The appearance of Deepseek-R1 is a Chinese AI model that seems to have shocked Openai’s latest products at a much lower price this week and has lost $ 1 trillion from US stocks. The event emphasized how close the race of American technology racing with China and how close the race was. However, not only competing via AI models such as Chatgpt and Deepseek-R1, but also these two technical superpower is a new award. Robots like humans, heads, torso, arms, (often) legs. Such humanoid robots are the most recently announced plans recently announced by Jensen Huang, the CEO of CHIPMAKER NVIDIA. Huang’s vision has made NVIDIA an ascending one of the world’s largest listed companies. ELON MUSK correctly predicted a vast market for space launch and electric vehicles. MUSK predicts that the long -term value of Tesla’s humanoid robot is “beyond the combination of all other Tesla,” and has a $ 25 trillion company. On the other hand, China’s industrial policy has poured the startup fire department into a humanoid robot.
With the progress of the generated artificial intelligence since 2022, the development of humanoid robots has been turbocharged, accelerating. According to what is reported, the 27 humanoid robot model debuted in August 2024 at the Beijing World Robot Conference. A few months ago, Huang announced a new basic model developed to control the humanoid robot (a basic model for a specific application). Humanoid robots are not yet executed in many real world environments, but are starting to operate in Mercedes -Benz and BMW Amazon warehouses and factories. Goldman Sachs Research expects a market equivalent to $ 38 billion by 2035, and CITIBANK has estimated 648 million humanoid robots by 2050.
The appearance of Deepseek-R1 is a Chinese AI model that seems to have shocked Openai’s latest products at a much lower price this week and has lost $ 1 trillion from US stocks. The event emphasized how close the race of American technology racing with China and how close the race was. However, not only competing via AI models such as Chatgpt and Deepseek-R1, but also these two technical superpower is a new award. Robots like humans, heads, torso, arms, (often) legs. Such humanoid robots are the most recently announced plans recently announced by Jensen Huang, the CEO of CHIPMAKER NVIDIA. Huang’s vision has made NVIDIA an ascending one of the world’s largest listed companies. ELON MUSK correctly predicted a vast market for space launch and electric vehicles. MUSK predicts that the long -term value of Tesla’s humanoid robot is “beyond the combination of all other Tesla,” and has a $ 25 trillion company. On the other hand, China’s industrial policy has poured the startup fire department into a humanoid robot.
With the progress of the generated artificial intelligence since 2022, the development of humanoid robots has been turbocharged, accelerating. According to what is reported, the 27 humanoid robot model debuted in August 2024 at the Beijing World Robot Conference. A few months ago, Huang announced a new basic model developed to control the humanoid robot (a basic model for a specific application). Humanoid robots are not yet executed in many real world environments, but are starting to operate in Mercedes -Benz and BMW Amazon warehouses and factories. Goldman Sachs Research expects a market equivalent to $ 38 billion by 2035, and CITIBANK has estimated 648 million humanoid robots by 2050.
However, robots are machines that can automatically execute complex tasks, and robots can take many shapes, such as factories, automatic driving cars, or military drone robots. Humanoid robots are two reasons for two reasons and promise a spiral to strengthen together. One is a large potential market for robots that can use human tools and functions in the human environment. Second, when artificial intelligence opposes the restrictions on data available, humanoid robots provide routes to convert the AI itself. It is no wonder that competitors in the world, which are similar to the world, are looking for every edge in human -shaped contests in this large prize.
Give a humanoid to the robot Forms, or at least aspects, have a significant possibility that these robots will act useful in the human world that spent thousands of years, and the tools and environment for humans shaped like us. Making,. A world full of stairs, tables, drivers, medical equipment, etc. Robots are very useful if a robot moves around the Victorian House in London and uses my tools to clean and cook. The same is true for factories, hospitals, elderly nursing homes, and battlefields. In addition, a robot, which is formed around our world, can accompany us to cooperate with the team’s human beings when we work.
Human forms also stimulate new ways to make useful robots. For example, walking like a straight foot can save energy compared to a typical robot walking with a bent foot. Human hands are wonderful. Evolution has given humans a great ability. For this reason, many AIs are benchmarked on human abilities, from language tuillering to computer vision and board game GO. Early generation robots have drawn out inspiration from simpler creatures such as insects and dogs. Today’s AI is progressing in fields such as language and planning.
There is no mass development of humanoid robots immediately. The AI’s recognition advances have been many years to reach a large scale on Amazon Alexas. Robotaxis has been developed for many years in the United States and Chinese cities, but in 2024, Google’s Waymo reached the transformation, reaching a 1 million to 4 million passenger trip. However, since the humanoid format is very convenient, as technology develops, there are markets that can be large for them. And it’s just half of the spiral promised by humanoid robots. The second is to promote the new leap of AI itself.
