Dialogue with Wang Xingxing of Unitree Technology, the post-90s talent who created a humanoid robot of 99,000 yuan

Late LatePost

2024-05-15 23:07Posted on the official account of Beijing LatePost

"I've been thinking about a problem for 24 hours, and I don't believe it can't be solved."

Text丨He Qianming

Editor丨Cheng Manqi

In August 2023, Unitree Technology released the humanoid robot H1 made in half a year, without a very delicate appearance, the hands are two round bumps like Doraemon, and they appeared in a black dress, they can't backflip, but they walk very steadily, and people will not fall down if they kick it hard, combined with the price of $90,000, H1 has attracted a lot of attention.

Andrej Karpathy, then a research scientist at OpenAI and a former AI director at Tesla, said on social media that he wanted to buy one or two, or some; It is also recommended not to kick the robot anymore.

Wang Xingxing, founder and CEO of Unitree, apparently didn't listen. On May 13 this year, Unitree released a new humanoid robot G1, and in the demonstration video released at the same time, Wang Xingxing not only kicked it, but also put on boxing gloves to launch a frontal attack. In his opinion, this is the most intuitive way to test the movement ability and stability of a humanoid robot, which many companies dare not do.

The picture above is the humanoid robot H1 released by Unitree in August 2023, and the picture below is the G1 released in May this year. Image courtesy of Unitree.

Compared to the first generation, the new robot, which debuted eight months later, has significantly improved capabilities: its body and legs can rotate nearly 360°, it can "slowly" complete carp straightening, and its hands have changed from round bumps to dexterous hands with three fingers, and can dance sticks, open cokes, and smash walnuts.

This time, Unitree lowered the entry price of humanoid robots to 99,000 yuan, which is more than 80% cheaper, and has once again become a hot topic in the field of humanoid robots.

In the past year, riding the wave of large language models, a group of robotics companies have released humanoid robots. But few companies can sell humanoid robots as commodities.

Unitree is an exception. That "rudimentary" humanoid robot, H1, was sold to universities, research institutes, and AI companies around the world. OpenAI also bought it, a person familiar with the matter said.

Wang Xingxing, born in 1990, is not a scholarly entrepreneur with a bright resume. Among his competitors are many talented teenagers who were born in Qingbei, Harbin Institute of Technology or have studied in prestigious overseas schools.

Wang Xingxing graduated from Zhejiang Sci-Tech University with a bachelor's degree, although the total score was enough when he was admitted to graduate school, because his English was too poor, he was not listed in Zhejiang University and was transferred to Shanghai University.

"In the three years of high school, all the exams combined, and I only passed English three times." He said.

This does not stop some investors from thinking that he is a genius. From 2013 to 2015, when he was in graduate school, he didn't have much resources or budget, so he designed the hardware and control algorithms from scratch, and made the robot dog XDog with industrial motors, and won the second prize in the Shanghai Robot Design Competition—at this time, there were still 3 years before the MIT open-source robot dog algorithm.

Xingxing Wang's XDog was developed during his master's degree. The screenshot of the video is from Wang Xingxing's B station account.

After XDog attracted attention, Wang Xingxing received an angel investment of 2 million yuan, and founded Unitree Technology in 2016, serving as CEO and CTO. In the next 8 years, he took the company to design most of the parts used in the robot dog from scratch, such as motors, 3D lidars, etc., and sold the robot dog to all over the world, brought it to the stage of the Spring Festival Gala and the Asian Games, and received more than one billion yuan in financing, becoming the head company in the quadruped robot industry.

Wang Xingxing said that when Unitree made its first humanoid robot in 2023, they only invested three full-time employees, and the rest were people from the quadruped robot team. Because the development of robot dogs over the past years has provided them with a lot of hardware and software that can be reused.

He explains the results with enough focus and hands-on experience cultivated since childhood. Only in his 30s this year, Wang Xingxing would say that he has "more than 20 years of experience in mechanical design".

"A lot of the time it's just inspiration." He said that as long as you are focused enough, you will naturally come up with solutions. "You've been thinking about a problem 24 hours a day, and I don't believe it can't be solved."

Wang Xingxing still remembers the first time he painted in kindergarten, the butterfly he painted surprised his parents, and when he was in elementary school, he watched documentaries on cutting-edge aerospace, biochemistry and other scientific and technological topics introduced in China, and used his pocket money to make electric toys, he said that he liked the starting point of robots, and it is also the basis for the development of low-cost and high-performance robots.

