Inside WAIC 2026: a glimpse of embodied AI innovations, humanoid robots

[00:00:20]So this >> Hello everyone, welcome to Shinua. Live, this is Ding with China's Chinua News Agency. I am now at the 2026 World Artificial Intelligence Conference in a high level meeting on global AI governance in Shanghai. So we are now bringing you a front row view of the latest embody AI innovation and humanoid robots. So look around here. So I think this is maybe the place that um you can see the most kind of the robots on this planet. So here I see a very interesting thing that's at table tennis. Let's get a closer look.

[00:01:02]So look at this live rally. So this is the uh hope AI table tennis challenge.

[00:01:07]So it is a top tier physical AI test that even become an official event of the world humanoid robot games. So let's get inside.

[00:01:17]So we can see a human player is just playing with the robot player. So who's playing better? [laughter] What do you think? Yeah, you can see I think right now like uh robot still like learning how to become a better athletes. Um but maybe in the two or three years uh I think it has the confidence to compete with the Olympic champions. [clears throat] >> So how does it work? What do you think is the robot? Uh how can he read you know the ball the speed spin you know the speed of the table tennis ball is so quickly that for the human beings it's very it's even hard for us to capture everything. So how can the robots do it?

[00:01:59]Okay, that's a good question. Uh getting a robot to rally back and forth with the person is genuinely hard and the difficulty is uh spread across several front at once. the embodied AI AI that acts through a physical body in the real world not just on a screen. Uh still it runs into real bottlenecks when it tries to match human physical skill. The hope table tennis platform exists to bring together the best of industry research and sports science to break through those bottlenecks with the goal of opening sharing within three years the training model that gives robot something close to human level athletic ability. So those bottlenecks uh fall into four areas that we see uh the robot body uh and you can see we're using the uh aboard uh like A3 right now one of the best uh like humanoid body that can play ping pong and 3D perception which is how it sees the world in the three dimensions and the physical training and also the data acquisition and understanding. So playing at the level of a human athlet dress all four at once. That's what makes it so difficult.

[00:03:09]And on the robot body, it must make fast balance sideway movements, swing the paddle, and also even perform landing saves while actually unstable. So you can see how the body some like moves. Um that shows how difficult to really manage the like human body the robot body that is in the posture that stands alone can keep upright.

[00:03:31]>> Oh can I have a try? Uh I really want to have a try.

[00:03:37]>> Oh can you hold this? Thank you.

[00:03:42]>> I don't know whether it is hard for me as I'm a beginner of the table tennis.

[00:03:52]So I can see the bowl is quite different from the normal one. The color the color is silver.

[00:04:05]Oh, >> so fast. I lose. [laughter] >> So, as I just mentioned that the color of the table small is quite different from the normal one. So, the normal uh the color is like the white or the yellow but you see you can see the color is silver. Also I think maybe it is better for the whole system to capture the speed or the spin. Is it right?

[00:04:37]>> Right.

[00:04:38]>> So we can see there is a pearl circle of the camera just on our top. So what does that work?

[00:04:45]>> Uh so honestly right now I think um as of 2025 no humanoid table tennis anywhere in the world has the purpose-built solution for screenshot and near net walls. So there are two reasons. First, the robot's 3D perception system still cannot capture the spinning parameters of the ball as a small and fast as a table tennis ball.

[00:05:06]Second, returning a near net ball requires a robot body to extend out of behind table strike and recover the stable standing postures. All while statically unstable, meaning it stands alone, won't keep it balanced. So we are uh using the cameras and also uh the algorithm to help uh see where's the ball coming from every demonstration of the core principle of the inbody AI that is the perceive decide act and adapt in the real time.

[00:05:38]If you are just tuning in we are here at the 2026 world AI conference and a high level meeting on a global AI governance in Shanghai. This year's event draws over 1,100 global tech firms. Organizers stash exhibitors will show more than 4,400 products and technologies with over 300 products making their global debut.

[00:06:05]So here where we are is the booth of the HHS robotics. So here we can see uh there are so many uh robot robots on their booth. So, let's talk to their staff here. Hello. Good morning.

[00:06:20]Morning. Morning.

[00:06:21]>> Yeah. How are you? Yeah, fine.

[00:06:23]>> Yeah. Can you give us a brief introduction about your booth?

