Unitree Robotics founder Wang Xingxing

TMTPOST -- A highly anticipated symposium on China's private sector was held on Monday in Beijing, with Chinese President Xi Jinping delivering a speech after hearing comments from prominent entrepreneurs.

Wang Xingxing, the founder of Hangzhou-based startup Unitree Robotics born in the 1990s, was present at the symposium. Wang was seated in the front row, alongside seasoned business figures such as the world’s leading battery maker CATL’s founder Zeng Yuqun, e-commerce giant Alibaba’s founder Jack Ma, telecom giant Huawai’s founder Ren Zhengfei, leading electric vehicle maker BYD’s founder Wang Chuanfu, leading smartphone and EV maker Xiaomi’s founder Lei Jun, and tech giant Tencent’s founder Pony Ma. Among them, the BYD founder, agricultural company New Hope’s founder Liu Yonghao, chipmaker Will Semiconductor’s founder Yu Renrong, Wang and Lei spoke as representatives at the symposium.

Unitree Robotics and its founder Wang's growth trajectory can be described as a "rocket-like leap." Wang rose from humble beginnings. He graduated with a bachelor's degree from Zhejiang Sci-Tech University. During his postgraduate entrance exams, he failed to gain admission to Zhejiang University due to poor English scores and was instead admitted to Shanghai University. In an interview with the media, Wang once remarked, "During my three years in high school, I only passed my English exams three times in total." 

Between 2013 and 2015, while pursuing his graduate studies, Wang, with limited resources and funding, independently designed hardware and control algorithms and used industrial motors to create the robotic dog XDog. This achievement won him second place in the Shanghai Robotics Design Competition, and after completing his graduate studies, he embarked on a journey as an entrepreneur in robotic dogs.

Unitree Robotics, founded in 2016, initially focused on developing quadruped robotic dogs and successfully marketed them globally, becoming one of the top players in terms of product shipments in this field. By 2023, Unitree Robotics had ventured into humanoid robots and quickly became the most talked-about company in the industry. In January 2025, Unitree Robotics' latest humanoid robot product made its debut on the stage of the CCTV Spring Festival Gala, garnering even wider attention.

In April 2024, Wang had an exclusive conversation with the AsianFin. Wang told the AsianFin that the fundamental reason behind the explosive growth of the humanoid robot industry is the emergence of large language models (LLMs). Previously, it would take one to two years to teach a humanoid robot how to walk, but now, with AI algorithms, this can be achieved in just one month.

Regarding the future development of humanoid robots, Wang was upbeat. He said that by the end of 2025, at least one company globally will have developed a relatively general-purpose robot foundational model. This foundational model will function like a complete set of building blocks, with LLMs being just one piece. Other components will include visual perception, tactile perception, decision-making, and interaction capabilities.

Looking further into the future, Wang told the AsianFin: "Within our lifetime, humanoid robots will be able to revolutionize every industry, from industrial and service sectors to agriculture and manufacturing. On an even grander scale, governments could deploy 100,000 humanoid robots to build an entire city. On a smaller scale, robots could shrink to the size of cells, transforming all aspects of the natural environment as we know it."

Below is the full text of the conversation between the AsianFin and Wang, with edits for clarity and brevity.

"The Inflection Point for Humanoid Robots Has Yet to Arrive"

AsianFin: Just a few days before our meeting, Boston Dynamics, a star company in the robotics field, announced that its hydraulic-powered humanoid robot would be retired, and future developments would focus on electric-powered products. What’s your take on this?

Wang: Boston Dynamics has been working on robots for many years and has also been pursuing commercialization for a long time.

As for hydraulic drives, I had already concluded around 2013 that this approach would not be commercially viable. The reason is simple: hydraulic systems rely on precision mechanical components, and whenever precision mechanics are involved, costs cannot be reduced. Moreover, all hydraulic systems are prone to oil leakage. That’s why, as you can see, even household cars rarely use hydraulic systems anymore—they’ve all been replaced by electric drives.