The following is the dialogue between "Later" and Wang Xingxing:

"To develop the first version of the humanoid robot, we only used 3 full-time employees"

Later: You once said that you were not optimistic about humanoid robots, why are you involved now?

Wang Xingxing: Humanoid robots are so popular now, one is that Musk is doing it, and the other is AI.

Musk is going to do it in 2021, and someone asks us if we can do it. To be honest, I didn't feel much at the time. Humanoid robots have been popular for many years, and before Musk did it, this direction had reached a low point. I haven't heard of anyone buying (humanoid robots) either, and the circle is full of pessimism.

What made me decide to do it was because I became more and more convinced of AI. In 2010, when AI was still in the doldrums, I was a big fan of neural networks, and I played with them a little bit. Later, I went to the event and said that AI was promising, but I didn't believe it enough at the time. A lot of people still don't believe it enough.

In the past two years, the technology has progressed really fast, and some people may not have noticed that ChatGPT has had some attempts to control robots with AI before, and the results are not bad.

"Later": You set up a project to make a humanoid robot in February last year, and released the first prototype product in August, and it only took six months to make it?

Wang Xingxing: Actually, there are only three of us who do this full-time, and there are less than 10 people involved. We invest relatively little in humanoid robots. Everything is ready-made, and the hardware and software algorithms accumulated by the company can be used directly.

"Later": Your experience in making robot dogs can be directly applied to humanoid robots?

Wang Xingxing: In fact, there is not much difficulty, it may be that the torque requirements of the joint motor are a little higher. Four legs become two legs, and the load on each leg will be a little greater. There are many control algorithms that can be used directly from the quadruped robot.

Later: So your first humanoid robot was actually a dog standing up directly.

Wang Xingxing: Pretty much. Humans also evolved from tetraped mammals.

"Late": Four-legged directly changed to two-legged, won't it be easy to wrestle?

Wang Xingxing: Stability is definitely worse than four-legged, but it's definitely okay to walk. Of course, there are improvements in the future, and we still have upgrades and progress. (Note: In March of this year, Unitree released a video of the H1 backflip)

"Later": You guys did it quickly, but the first humanoid robot released last year had no shell, only a piece of clothing, and it looked very shabby.

Wang Xingxing: At that time, it was to catch up with the World Robot Conference.

I think our approach is very differentiating and advantageous. For example: first, our shell is very resistant to falling, and the shell of some companies will break if you fall casually, because it is a plastic shell, and our clothes have an impact-proof sponge lining, and you can change it at any time if you want to. And the cost is very low, how much is a piece of clothing, how much is a shell.

"Late": It also doesn't have a dexterous hand (a simulated manipulator).

Wang Xingxing: Because at that time, many domestic peers may have bought third-party dexterous hands, and we were making dexterous hands ourselves, and we didn't do it well at that time, so we didn't release it first.

"Later": One point of view is that the dog stands up directly and becomes a human, although it is done quickly, but the upper limit may be limited, because many of the ways of human movement are different from dogs, and people can pull their knees over and knock Erlang's legs, but dogs can't. Limiting the way you move limits the data you can access and the speed at which you can iterate on the software. What do you think of this assessment?

Wang Xingxing: This question itself is a big misunderstanding. The mechanical structure and control algorithm of almost all humanoid robot companies in China are transplanted and borrowed from quadruped robots.

In addition, our H1 took a photo of the stilted legs in August last year, and the latest G1 can do a lot of actions that people can't do.

"Later": You just released the new humanoid robot G1, it really looks much more flexible, is it also optimized for the way dogs stand up and become humans? Or is it a new technical route?

Wang Xingxing: There is no problem between dogs and people, if you have to say it, it is still the same sentence, all people have evolved from four-legged mammals.

"Later": The G1 high-performance version has a dexterous hand, is it your own development? Why only 3 fingers?

Wang Xingxing: We developed it ourselves, because 3 fingers are enough to do most things. In addition, Boston Dynamics' latest humanoid robot is also 3 fingers, which also shows from the side that 3 fingers are basically enough.

Later: When you were developing your first humanoid robot, you said that your company's principle was "don't do it if you don't make money", and the shabby-looking humanoid robot at the time cost $90,000. Now that the low-performance version of the G1, which is more agile in action, is only selling for 99,000 yuan, can it still make money?