[00:06:26]>> Yes, of course. Uh, yeah.

[00:06:30]This is our household scenario area. We actually um show up our uh Juliet real robot. What? Her name is Julia.

[00:06:39]>> Yes, she's a female. What a what a lovely name.

[00:06:42]>> Uh she's actually collecting things. um um in a uh household scenario.

[00:06:47]>> Yes.

[00:06:48]>> Um he when she monitor something fall on the ground like closes or rubbishes.

[00:06:55]>> So the engineer here they are still uh telling her how to do what to do.

[00:06:59]>> Not actually uh she actually can uh make the decision back by him by by herself.

[00:07:06]>> So when she collecting things they don't have to be engineer uh to control things. she can do it by herself because we have our V model which is actually a uh advanced uh model which can facing um sophisticated scenarios like in in a home. So um even though this scenario is not that mature in the for the market but I assume 3 to 5 years there will be a very good uh popular um products for the for the market use. Yes, I can see Julia is based on something that's with the wheel. So, it can move very smoothly like inside inside.

[00:07:48]>> Yes. Uh it can move very thly but also very quietly. Um if you when you use um 5 type, it's not very quiet. So, it will be noisy in the home. So we choose the real one to collect things even though she's she's doing something not that perfectly but we we're using that and we I think our technology is uh going to that way very soon.

[00:08:16]>> Yes. Is Julia the same robot that on the picture? Yes.

[00:08:20]>> It's the same one.

[00:08:21]>> Yeah. She she she not only doing things in our home scenario but also in our retail scenarios. I can show you some retail scenario.

[00:08:30]>> Yeah. Yeah.

[00:08:31]>> Yeah. This is embodying AI for retail scenario. We're not for sure. We are for business you. We have a business circle.

[00:08:39]Uh when put a such a humanoid robots in a box, it's very attractive for uh the customers. Uh this is Najing Road. In in this year festival, we put out a soul for 11 days.

[00:08:58]Uh yeah but this show actually give us a uh a full circle business scenarios which can uh show the inbody AI can really make money.

[00:09:10]>> Yes.

[00:09:10]>> So this this type of business mod actually we got a copy to all of our uh market in not just in China but also in worldwide. I think I have noticed this retail uh space just uh on the uh other side of our pavilion there that is the first world's first humanoid operated media space.

[00:09:33]>> Yes.

[00:09:34]>> Ah I think I think I already noticed that yesterday.

[00:09:37]>> Yes. Because that's that's the first one in the world that can show the robot can real sell things like a salesman. So this this kind of capability actually we're not to replace human beings with labor. We're doing that things for more opportunities to show uh the retail a new types of a retail things. So the robots can do that.

[00:10:04]>> So how long time can it works every day?

[00:10:07]>> Actually uh 20 24 hours a day 7 days.

[00:10:11]>> What a hardworking employee.

[00:10:13]>> Yes. because uh she can go to go to the charge station automatically >> by by herself and so can can do consecutive things.

[00:10:23]>> So it has a smart brain and it can work with non-stop.

[00:10:27]>> Yes, of course. Oh, >> so now let's move to this part. Okay.

[00:10:31]>> So there are so many robots here. So we can see like the robot dock.

[00:10:35]>> Yeah.

[00:10:36]>> And also the hand is it the robot hand?

[00:10:38]>> Yes. So we can see uh if [singing] their staff or customer so they're like uh remote controlling. So what are they doing?

[00:10:46]>> Yeah, this is our quadruped uh new types of robot Nebula. It's called Nebula.

[00:10:52]>> Nebula.

[00:10:52]>> Nebula. Yeah, this one >> is the name of the robot. Not only for the research use but also for some uh special circumstances when human cannot enter some spaces or environment [singing] that very dangerous [music] or were very harm to the human health. So he can do the carry things. And this one is our open arm. Uh this one is a uh is a drive arm. That one is a slave arm.

[00:11:20]the when you use the drive arm you can do the things and and that one can do the same exactly the same thing you can do. [music] >> So if you want to try to >> Oh sure sure. So can you teach me how can I >> Okay.

[00:11:41]So we are controlling the robot robbo arm on the same side.

[00:11:44]>> Yeah.