So, if this company intends to continue developing humanoid robots, adopting an electric drive solution is undoubtedly the correct path. However, the only surprising thing is that back in 2018, I assumed Boston Dynamics had already started developing an electric-driven version. But later, I realized they hadn't made much progress, and I eventually forgot about it.

AsianFin: Compared to hydraulics, is electric drive better suited for LLMs?

Wang: Compared to hydraulics, electric drive can be said to have all advantages and no disadvantages. As for whether electric drive is better suited for AI, it's hard to evaluate. However, electric drive is lower in cost, offers better motion flexibility, is safer, and is also lighter in weight.

AsianFin: After Boston Dynamics transitions to electric drive, combined with their previous training data, is it possible for them to iterate faster than their competitors in this new race?

Wang: It's hard for me to comment on this. However, we remain relatively confident because we've been working on quadruped robots for many years. A lot of the algorithms and components developed for them can be directly applied to humanoid robots.

Another point to note is that a significant portion of AI talents in the U.S. do not work at Boston Dynamics but at companies like Google, Nvidia, and OpenAI. Boston Dynamics' strengths may lie in their hardware capabilities and traditional humanoid robot control expertise.

AsianFin: So, does the emergence of LLMs represent a major turning point for humanoid robots?

Wang: I don't think we've reached that turning point yet.

It's merely the beginning of a direction. There seems to be a misconception that LLMs like ChatGPT can be directly applied to robots. But in reality, they can't be used in this way at the moment.

AsianFin: Why not?

Wang: Because it's not designed for robots. ChatGPT is based on text logic, and its dataset and training methods are entirely text-based. Its performance in environmental perception for robots is currently not desirable, not just for us but globally.

So, in the current humanoid robot industry, you could say it uses AI technology, but there's still a significant difference between this and LLMs.

AsianFin: But some companies have already publicly stated that they are using large models to recognize different types of plates and then instruct robots to identify and pick them up.

Wang: That’s actually hard to assess. After all, it’s just a video, and there hasn’t been any confirmation to the public. Moreover, there’s no data to prove whether the robot could still distinguish objects if the plate were replaced with an apple, a pear, or something else. Personally, I feel that Silicon Valley hasn’t achieved any significant technological breakthroughs in this area; their progress is still quite conventional and by the book.

AsianFin: So, does this mean that LLMs are not the turning point for humanoid robot development? Are they not as important as people think?

Wang: For robots per se, they’re insignificant. However, the underlying technological direction they represent is very significant. Currently, large models are primarily large language models, and no one has yet succeeded in creating a large model specifically for robots.

AsianFin: That brings us to the question—what exactly is driving this wave of humanoid robot startups in 2023?

Wang: The driver is actually quite simple: Tesla has started working on humanoid robots.

Elon Musk has achieved considerable success in the automotive and rocket industries, significantly expanding those sectors. Now that Musk has ventured into humanoid robots, governments and various organizations also want to get a head start instead of waiting for Musk to succeed and then playing catch-up.

At the same time, the launch of ChatGPT’s large model has opened up the imagination space for the entire AI field. You could say it has ignited everyone’s imagination and enthusiasm. The current buzz is just the beginning; it will grow even stronger in the future. With annual advancements in hardware and AI technology, the disruptive impact of this industry on the world will be immense.

"It’s quite simple, not as complicated as people imagine."

AsianFin: If today’s AI large models are just the beginning, what will be the future direction of the industry or the focus of everyone’s efforts?

Wang: There are many directions. The first step is to adapt AI for robots, enabling them to handle tasks like visual perception, understanding, execution, and planning.

Like everyone else, I’m very excited. Personally, I feel this industry will develop rapidly. Whether it’s robots, large models, or AI in general, I believe that by the end of 2025, at least one company globally will have developed a relatively general-purpose large model for robots.