Wang Xingxing: The H1 was the first humanoid robot we officially developed, and the idea of industrial design at that time was simple, so it was not particularly beautiful.

The business logic of our company is that there must be a reasonable business profit in the business behavior. Cost has always been the KPI we do for everything, and at its core, it's all about making money.

"Late": How to continue to reduce costs next?

Wang Xingxing: I still have a good feeling about cost reduction, because I have spent my own pocket money (doing various things) since I was a child. In 2010, I made a bipedal robot, and the cost of machinery and everything added up to 200 yuan, so don't compare the cost with us, we can continue to reduce it a lot.

"Late": This is your history of success. If you sum it up, what are the ways to reduce costs?

Wang Xingxing: We must have a sufficient understanding of the entire production chain, production process, including the entire design.

"Later": Isn't it by mass production?

Wang Xingxing: Design is definitely the main reason, followed by mass production, and relying on mass production alone is a detour.

Many people think that I can reduce the cost if I have the volume, but this is purely an illusion.

"Later": Robot dogs and people, in fact, there is no way to mass produce them at this stage, and there may not be so much demand.

Wang Xingxing: That's right. We currently have a lot of dogs, but the biggest point to really reduce costs is to rely on design, which is the most fundamental path. For example, the material is cheaper, the optimized structure can reduce some production costs, and the mold cost will also be reduced.

"Later": The logic of cost reduction for many companies is vertical integration, and everything is done by themselves.

Wang Xingxing: Industrial integration is what we have been doing. When I did XDog in 2015, the motor was bought, and the motor driver inside was designed by myself, so the cost was very low. After the establishment of the company, the first robot we made, the motor in it was made by ourselves, and the lithium battery pack in the robot in 2017 was also developed by myself, and there are some wireless communication and vision modules in the back, including the current 3D lidar, we will go further every year, and we can develop all self-developed. We also have our own factory.

Later: When did your humanoid robots go into mass production?

Wang Xingxing: It was shipped in the second half of last year.

"Late": Who are the customers?

Wang Xingxing: The customers are quite diverse, all kinds of companies, scientific research institutions, including AI companies, all kinds of things.

"Late": What do you think the humanoid robot will look like in the end?

Wang Xingxing: I don't think it's necessarily humanoid. Why do you want to be humanoid now? We're just pandering to the social consensus. Everyone now thinks that intelligent robots and general-purpose robots should look like people, which is the most socially influential, because people look like this.

In fact, it doesn't matter to me what form it is, the most important thing is intelligence, whether it is the image side, the execution side, it can do what people can do, and even what people can't do.

I've always felt that intelligence is not very scarce, it's very broad, because people have all of them, and some animals have a very high level of intelligence, for example, there are many animals that can use tools.

"Late": There are no new intelligent species after man.

Wang Xingxing: Because it is only a million years since the birth of human beings, intelligent human beings are actually very short.

Later: Do you think that as long as it takes long enough, there will be a second or third intelligent species?

Wang Xingxing: That's right. The odds are very high, and I think it's a coincidence. I have been reading for a long time and have always had the idea that many animals can't speak, a big reason is physical limitations, its brain is enough, but its vocal cords can't interact with each other, so its culture can't be established.

Late: How close are you to realizing the general-purpose robot you envisioned?

Wang Xingxing: At present, the world is still relatively slow. I think AI technology is sufficient, but the field of robotics is at least ten years behind AI research, because the hardware threshold is high and the talent base is small, and the whole pace will be much slower.

What people are doing right now is very simple, which is to bring AI technology to robots. However, I think that the current technical tipping point for the combination of AI and robots has not yet been reached, and the entire technical solution may not have been run through.

"Late": When will it come?

Wang Xingxing: I think it could be within three to five years.

Now that everyone has seen the signs of general-purpose AI, more people want to combine robots with AI, and more people are joining the industry.

"I'm in my 30s and have more than 20 years of experience in mechanical design"

"Later": Many of the founders of robot companies graduated from Tsinghua University and Harbin Institute of Technology, or came back from studying abroad, and your background is more "grassroots". Why are you still ahead of the curve?

Wang Xingxing: I have a serious bias in English, and I have only passed English three times in three years of high school. But I have been playing a lot with technology since I was a child, I started to draw at a young age, make all kinds of things by hand, play with machinery, and in 2009, when I was a freshman, I made a little humanoid robot.