[00:11:48]While clicking the button, you can just uh clip the the clap.

[00:11:53]>> Oh, let me try it. Let me try it.

[00:11:55]>> Let me hold on.

[00:11:59]>> Okay. So, now I want to >> Oh, I have to move. Oh.

[00:12:15]So, look at that carrot. Oh, one carrot.

[00:12:18]All my R.

[00:12:20]Oh, this side.

[00:12:23]Whoa.

[00:12:27]So, let me try the right side.

[00:12:31]Is it the same way? [music] [laughter] >> Oh, no. No. Further, further. Oh, no.

[00:12:37]No, no.

[00:12:42]>> Oh.

[00:12:49]Yes, that's the same uh situation.

[00:12:51]>> I think I should stop that.

[00:12:52]>> Yeah.

[00:12:52]>> So for maybe the ma mass, >> we actually using the this not for entertainment. We using this for academic research.

[00:13:01]>> We sell them to the >> So this is not about the entertainment.

[00:13:05]So you are you are doing the experiment like?

[00:13:07]>> Yeah, experiment. Uh we we call it embodied AI data collection. You follow me now.

[00:13:13]>> Okay. Okay.

[00:13:17]We have >> wow look at >> embodied AI collection. We're not just doing the data collection but we also do the uh data training and modules training and model uh deployment.

[00:13:32]>> So what do you think of the open ecosystem you are doing? What do you think of it?

[00:13:37]>> Uh we actually making the ecosystem not just for ourselves. We for the site for research it for example we doing that for Ning universities we put all our data collecting uh series robots to the in the university to build up a um uh academic environment so they can do the the whole data process by themselves [music] so we're not just doing for ourselves >> I think this is very good for the development of the whole industry >> yeah not just uh in the academic uh industry but also in the occas occasional college because China's um industrial upgrading actually drives more uh occasional colleagues to set up new majors in the embodied AI related.

[00:14:31]>> Yeah. So for all the scenarios that these are all the scenario for you to train the embodied AI to be smarter.

[00:14:39]>> Yeah. But also we can provide them our textbooks bilingual textbooks for the inbody AI training as well.

[00:14:48]>> Wow. Thank you. Thank you. Thank you so much. Okay. So our tour is just go on.

[00:14:54]Let's move to the other booth. Let's go.

[00:15:00]>> Okay. Next up is Unix AI. It is showcasing how embodied AI seamlessly blends culture, catering, craftsmanship, and family life. So now let's talk to the staff here. Hello. Good morning. How are you?

[00:15:15]>> Great.

[00:15:15]>> Okay. So I can see so many robots at your booth. So some of them like playing concert, their bands, some of them they're like working in the kitchen. So can you give us a brief introduction?

[00:15:27]>> Uh so basically our bus is focusing on the household environment. So you can see that actually we have uh a couple of the uh activities we're having here. Oh, so you are showcasing so many demo scheduled here.

[00:15:40]>> Yeah, it's pretty tight. See, it's pretty tight. So, we have the uh uh household breakfast service uh together with the uh hotel service. Very interesting.

[00:15:49]>> Yeah, it's on that corner. We have a white rob basically uh after finishing the house uh doing the making the breakfast, but go to the uh the bat, making a bath, washing the dishes, uh washing the uh uh making up the beds, and then finish taking care of all the household areas. And over there we actually we have a very interesting pizza cooking kitchen over there. So actually you can interact with the robot. You can ask the robot what kind of pizza >> what kind Yeah. What kind of flavor you want. Actually we have prepared three flavors you can choose between. So uh actually the robot over there uh it will make the pizza uh from the very beginning uh all the way to putting the pizza uh into the uh oven taking out uh and also will actually uh use a glass of uh also use take the glass to get some coke make a pizza combo and uh over this line we are having a robot band. So we are the only robot band uh in this uh uh in this conference >> in this conference. Yeah, in the WIC that actually the robot is uh is really playing the band. You can you if you listen [music] carefully uh you see >> their name is Xband Xan [music] Xan.

[00:17:04]>> So we have the Dra Dra hand.

[00:17:07]>> Yeah, [music] it's it's working. It's working. It's just not the part of the piano is working but it it is working.

[00:17:14]>> Let me try.

[00:17:14]>> Let's we can we can we can get a song.