Of course, our company hopes to be the one to achieve this, but realistically speaking, the probability is higher that it will happen in the U.S.

AsianFin: That brings up the issue of whether to go open-source or not.

Wang: Anyway, whatever we develop, it definitely won’t be open source.

AsianFin: Is there a unified model that applies to both large-scale robot models and robotic dogs?

Wang: Most robotic dogs are implemented through reinforcement learning, and the technology is actually quite mature.

As for large-scale robot models or world models for robots, they can be applied to all kinds of robots—not necessarily humanoid or robotic dogs. It’s a general-purpose concept. I’ve always believed that the form of a robot doesn’t have to be humanoid. Humanoid is just one of many forms, and I’ve never insisted that it must be humanoid.

AsianFin: But the mainstream opinion seems to favor humanoid robots, arguing that the basic framework of our entire civilized society is built around the human form.

Wang: That’s just what they like to say. I’ve never agreed with that.

You can completely construct a new physical world. For example, if you’re mining, why does it have to be humanoid? If you’re building houses, why does it have to be human-shaped? Of course, humanoid robots are an important part—or relatively important—but they’re not everything.

For instance, if you’re using a robot at home, people might prefer a humanoid form for things like acting out a storyline or accompanying you on a trip. But for tasks like building houses or carrying heavy loads, there’s no need for it to be humanoid. Moreover, people might feel uncomfortable seeing a humanoid robot doing labor-intensive tasks, as it could evoke a sense of slavery. It just doesn’t sit well with some people.

AsianFin: Would you feel sorry for them?

Wang: Current AI hasn’t reached that level yet; it can’t perceive such things. But if AI were to develop to the point where it could sense pain or negative emotions, well, that might be problematic. For now, though, there’s no need to feel sorry for them, because they’re still just inanimate objects with limited intelligence.

AsianFin: Here’s something I’m curious about: even though their intelligence is limited, why do they exhibit a human-like stumbling motion when you push them?

Wang: That’s because it’s trained through AI and reinforcement learning.

AsianFin : So it’s essentially mimicking human behavior?

Wang: Some things are not necessarily imitations; they are determined by natural laws. You could say that the logistics of these robots naturally limit their general form to something like this. If there were an alien who looked similar to a human, then their behavior and mannerisms would likely be quite similar to ours as well. 

Robots made by Unitree Robotics are dancing at the CCTV Spring Festival Gala

AsianFin : Nowadays, people tend to break down robots into components like the cerebrum, cerebellum, and body. What’s your perspective on this?

Wang: I’ve never been a fan of dividing the brain clearly into "cerebrum" and "cerebellum." One model is enough—why split it into two? I think it’s unnecessary.

Of course, within the model, there might be various modules, but overall, I prefer to treat it as a single model. Right now, everything we do, from walking to fine motor operations, is achieved entirely through AI in an end-to-end manner. From visual perception to leg movement, one model handles it all—there’s no intermediary mathematical formula or anything like that.

AsianFin : Is the hardware capable of keeping up?

Wang: For robots, it’s just a few joints—there’s nothing particularly challenging. It’s just sensors sending data to the model, and the model sending commands to the joints. That’s it.

AsianFin : Your version of humanoid robots seems much simpler than others'.

Wang: It really is simple—it’s not that complicated.

AsianFin : For example, others might think that dexterous hands are a particularly challenging area in fine motor operations because they require more precise recognition and finer motion control.

Wang: If you use traditional techniques, it’s indeed very difficult. But you can’t rely on traditional techniques. If your technology lacks innovation, then it’s meaningless. Of course, you can’t express this too bluntly either. It’s better not to go too far beyond the general public’s understanding; otherwise, I’d probably get criticized to no end.

AsianFin : What exactly do you mean by "non-traditional"?

Wang: It's the new AI, end-to-end. It means there's no need for manually writing too many software programming rules in the middle, nor do we need to rely on traditional image recognition methods.