Maybe my school doesn't look particularly good, but I do have a lot more experience than others. I will casually say that I have more than 20 years of experience in mechanical design.

Late: You're only in your 30s, and you have more than 20 years of experience in mechanical design?

Wang Xingxing: That's right. I have a lot of experience with this, and I like a lot of techniques, and I play all kinds of techniques, so the overall control will be better.

Late: Is that why you can compete with some founders whose backgrounds look greater?

Wang Xingxing: That's right. Although my grades are not very good overall, at least in the field I like, I have been basically cutting-edge since I was a child.

Before college, I probably had a little low self-esteem, thinking that I was much worse than those who were better at school, and I didn't take it seriously later.

"Later": What happened to make you no longer feel inferior?

Wang Xingxing: I originally lived in a small city, and I taught myself a lot of things, and I didn't have a teacher to teach me, so I could find a few books to read at most. I was a person who was suppressed by the education system since I was a child, and I didn't have any sense of accomplishment in my studies.

Later, I made a small robot in my freshman year, and I didn't think I was doing anything particularly good, so I just played around.

In fact, looking back now, it was already very good at that time. In addition, I made the XDog robot in 2015, and I can basically play directly with those masters and PhDs at MIT.

Later: Why do you think your XDog and MIT master's and Ph.D. are doing the same?

Wang Xingxing: At least I did better than them in that direction at that time. They invested a lot of people, spent a lot of time, money, and in the end they didn't do it perfectly.

At that time, my XDog was very cheap to make, the R&D cost was only 10,000 or 20,000, and I could do all kinds of sports, and it was very flexible.

Late: What metrics do you use to compare your stuff with theirs?

Wang Xingxing: It's not like that, their dogs are much bigger than mine, but my movement flexibility and stability are better.

"Late": Did you compare by looking at the videos they posted?

Wang Xingxing: That's right. Later, MIT engaged in an open-source solution, and many robots in China basically use this solution.

The open source solution uses the same motor as my XDog. In 2016, a Ph.D. student at MIT asked me for a master's thesis.

The global low-cost and high-performance legged robot is the direction I pioneered, which is why we sell very well overseas. Our products are also the first to be marketed, and the product cost performance is the highest.

"Later": You said that the R&D cost of XDog is only 1-20,000 yuan, how did you make it?

Wang Xingxing: I started doing it when I was in my first year of graduate school. My mentor originally wanted to make that kind of hydraulic robot, but the conditions were not enough to do it. I always wanted to make the robot better, so I remade the hardware of the robot, took the industrial-grade motor, made a small drive, and integrated it with other components to make XDog, and the overall mechanical structure and algorithm were all developed by myself.

Later: How did you come up with the idea of bringing these things together?

Wang Xingxing: I suddenly thought of this plan, and I thought it was a good plan and worth doing. I even thought that this program was worth commercializing, and I had to drop out of school and start a business or something. Originally, I was going to graduate in the second half of 2015, but I only made half of this thing, and I felt that half of it was worthless, so I took the initiative to postpone it, and I also participated in the competition in Shanghai and won the second prize.

Later: It came naturally to you to come up with this plan?

Wang Xingxing: A lot of times it's just inspiration, you're thinking about something, and suddenly you think that something is good, you can combine them with each other, or there is something innovative.

Later: Did you later summarize your R&D approach?

Wang Xingxing: I think the method is also very simple. You just think about it 24 hours a day, and now it's busy. Reading a book will be very simple, you think more about it when you go to bed, and suddenly you will be inspired.

Late: So do you have any methods that your colleagues and team can replicate?

Wang Xingxing: I think the main thing is that you are focused enough. You've been thinking about this 24 hours a day, and I don't believe it's going to be solved.

"Later": In the early stage of cutting-edge technology, investors prefer to invest in founders with a particularly good technical background, and the robot dogs and robots made by Yushu belong to this direction. Was it difficult for you to get your first investment?

Wang Xingxing: In 2015, XDog was already made, and I participated in the competition to meet a few investors who wanted to attract some investment and entrepreneurship, but they didn't invest in me.

I don't think that my personal ability, social resources and financial resources have arrived, so I didn't play myself to death when I came out to start my own business? So I went to DJI for three months. At that time, I thought about it more clearly, and it was a good thing to go to a large company for a long time.