[00:17:25]So your word uh your staff is controlling there.

[00:17:28]>> Uh it's not controlling.

[00:17:30]>> It's a remote control.

[00:17:31]>> No, it's fully autonomous.

[00:17:34]What our staff is doing is just picking a kind of song. It's going to We have iPad as a UI interface. That's what we are having here. So you can see [music] that 80% of our staff Yeah. You see that head is shaking. Yeah. Also you see the trump [music] if you listen carefully actually the uh the dextra hand is really hitting the button of the piano. [singing and music] It's really making a sound here compared to some other stuff that's basically having the music coming from the uh component [music] background. We're really playing something.

[00:18:15]You see also the drum. [music] >> So we are making a traditional Chinese song [music] in a relatively guitar way that we Yeah, we can it's making [music] some uh oldfashioned sound in a new new kind of way. So if you uh look over there actually the [music] the pizza machine the pizza robot.

[00:18:37]>> Oh, you just mentioned that I can choose like three flavor.

[00:18:40]>> Three flavors. We have >> Where are they?

[00:18:43]>> Okay, it's in the iPad. What?

[00:18:44]>> Okay.

[00:18:47]>> So, it's cutting the pizza. I think it's the last step.

[00:18:50]>> So, after >> It's the last step.

[00:18:51]>> Yeah. I think it's the last step.

[00:18:52]>> They're doing the packaging.

[00:18:54]>> Yeah. Yeah. Doing the packaging and and and splitting.

[00:18:56]>> Uh-huh.

[00:18:59]>> It's a very long horizon and impulation task. Basically, uh two of our task will last around 10 to 15 minutes. We're making one or two pizza together with the uh the breakfast.

[00:19:12]I think we are the longest horizon of the manipulation task we can see in in uh in the WIC there or sometimes I stuck there and let's see how how will we be interact with the uh uh with the with the with the machine.

[00:19:29]>> You see it's changing the hand. It's changing the hand.

[00:19:33]>> It's pretty smart, right?

[00:19:34]>> Yes. [music] So it's trying to find a relatively smart way for manipulate with that object.

[00:19:46][music] >> Yeah. You see it's pretty it's pretty tight. You see only a very very a small.

[00:19:52][music] >> Yeah. You see it's putting inside adjusting the right position.

[00:19:55]>> Let's see. Let's let's see how it's doing.

[00:19:57]>> It's not it's not finished yet.

[00:20:00]>> He's still it's still trying to adjust it other stuff. Yes.

[00:20:05]We have to find the right angle. Wa!

[00:20:07]>> Finally. Finally. Good job. Good job.

[00:20:10]>> So, so we can order a pizza and see how how it tastes.

[00:20:15]>> Yes. Yes. After after I finish this pizza.

[00:20:17]>> Oh, so here we can see there are three flavor that for customers can choose.

[00:20:22]And >> I haven't finished it. He needs to wait until this one is finished.

[00:20:27]>> Okay. The robot is making a pizza now.

[00:20:30]So the current one is a vegetarian vegetarian.

[00:20:34]>> Yeah, it's a vegetable pizza.

[00:20:36]>> And we also have pepperoni and the classical cheese pizza.

[00:20:40]>> Yeah, >> I think >> it makes me think of the game like the crazy kitchen.

[00:20:47]>> Crazy [laughter] kitchen. Yeah.

[00:20:49]>> Uh yes, actually it is the crazy kitchen cuz you have two robots actually over there. They need to collaborate with each other. Which kind of pizza you want? We can order a new one right now.

[00:21:00]Classic cheese.

[00:21:01]>> Okay, let's let's order classic cheese pizza.

[00:21:07]>> Okay, >> it's activating the machine right now.

[00:21:12]>> Oh, it's working. It's not working.

[00:21:15]It's >> not working.

[00:21:20]>> You see there are different different flavors over there. You can adjust uh adjust the recipe according to whatever you have ordered over there. So actually in the first app I think it's a I'll say it's a it's a papar sauce or it's a tomato sauce over there.

[00:21:36]>> Tomato >> tomato sauce. Right. So uh the idea the first step the idea is that you have the tomato sauce first in the put first. Put it in the middle and use that the stuff to spread >> relatively evenly on that on that sphere.