AsianFin : How is this achieved?

Wang: By modifying the model. The underlying AI remains the same, but the entire model structure and algorithms are different. This might be hard to explain in detail because it could be difficult to understand. For example, you don't need traditional image labeling or image understanding at all. You can input images and videos into a model, and the output would be the robot's joint trajectories, which can then be trained directly.

Image labeling can still be done, like labeling an image of an apple. The purpose of labeling is solely for interacting with humans, helping the AI better understand human intentions. But for a robot, there's no difference between an apple and a pear.

AsianFin : Compared to the mainstream, your logic and industry perspective seem quite unique.

Wang: The mainstream approach still has many issues. For a startup, if your thinking is purely mainstream, it’s simply not enough. You must foresee the development direction of the next few years. Once you see it, you need to start planning ahead, and then you’re bound to succeed—or at least, you won’t lose. If you only focus on what everyone else is talking about, others can obviously do it better than you. How could you possibly stand out?

AsianFin : In your view, what will the next few years look like?

Wang: I can’t go into too much detail, but one thing is certain: the pace of progress in the industry will be incredibly fast.

AsianFin: How fast are we talking?

Wang: It’s almost beyond imagination. For instance, the current pace of AI being applied in factories globally is advancing at an extremely rapid rate. The technology is already close to being fully operational.

AsianFin: So far, no company has been able to fully utilize robots for work.

Wang: But the entire logic is almost fully developed. This doesn’t mean robots can do every type of work, but end-to-end robots capable of performing tasks are nearing maturity. A more generalized robot model is expected to be developed globally by the end of 2025.

AsianFin: That soon?

Wang: It might even happen faster. Some people have already seen the direction, and while it might sound a bit boastful, I also feel like I’ve seen the direction. With this direction in mind, if we invest some more time, manpower, and money, we can basically make it happen.

"Luck plays a significant role in all technological breakthroughs"

AsianFin: What exactly do you mean by robot models?

Wang: You can think of it as, first and foremost, having very strong mobility, suitable for most terrains, and perhaps even surpassing human capabilities in some aspects. For example, its ability to overcome obstacles, speed, and jumping capability might be slightly better than humans.

Another aspect is performing tasks in factories. Robots can handle many tasks without requiring manual programming. Leveraging the capabilities of large models, you just need to teach them a little, and they can learn on their own and get the job done.

AsianFin: Does it still require simulation training in a virtual environment?

Wang: Probably not. Once you’ve trained and verified it, simulation might not be necessary anymore. Of course, completing the hardware might take some time, but I think that’s merely a matter of time.

As for AI, there’s still some uncertainty. While I just mentioned that I’m personally optimistic that it could happen by the end of 2025, it’s also possible that it won’t, and it might take 3–5 years or even longer. It depends on the collective luck of humanity. Sometimes, it really comes down to luck, doesn’t it?

AsianFin: How should we understand this concept of luck?

Wang: Luck plays a big role in many technological breakthroughs. For example, if Einstein hadn’t existed, someone else would probably have discovered his theories. But it might have taken a few more years, or even decades. You could say that luck is a significant factor in all technological breakthroughs.

AsianFin : Another point is that breakthroughs in large models require not just algorithms and models but also data. Is data collection currently a major challenge?

Wang: There’s indeed a lot to be done, but there are methods to address it. It’s not as complicated as people think. Many problems aren’t as complex as they seem. As you know, in all current fields of technology, if you really look into it, there’s nothing overly complicated. Things are relatively straightforward and simple. Even with something like a lithography machine, it’s just a matter of a few components.

AsianFin : So, would you say your industry is divided into two camps—optimists and pessimists? For example, you seem to lean toward optimism, believing that the challenges are not insurmountable.

Wang: It definitely requires time and intellectual effort, but these are solvable and actionable issues. They are not like room-temperature superconductors or controlled nuclear fusion.