In June 2016, my XDog became popular in the media and robotics circles at home and abroad, and some people were willing to buy my robots, and some people were willing to invest, so I came out and started my own business.

"Later": You didn't go out of your way to find investors, but investors found you?

Wang Xingxing: I also took the initiative to contact a few companies, but the effect was not very good, and finally I took the money of a person in Beijing, and he took the initiative to find me. The terms he gave were very broad, and he didn't sign an agreement when he paid for it.

"Late": How much?

Wang Xingxing: Two million yuan.

"Robot dogs are a niche market, and they have to make all the money they can make"

"Later": From your business to the present, most of the time the robot dog users are not very clear. How do you determine how each new generation should iterate?

Wang Xingxing: Some of it is customer feedback. The overall direction is actually that the hardware performance is better, the software performance is better, and then the cost can be lower, and the reliability is higher, and these directions are generally the same.

"Late": Low cost is a clear requirement, what does good hardware and software mean?

Wang Xingxing: For example, the stability of the hardware, the appearance is better, including the whole structure is better, in fact, many of them are not visible. A lot of people can't even tell the difference between good and bad things, and I think I'm pretty good at that.

"Late": How would you define good-looking? Some people think that the head of the early Uki robot dog is a bit like a cockroach.

Wang Xingxing: Good-looking is related to industrial design, and there is a bit of personal aesthetics or technological aesthetics in it. But sometimes for mass production, you have to make a compromise, and you can't make it like a car. But some cars are even stranger than that.

"Late": Good sports performance, in addition to stability, what other indicators are there?

Wang Xingxing: For example, athletic ability can adapt to various complex terrains, the speed of walking stairs can be faster and faster, the speed of running can be faster and faster, and the endurance is stronger; Higher protection level such as waterproof and dustproof; Then it is more user-friendly, and customers don't have too much trouble when using it. These are some of the more straightforward indicators.

"Later": Did you step on any holes in the research and development process, and how did you find these things?

Wang Xingxing: Actually, it's still a bit of a pit, for example, I don't pay enough attention to our products before 2019, and I feel a little confused. For example, the load capacity is not enough. Some products I think are a bit big, and it would be better if they were made a little smaller.

I now think that making products is a very rational thing, whether it is the appearance of various indicators, everything must be rationally analyzed, and to prove that in this market, your product is competitive with other products. If you're feeling a little emotional, you're definitely not thinking about it.

"Later": You have been selling robot dogs since 2016, and the users are not clear, can the pricing be very rational?

Wang Xingxing: The price is hundreds of thousands of yuan. But at that time, there was no bottom, and our pricing was a little more rational later.

"Late": What does it mean to have no bottom?

Wang Xingxing: It's just that the price has been adjusted several times. Because the product itself is not for consumers, we set the price at the beginning and grind it out with consumers.

The pricing of all products is very mysterious and very testing of human nature. In many cases, it is not that you are cheap, but in many cases, it is anti-human, because in most people's perception, the price is good.

Sometimes you sell cheaply and suffer a loss, which will affect your brand's style, which is why luxury goods can be sold at such high prices.

Later: So you want to make luxury goods?

Wang Xingxing: No, our idea is very simple, because this is a niche industry, and I want to make all the money that can be made in this industry.

Later: What do the people who buy your products mainly do?

Wang Xingxing: There are three main parts of the dog market, one is pure consumers, buy to play, or do a little simple educational programming. There is also a kind of education and scientific research, which is bought by various universities or companies to do some scientific research, publish papers, or do some teaching. There is also an industry application, such as power grid inspection, some power plants, chemical plant inspection.

"Later": Each of these scenes you mentioned has a company dedicated to doing it, why do you do it at the same time?

Wang Xingxing: I think all three are very important. A misconception of many people is that I only need to eat a part of this industry.

"Later": Is this logic only suitable for niche industries? For example, car brands generally anchor a positioning, and will not do low, mid and high-end as soon as they come up, and Apple only does high-end, maintaining high R&D investment with high gross profits.

Wang Xingxing: Everyone has always felt that Apple does not have too low-end products, but I found that Apple has a very good idea, it uses products from a year or two ago to sell as low-end products, so I think Apple also has this logic in it.

"Later": In recent years, there have been more and more companies making robot dogs. Will you adjust your company's strategy?

