to drive breakthroughs in technology. In the end, human science is still in the stage of weak artificial intelligence, and humanoid robots are still far from 'learning and understanding like people'. The most important thing at present is to go out of the laboratory and really go to the market." Ding Han, an academician of the CAS Member, said.