The biggest problem with room-temperature superconductors and controlled nuclear fusion is whether they are even physically possible. It’s possible that the universe simply doesn’t allow such things to exist, and no matter how much time and effort humanity invests, they may never be achievable.

Artificial intelligence robots, on the other hand, are a much more common concept. It’s not some lofty, unattainable goal—it’s essentially the intelligence of a group of people or animals. Intelligence is a very prevalent phenomenon. Some animals are incredibly smart; many can understand human speech, though they can’t speak themselves. For example, crows—some crows can even use tools directly.

So, intelligence as a concept doesn’t have many constraints or physical limitations. It’s something that can be replicated.

AsianFin: What is the biggest driving force behind your decision to pursue this endeavor?

Wang: Honestly, the biggest motivator for me personally is AI.

Years ago, an investor asked me if our company would ever work on humanoid robots. I told him we would never do it, not in a million years.

Traditional humanoid robots were far too complex. With conventional algorithms, you simply couldn’t manage such intricate machines. The training algorithms for traditional humanoid robots relied on clever human minds writing mathematical equations and solving them to determine the robot’s movement trajectories. But these equations had significant limitations. If the environment changed, they might no longer work, and you’d have to design entirely new equations from scratch.

This type of training approach can lead to a massive amount of code, and when a system becomes sufficiently complex, it’s impossible to maintain it solely through human effort. However, for AI, as long as the model is well-constructed and continuously fed with data and computational power, it can keep iterating through trial and error. By leveraging the reward mechanism in reinforcement learning algorithms, AI can automatically retain successful training outcomes and discard the unsuccessful ones, significantly improving training efficiency.

Currently, the progress in AI technology, including its capabilities, has far exceeded my personal expectations. As a result, you can see that our humanoid robot, which we’ve been working on for just over a year, has already achieved remarkable performance. The reason we’ve progressed so quickly is simple: the rapid advancement of AI technology.

The advantage of AI is that once the model is well-constructed, the remaining challenges can be handed over to computational power, and you don’t need to intervene. If you think there’s a scenario that needs testing, fine, just feed it some more data, and again, you don’t need to worry about it. That’s why you’ll notice that Tesla’s autonomous driving team is significantly smaller than domestic autonomous driving teams—by a huge margin. I know Tesla’s team consists of only a few hundred people, whereas many Chinese companies have teams numbering in the thousands.

AsianFin: Is this also the reason why latecomers can surpass Boston Dynamics?

Wang: Exactly. Compared to traditional algorithms, we can’t compete with Boston Dynamics at all. The reason is simple: Boston Dynamics has a bunch of MIT PhDs, and China simply can’t match that.

A robot produced by Unitree Robotics

AsianFin: In your opinion, what will be the most critical differentiating factor for humanoid robots in the future?

Wang: Robots are an integrated product. They won’t have the kind of stark differences you see between fuel-powered cars and new energy vehicles. Instead, the differences will lie in small technical solutions. For example, the size of the motor, where the motor is placed, the working range, the general design of the appearance, the shape of the legs—these are the areas where differences will emerge.

This applies to AI as well. For instance, with large language models, everyone is more or less on the same page, right? The main differences at the moment lie in various details. The GPT architecture, for example, is still relatively clean.

"In our lifetime, humanoid robots can reshape all industries and natural environments"

AsianFin: Commercialization is also crucial. How can startups survive in an increasingly competitive landscape?

Wang: The business logic is very simple. As long as your product is better than your competitors in all aspects, you are almost guaranteed to make money. The remaining question is: how big is the market itself? Currently, our company has a relatively strong foundation in this industry, so we are essentially capturing the more lucrative opportunities within the sector.

AsianFin: What do you mean by "lucrative opportunities"?

Wang: It means achieving higher shipment volumes. For example, with quadruped robots and humanoid robots, we sold quite a few humanoid models last year.

AsianFin: How many did you sell?

Wang: It's probably not appropriate to disclose the exact number, but it’s less than a few hundred. However, we definitely sold the most in the domestic market.