AI spent several decades in DOLDRUMS for decades to make a big leap in computer vision in 2012. This required a large new dataset of visual image to train AI. In 2022, the AI using ChatGptt had a big leap, and AI again trained the data again. Both progress also needed to enhance the power and calculation method of the computer, but the data was important. Today, we have run out of data that can be easily accessed by the world for training models. Data simulation is useful, but you need real data. Humanoid robots help provide a huge amount of new data linked to reality to learn how to act in the physical world.
The robot is physically embarrassed. Today’s AI is a robot vision language action model that can shoot texts such as text (such as ChatGpt), as well as the data on the physical environment of the robot (via a camera or microphone) and the internal sensor (for example, the robot joint position. It can be provided. The amount of data obtained from a single human infant is incredible. In recent studies, we used video and audio from head -mounted cameras for one infant. With that data alone, AI was able to learn many words and concepts. Adding the effects of infants and robots to get more useful data on how the world works. And by combining all of these types of data, the stubing of the toes can be reduced, for example, because it is a collision with reality.
Humanoid robots have additional benefits. They can learn from the most surprising intellectual behavior of the earth: humans. For this reason, NVIDIA’s Project GR00T aims to develop an AI model that Humanoid Robots observe human demonstrations and help human telepelers to practice actions. How humans execute tasks can give humanoid robots a “correct answer” to train. Such help is essential for Moravec’s paradox. Tasks (for example, chess), which seem to be difficult for humans, are often simple, but tasks (for example, putting a shirt on a hanger) are very difficult for AI. The correct answer to actions, such as putting a shirt on a hanger, seems to be simple for humans, but robots can learn a lot about how humans succeed in such actions. When a robot shares our human form, it is easier to teach robots. We will explain how to do what you do in your body and provide feedback on your robots for your efforts to learn.
This kind of learning is a large -scale, a large scale, and the humanoid robot praises a global technology competition for half of both of the enhanced spirals. 。 More robots in dialogue with humans can lead to more data that AIS can be learned, connect to better AI, which enhances robots and undertakes more jobs, so that AIS can learn more data. After the spiral.
China and the United States have a variety of strengths in global competitions to win the dominance of the humanoid robot. Each is associated with different half of this spiral, including mass production and AI learning.
Chinese big edges are large -scale manufacturing. China is the only superpower of the world’s manufacturing industry, and the production exceeds the next nine large -scale manufacturers. China dominates several major robot markets, including drones, which supplied 70 % of the world’s drone users in 2023. China seems to be the same as the Robotaxis USA. Also, as Chinese car executives stated as “wheel robots,” huge Chinese EV companies make a large amount of investment in driver support software. In general, robot engineering was identified as one of the Chinese President Xi Jinping’s President Xi Jinping, and in 2022, China set up more than half of the world’s industrial robots.
Humanoid robots have been identified by China as an important field of technical competition in 2023, when the Ministry of Industrial Information Technology announced the “Guidelines for Innovation and Development of Humanoid Robot”. This year, China is trying to establish a humanoid robot world -class innovative ecosystem, and by 2027, I would like to integrate humanoid robots into manufacturing supply chains, use it on a scale, and expand the use of humanoid robots throughout society. I think.
Big edges in the United States are the most advanced technology in companies, emerging companies, universities, and universities such as NVIDIA. The humanoid software “brain” accounts for about 80 % of its value, and the NVIDIA chip seems to be more preferred than the state -of -the -art AI in China. U.S. companies have built Chatgpt, and US researchers are still led China to the second position and publish the AI research quoted to the top. Allies are also important, for example, the United Kingdom has published the third research of AI, but Switzerland, Germany and Japan manufacture many world -global industrial robots.
However, China and the United States are equally consistent today, which can change. For example, the United States was a pioneer in the manufacturing industry and industrial robots, but ultimately delayed. China’s new Deepseek-R1 suggests that American edges in AI can hardly attack. So what can Washington do?
Leading competition In humanoids, robots may be the most results in the next 10 years, but most people do not even know this race, except for unusual numbers such as NVIDIA’s Huang. The first step is to better recognize this important race between the US policy community. Secondly, understanding the issues can help navigate the policy device to navigate the trade -offs needed to win this race. This is because inevitable political pressure on unemployment, privacy, and political control affects these developed industries. Policy lever such as industrial strategies and tariffs (according to political preferences) can help protect these important new industries through these difficult initial stages in accordance with China’s efforts.
The United States must not lose its advantage in the most advanced research, but many of the things needed to compete in this race need to return to the outstandingness of a large -scale manufacturing industry. And this itself reveals an interesting change in the relationship with AI. Increasingly, AI leaves the virtual regions of Tiktok, X, or Instagram feed, and enter the strict reality of the physical world around us.
