From quadrupeds to humanoids, embodied intelligence is moving into people's living spaces.

What will the robots of the future be like?

Dr. Yu Yinan, Founder & CEO of Vbot, alumnus of the 6th cohort of Hundun Academy, has put forward his unique insight: the future robot is not only equipped with stronger technologies and more complex capabilities, but also closer to people's daily lives.

It can walk, interact with people and take on tasks, but more importantly, it can be accepted and trusted by people, and stay in real-life scenarios.

From the AI cargo-carrying robotic dog "Datou" to the humanoid robot "Vbot ATOM", Dr. Yu Yinan talks about the next stage of embodied intelligence: how robots evolve from technical products to a part of people's living space.

The following is the transcribed text of the video content:

Only products that stay with users are the real valuable ones

Our quadruped robots have accompanied our users to walk a round trip from Beijing to New York in total. Among all users, the one who has the longest usage record has a total walking mileage of about 300 kilometers, which basically verifies the matching degree between the product and its applicable scenarios.

We believe that the entire embodied technology stack requires three core capabilities.

The first capability is the interaction capability between robots and human beings. I think current large language models have basically solved this problem at the technical principle level, and the remaining work is more about productization and engineering. The second very important capability is the locomotion capability. The third capability is the dual-handed manipulation capability.

These three capabilities are the most fundamental and core capabilities we have identified for embodied intelligence. In fact, for future humanoid robots or all robots in the future, these three capabilities are all indispensable. It's just that in certain scenarios, we may weaken or ignore the demand for one of the capabilities.

We keenly realize that the locomotion capability has been very mature at the current technical principle level, but no company has done an excellent job in its productization stage. In our opinion, it actually falls into the "gap between principle-level technology and engineering-level technology", a gap that no one has bridged yet. Therefore, we believe this is the first market opportunity.

This kind of value and capability itself has great potential. So why can't we be the one to achieve it? We soon invested a lot of work in this direction. In fact, we are indeed the first player in the industry to push this type of product, or this kind of capability, to a state that can truly satisfy users.

When the robotic dog first arrives at home, users will definitely unbox it and interact with it at the very beginning. But you will find that for long-term users, the most important usage scenarios are endurance, load bearing and movement. For example, some of our users take the robotic dog to shopping malls, some take it to climb mountains, and others take it to visit parks.

Frankly speaking, I don't chat with our family's robotic dog very often, but my daughter talks to it a lot. What I do more often is take it out, let it help me carry my bag and bear the load. As a movable robotic dog accompanying you around, it is already very nice. Therefore, endurance and load bearing are in my opinion the very obvious external and functional values of the product.

From Quadruped Robots to Humanoid Robots, Technology Is Only the Starting Point

In fact, we didn't spend a very long time developing our humanoid robot Vbot ATOM. Moreover, based on the solid foundation of our quadruped robot, which has already completed the closed loop of the entire business chain and broken through every link of the whole chain.

The full breakthrough of the whole chain has accumulated a lot of partial capabilities for us, such as supply chain capability, R&D and design capability for basic product components, including our models, algorithms, data and software. We quickly migrated all these achievements to the humanoid robot. In fact, the technology stack of our current humanoid robot is basically similar to that of our quadruped robot.

Of course, we have some basic principles for technology selection and judgment. First, the technology must be low-cost. It should not only become cheap when the production volume increases, but be inherently low-cost based on its basic technical principles. Second, it must be fully verified. Third, it must be algorithm-friendly. Therefore, we later came to a conclusion: robots should be the combination of matching capabilities, roles and scenarios.

What are my current capabilities? For example, I have the interaction capability, the locomotion capability, and the generalized manipulation capability — these are my capabilities. But there is still a gap between these capabilities and scenarios. What is this gap? It is actually the role. How do you turn these capabilities into a specific role?

The same logic applies to humanoid robots. First of all, they have strong basic capabilities, and meanwhile we place them in specific scenarios and endow them with a certain role. Then this role can complete certain tasks or provide certain services, which finally bring value to users.

When we connect the entire chain together, you will find that the most important thing to do right now is actually the part before role definition. Only after you do this part well, and endow it with different roles in different scenarios, can it deliver real service value to users.

I think the first and foremost thing for a C-end robot is to look good — appearance speaks for itself, and it should be friendly and approachable. Users and customers will genuinely fall in love with your product the moment they see it.

I have full confidence in my team's technical capabilities, and I am not worried about any technical shortcomings at all. What I value more is user experience: what feeling do I get when I stand in front of this robot? What feeling do my family members, my children and my friends get when they see this robot?

I think human's feeling is in fact the most, most core point of a product. Apart from endurance and load bearing, user experience is the top priority. The same is true for current humanoid robots. I believe our humanoid robot will be able to do some work and deliver functional value in the next three to six months, but that is far from enough.

I think the most important thing is to return to human's feeling: what kind of mood do I have when I watch it work? This is a point our company attaches great importance to: we put human beings at the center.

What is the vision of our company? Our definition is "to create robots for living spaces". We believe that the products we are developing now are not designed to replace human beings, but to serve human beings, and to do the things that human beings want to do but are not willing to do. Therefore, we still put human beings at the core, and the initiative is always in the hands of human beings.

Recently there was a very interesting event: we jointly established a "Aerospace Embodied Intelligence Laboratory" with a university in Beijing. This sounds very distant, but also very interesting. Because you can imagine, if human beings are really going to explore outer space in the future, there will definitely be a shortage of space labor force. At that time, robots in space will not replace human beings, but actually enhance human capabilities. It is very likely that three astronauts will bring hundreds of robots to develop and build in space, no matter on other planets or in the universe, and this is exactly what we are eager to do.

Embodied Intelligence Is Evolving From the "Spring and Autumn Period" to the "Warring States Period"

My judgment for 2025 is that we are currently in the "Spring and Autumn Period" of embodied intelligence. The "Spring and Autumn Period" means that everyone holds their own different opinions, and different players take turns to show their ideas.

The "Spring and Autumn Period" is defined by my own judgment of what this industry should be. When we enter the "Warring States Period", the market will be the final judge: if the market recognizes you, you are qualified; if the market doesn't recognize you, you are out.

But what you can see right in front of you? It is whether you are standing on quicksand or on solid rock under your feet. That is something you can see clearly. Therefore, every time you make a leap forward, there are two most important points. I think the first point is that you have to make sure this leap can bring you to a broader new sky.

But at the same time, you should not jump forward if you see quicksand ahead. The ideal situation is that there is a higher solid rock in front of you. The so-called risk aversion is actually the two sides of this matter: you need to focus on two conflicting goals at the same time: you can get the chance to touch a higher ceiling, and you won't fall into the river when you land.

Therefore, you don't have to jump in recklessly just because the ceiling ahead is high enough; or you can't be too conservative to make a choice that the ceiling is very low but the bottom line is very high, then you jump over. I think every major choice is to consider both the ceiling and the landing point at the same time.

Hundun Is Not the Answer, But a Space for Thinking

First, it is about the thinking of First Principles. I think the entire framework put forward by Professor Shanyou is very fundamental. It is not a specific solution to your specific difficulty or problem, but more like the origin of your thinking. But I think the most important thing is that the environment of Hundun is very good, which provides you with a space for thinking.

And there is a group of like-minded friends. Since all of us are entrepreneurs, when we encounter similar problems, we will communicate with each other and learn from each other; when we face current difficulties, we will comfort each other, pat our friends on the shoulder and say "everyone is going through the same thing", then you will feel much better.

This article is from the WeChat Official Account "Hundun University" (ID: hundun-university), written by Hundun Academy, and authorized for release by 36Kr.