AsianFin : Who are your customers?

Wang: A variety of clients, including research institutions, AI companies, and project implementation teams, among others.

AsianFin: Why are you able to move so quickly and even sell products?

Wang: Because we have a strong foundation. Quadruped robots and humanoid robots are actually quite similar. In this industry, we have relative advantages in areas such as technological R&D capabilities, AI algorithms, manufacturing capacity, and sales channels. Our clients are already in place, and our products are ready to go. For other companies, they would need to start from scratch in all these areas, which requires time to build up.

AsianFin: Are your revenues sufficient to support your R&D efforts?

Wang: The company’s overall gross profit is quite good, and we also have some funding from investments.

AsianFin: For startups in the humanoid robot space, would you say fundraising ability is a core competency?

Wang: Right now, it's hard for me to evaluate this industry because it's just too hot. Many companies with some foundation have secured significant funding, and this money is enough to keep them afloat for a while.

Everyone has plenty of money. When we started, we were very poor. Compared to our early days, it's a completely different story now. Some companies are getting valuations of 1 billion yuan just a year after being founded—it's absolutely absurd. The industry isn't short of money, and neither are they.

However, I feel that when the industry hasn't truly taken off yet, having too much money isn't necessarily useful. If it's not spent wisely, it could just end up being wasted. Right now, neither the technology nor the business models have been fully validated. Throwing money around recklessly at this stage is really unwise.

Take shared bikes, for example—they succeeded because their business model worked. Once the model proved viable, the next step was to scale up, and the only way to do that was to pour in money. Otherwise, there wasn't much else to do.

AsianFin: What do you mean by "not fully validated"?

Wang: It means the technical direction and commercialization haven't been figured out yet. Even if you have money, you don't know where to spend it, or how to spend it effectively.

AsianFin: What are the challenges in terms of technical direction?

Wang: (For humanoid robots) it's about figuring out how to pair them with AI models. That part is still unclear.

AsianFin: Another thing—most entrepreneurs in the humanoid robot space are very young (Wang himself is part of the post-90s generation). Why is that?

Wang: The reason is simple: older people aren't interested in this stuff. Most of the recent advancements in AI technology have emerged in the past few years, and older knowledge is practically useless now. AI technology from more than five years ago is hardly worth looking at. Naturally, younger people are the ones learning and adapting to these new developments the fastest. Traditional internet startups had very low barriers to entry—almost anyone could be a product manager. But humanoid robotics is not a traditional industry.

AsianFin: You just mentioned the possibility of a "genius idea" emerging. Were you referring to the issue of how humanoid robots should pair with AI models?

Wang: Yes, that's about right.

AsianFin: Aren't AI models just like basic building blocks that you piece together?

Wang: The differences behind all this are actually quite significant. Starting from the foundational level of the models, there are countless areas that can be modified. Take the Transformer, for example. Even that can be altered, and now many people are researching how to improve it. Some are even considering abandoning the Transformer architecture altogether and finding something entirely new to replace it. In the field of AI, there are so many areas where you can make changes, and countless opportunities for exploration and innovation.

I predict that by 2025, we’ll see a fairly advanced model for general-purpose humanoid robots. If someone manages to develop this general-purpose model by then, the industry’s momentum will increase even further, and companies worldwide will rush into this field.

AsianFin: By that time, do you think software will achieve breakthroughs first, or will hardware lead the way?

Wang: It primarily depends on software. Hardware alone won’t cut it. Even if you build the best hardware, without software support, it’s useless—it’s just a pile of scrap metal.

AsianFin: So, at the current pace of development, as long as the software is ready, hardware can keep up?

Wang: Hardware definitely won’t be an issue. If there’s a real need, we could pour massive amounts of money into it. Honestly, within two to three months or up to a year, it could absolutely be done. Work overtime, triple salaries, buy up all the necessary equipment, and throw all the financial resources into it—it’s definitely achievable.