[00:21:55]>> Okay. Finish the first step.

[00:22:03]>> [music] >> Okay. Then for the classic pizza, right?

[00:22:09]Cheese over here.

[00:22:16]>> It commonly use like 8 minutes to finish each of the pizza.

[00:22:20]>> So I think the the point of that is not uh it's half about the machine. The other half is you need time to heat up the pizza. You order, you can eat the pizza actually.

[00:22:36]>> For this one is is a simple.

[00:22:43]>> Oh, we can see there are so many visitors at the booth.

[00:22:47]>> Whenever the machine is working, the robot is working, there's always crowd of people surrounded. Uh this is the robot that is the good at making the breakfast. It's making the salad >> is uh we uh it's the butler machine.

[00:22:59]It's the butler robot.

[00:23:01]>> So basically will it's making the uh uh it's making the salad. It's actually it's a combo for the whole breakfast. So after that it will goes to that area.

[00:23:09]You see it's a bedroom area. Uh it will make up the bath uh do the cleaning, do the laundry, uh do all all these kind of stuff. The future is already coming. I I cannot guarantee for today you can replace a 100% of the housekeeper but we can do two/3s. We can do two/3. So except for the very complicated bat making a bat changing we can do the half uh and some very very uh uh precise activities we can do the rest of the so we only 30 minutes we we make a we make a hotel run.

[00:23:39]>> Thank you.

[00:23:40]>> Thank you. Thank you so much.

[00:23:45]We've now arrived at hall three. This is the in body intelligence pavilion. So every booth here is full of Chinese developed humanoid roy and models.

[00:23:58]And you can see the flag here uh is we win the champion uh of global cup this year and also we have the championship last year. uh it has like 23 uh golf of degrees. So it's like you can see it's act very flexible on the field and uh our application is most like uh for the research and for the schools uh and we have like a whole package of the solution is from the robots itself for the software and the curriculum. Uh it's all combined in one.

[00:24:39]>> So we can see the star is just restart the game again. Yeah, >> exactly. And it's all AI without any remote control.

[00:24:51]>> Autonomous playing. Yeah.

[00:24:55]>> Yeah. And you can see they have the multi- aent.

[00:25:00]>> Never mind. Never mind.

[00:25:02]>> The football game should be always fiery.

[00:25:04]>> Yeah. Yeah. Yeah. For sure. And it's very interesting for the students like uh students with the uh king on this industry they can watch the game they can compete and they can learn something from this as well.

[00:25:27]>> Yeah. And as you can see here we have the space limited so it's only two versus two. Uh and uh like for our games actually we have five versus five. So it's going to be like more decision making uh more collaboration between different robots.

[00:25:51]>> So what is the highest speed for the robot can running? I think for now basically it's about 4 m per second. Uh it's enough for the so field. Uh but for our like flagship uh for for this one is kind of almost run fast.

[00:26:30]So looking into the future. So what's your most exciting real world application you expect maybe in the next one or two years? Uh I think the most uh exciting application in the future is like in a I think maybe well in two areas. uh the first uh education and research at scale like we are rolling out our 160 and 10,000 schools initiative uh which is bringing booster K1 like like this one to into the classroom uh as the first robots of students use to learn about the AI. So it's not just about selling hardware uh it's also selling a like complete ecosystem from the hardware to the software to the curriculum and also to the commentation as well. Uh so it's a first place and the uh second I think is the developer ecosystem uh like booster studio like this one is the first IDE uh specifically for the AI developer. So like developers can view can train and validate algorithm in the simulation like like this one. Uh so a like they can deploy to the real robots but just like and it's all beginner.

[00:27:56]So like some people like who has no experience in this industry they can learn something they can use this uh I say to to go into this industry. Yeah.

[00:28:10]>> Okay. Thank you so much. Thank you for introduction. So this wraps up our tour of the 2026 World AI conference for today. So from table tennis robots and championship software robots or even to the multi- scenario service robots. So everything we've seen here sends one Korea message that is China shares its innovation achievement openly with the world and is committed to building a smarter more human centered future alongside global partners. So instead of the closed tech barriers, we can see so many uh new development platforms share robot hardware and across border competition. Okay, so that's all our live today. Thank you for watching and I hope you like it. See you next time.

[00:29:00]Bye-bye.