Wang Xingxing: Now we have set a very simple direction, our main opponent is still ourselves, that is, we do better every day, every month, every year than we used to do, and our peers basically can't catch up.

In addition, the past few years have also proved that the entry of peers does not have much impact on our company.

"Late": How to prove it?

Wang Xingxing: For example, shipments and social influence. When Xiaomi released the robot dog in 2020, I also told a lot of people that it was a good thing that they came in, and it would lock the whole track to death.

Many people who want to engage in this industry will think, Xiaomi has done it, so is the success rate of my business very low? When investors want to invest in this industry, they will also reflect, Xiaomi has done it, is my investment failure rate very high?

"Late": Is Xiaomi no pressure on you?

Wang Xingxing: I think it's a good thing. We can borrow resources and strength from each other.

"Late": What impact will the Spring Festival Gala in early 2021 have on the sales of your robot dogs?

Wang Xingxing: There must be some impact, but there is no more intuitive data to prove it. Later, our price was lower, the cheapest was only more than 9,000, and the sales volume increased.

Unitree Technology did rehearsals for the robot to board the Spring Festival Gala. Image courtesy of Unitree.

"When what you do is in line with the main theme, good resources will be tilted towards you"

"Later": In your previous interview, you said that you must find a point that suits you and has the ability to do it. It's not easy for most people to find something they like and are good at, how do you do it?

Wang Xingxing: That's what I know when I'm old.

"Late": How do you know?

Wang Xingxing: I think it should be a positive iterative process. For example, when I was in kindergarten, my parents were shocked when I drew the first picture, and they called the neighbors to see it.

I drew a butterfly, I didn't learn to draw, I only did one thing at that time, I just drew the details. Isn't painting according to the painting? I'll draw whatever he draws.

"Later": Do you remember everything about kindergarten?

Wang Xingxing: I basically know what I was doing 20 years ago. Later, when I was in elementary school, I would make things and draw random things. I didn't have to learn anything, so I could do the best in the whole school, and it was relatively easy to do it well.

At that time, the country introduced a lot of documentaries, and my biggest hobby was watching documentaries on TV after school, and I watched everything about aviation, aerospace, and biology before I was almost ten years old.

So I know very well what I'm good at. I don't know if my language skills are born or caused by thinking backwards. I don't speak English very well, and I don't speak very well.

"Later": You didn't get admitted to Zhejiang University because of English, did bad English have any impact on your business?

Wang Xingxing: I think there are advantages and disadvantages. Because I was poor at language or other things, I had to make something else better to generate self-worth and a sense of self-fulfillment. The disadvantage is that your English is poor, and you may have a lot more difficulty reading some literature and reading some things than others, which will definitely suffer.

Later: Did you think of any way to learn English well?

Wang Xingxing: I thought about it a lot. I was very good at politics in middle school because I found that politics is logical, you can do well just by following the logic, I don't need to memorize it.

When I was in graduate school, I made up for a year of English, and I used all the methods I could use, and finally my English reached the best level so far, and I passed level 4, but I didn't pass level 6.

"Late": Is it reconciled now?

Wang Xingxing: Pretty much. Anyway, I don't take English classes now, and it will be embarrassing to communicate with foreigners when I go abroad.

"Later": You didn't have much work experience before starting your business, and you only worked at DJI for three months. How do you learn to manage a company and sell things as products?

Wang Xingxing: I have been sensitive to human nature since I was a child, so I still know a little bit about people. In addition, I am from Ningbo, Zhejiang, and I still have some knowledge of doing business since I was a child.

"Later": What inspired you the most?

Wang Xingxing: I think it's still a very rational thing to do business, whether it's products, sales strategies or pricing. It would be nice to follow rational standards in all aspects.

Later: You've always said that you want to learn from DJI, what do you want to learn most about DJI?

Wang Xingxing: Actually, I have a lot to learn. DJI is just one of them, and I'm interested in some of the companies that are doing better. I'm thinking about one thing over and over again, why those companies can do so well, so big.

Late: What inspired you the most?

Wang Xingxing: For example, as Mr. Lei (Lei Jun) has always said, this is also something I thought of a long time ago. When what you do is in line with the main theme of the times, some good resources will be inclined to you.

"Late": What is the current trend?

Wang Xingxing: Humanoid robots.

Title image source: Unitree Technology

View original image 116K

• 

• 

• 

• 

• 

•