AsianFin: Is there a gap between China’s hardware capabilities and those of other countries?

Wang: China has a stronger advantage in hardware, offering better cost-effectiveness.

AsianFin: Why is that the case?

Wang: First, in the U.S., they don’t currently place much emphasis on hardware development. Most of the smart people there are working on software. Secondly, their production and labor costs are much higher than in China.

AsianFin: It seems like they focus more on software, while our strength lies in hardware.

Wang: Exactly. At present, most major companies, especially in the U.S., are primarily focused on software. But for us at Unitree Robotics, we work on both software and hardware because we need to maintain our competitiveness. Unitree Robotics isn’t a large company. Big companies can afford to focus solely on software and ignore hardware, and that might work for them. But we can’t afford to do that. We need to stay competitive, and we can’t neglect either side. If we were to abandon hardware, it would be a losing proposition—too costly and inefficient.

AsianFin : Why do you think robotic dogs have developed more maturely compared to humanoid robots?

Wang: I think one reason is that robotic dogs have had a longer development timeline, and their stability is better. Currently, they don’t involve complex operations like grasping or handling objects.

Another crucial point is that fewer people are working on robotic dogs now compared to before. For example, large language models are far more advanced than robotic models, and the reason is simple—there are more people working on them. Take image AI as an example: the past decade was almost like its golden era, with a lot of people working on facial recognition and detection. Why were so many people working on image AI? Because it’s relatively simple—if you have a functional computer, you can run programs. But working on robots is much more complicated; it requires hardware simulation support. So, naturally, fewer people are in this field. As I mentioned earlier, the development of this industry is accelerating because, in the past year or two, many more people have entered the field. With more people, results naturally start to emerge.

AsianFin: Does Unitree Robotics have its own product roadmap and timeline?

Wang: We release new products every year.

AsianFin: What do you envision for the next generation of robots from Unitree Robotics?

Wang: I think their capabilities will definitely surpass those of our current robots in all aspects, including appearance, performance, and AI capabilities.

AsianFin: Can you provide a specific example?

Wang: We hope that robots can eventually work in factories, participating in tasks like production assembly and material handling.

AsianFin : Is there a timeline for the next generation of products?

Wang: It’s not convenient to disclose that at the moment.

AsianFin: Unitree Robotics has completed eight rounds of financing. Do you think the pace of financing will accelerate compared to before?

Wang: I think it’s about the same. It’s just that the industry is getting more attention, and more investors are reaching out to us.

AsianFin: In your opinion, what is the ultimate goal for humanoid robots?

Wang: In the future, humanoid robots might revolutionize every industry, including manufacturing, services, market production, agriculture, mining, and construction. If we think about the ultimate possibilities, governments could potentially deploy 100,000 humanoid robots, allocate a piece of land, and build a new city there. They could set up the infrastructure and provide housing for free. At that point, ordinary people wouldn’t need to work anymore, as everyone could be supported by robots. This is entirely achievable.

Moreover, the humanoid robots people talk about today are roughly human-sized. But in fact, humanoid robots could be designed to create smaller robots, which in turn could create even smaller ones, continuing this cycle and making them progressively smaller.

Eventually, they might even produce robots as small as cells. What would the world look like then? Who knows—what you perceive as a bacterium might actually be a robot. The entire natural environment could be reshaped. At that point, governments would certainly need to introduce regulations to prevent robots from becoming overly abundant, as they might consume all available resources.

AsianFin: Do you think we’ll witness such scenarios in our lifetime?

Wang: It’s entirely possible. The only missing piece is AI. Once AI breakthroughs occur, everything I just mentioned will naturally follow.

So, I believe this will bring profound changes to the world. I’ve always felt that when general-purpose robots or general AI become fully mature, looking back at today’s society will feel like how we view primitive humans now.

(Author | Rao Xiangyu, Editor | Zhong Yi)