Hello, welcome to Roundtable where we serve up piping hot debates on the issues that sizzle in China and beyond.
I'm Niu Honglin.
Humanoid robots are having a moment.
From factory floors in Shanghai to viral showrooms in Silicon Valley, lifelike machines are stepping into the spotlight can help but make you wonder which of these robots are truly capable breakthroughs and which ones are merely cleverly designed models designated to impress investors and social media.
As China sets 2025 as the first year of humanoid robot mass production and global competitors rush their own models to market, the question becomes more urgent.
How close are we, really, to welcoming a humanoid robot into our homes?
For this episode, I'm joined by Steve Hatherly and Yuxuan.
Now grab your virtual compass and follow us to the heart of the discussion.
Humanity has always built tools, but this time the tool looks back.
As humanoid robots grow more agile and more intelligent, they force us to face an uncomfortable question.
What happens when a machine begins to resemble us not just in shape, but in ability, in behavior and eventually, in judgment?
But before asking that question, we should first ask are we ready to hand over the factory keys or even the dishwasher?
Are these really smart robots, really smart enough?
Are they really as efficient and reliable as we imagine?
And are we on the cusp of an industrial revolution or simply enjoying the greatest tech demo of the decade?
So today we are going to do a relatively more comprehensive reality check on the development stage of humanoid robot.
It has been a year for this, but we pick this show today for a good reason because a heated discussion on the Internet has been sparked by a specific Yeah, this robot, or this newly released robot has like, raised a lot of discussion on it.
This is a humanoid robot recently unveiled by a Chinese carmaker, actually.
And it has drawn widespread attention recently.
Many foreign media outlets reporting this.
CN reported it.
CNBC covered this story.
And that robot is Xpeng's or Xiaopeng's new humanoid robot, Iron.
And...
The debate online was this.
Is that a human inside that robot suit?
Have you checked the video?
I thought it was a joke when I saw the news story.
How can you accuse a real news press conference with that?
I watched an Australian vlogger's comments on this.
He was at the unveiling of this robot and he said that.
He said in the crowd to the person next to him is that a person?
And he was there, right?
When I watched it, I thought, whoa, okay, this is actually... I know what you mean.
Right?
It's not fully like a human being.
The walk was very slow.
If you haven't seen it yet, which I'm sure our audience has, but just in case.
The walk is very slow and deliberate, almost like a model walking down a runway, but at half speed.
So I can see why people thought, oh, that's really quite human-like.
But at the same time, it still had some robotic element movements to it, in my opinion.
Yes.
Yeah, I can understand people are saying this, and I can understand why you said that Steve, because I think when people are questioning about this, it's because they think
It looks like a human aimed this robot, and this human is trying to mimic a robot walking.
Yeah, exactly.
That's why it is so good that that looks like a real human.
Who does that, though?
Anyways, I got to the point here.
So that was the moment that the entire Internet started to talk about how great the technology is already becoming, if it's really true.
And the CEO of the company even came out a bit later to explain this is really a humanoid robot.
Yeah, and not just explain.
He actually released a video showing the inner workings of the robot, including the muscles, the lattice work muscles, lattice-like muscles of the robot.
And that was done to prove that it was really a robot and not a person inside.
Yeah.
And they even cut open the robot's leg covering because it was wearing like a suit on the outside, which made it look more like a person when it was walking.
They cut that suit open. during a live event to expose the mechanical components inside.
But we have to admit we've been listening to and exposed to the idea of humanoid robots for a really long time and it has been developing rapidly.
So where are we now?
Where are we when it comes to humanoid robot status quo?
Actually, there has been a lot of discussions on it and many conferences are discussing the development of the whole humanoid robot industry and how we can develop it and what is the status quo.
So on November 5th at the 8th Hongqiao International Economic Forum the sub-forum is called innovation and development cooperation in humanoid robots was held and this event was strongly hosted by the Ministry of industry and information technology and also the Ministry of commerce and organized by the Chinese Institute of electronics, and it invited renowned domestic and also international experts and representatives from leading companies to discuss the current state of the human or robot industry, also technological breakthroughs and application prospects, trying to have a discussion on these kind of intelligent momentum into the new industrialization process.
If you're wondering why and I think this is a fair question why are they trying to make these robots look like people?
Not every robot does look like a person, right?
And sometimes it's not necessary.
But I did a little digging and I found the answer to be quite interesting, because a humanoid robot is designed to resemble a person.
It looks like one.
It has a head and a torso and arms and legs and it has a human shape.
And the reason is because...
We humans built the world for humans, for other humans.
We had that in mind when we made our decisions on construction and other things.
And since our environment, our buildings and our tools and our vehicles, they're designed differently.
For a human body with two arms and two legs and hands, and therefore the most efficient robots to operate in that world designed by humans, is to be shaped like a human.
They're kind of like a key, if you will, in a world that was built by us.
That's why some of these experts are saying that humanoid robot are expected to become the next generation of the super Yeah.
Following smartphones and also new energy vehicles.
Yeah, the super terminal, like the all-in-one interface.
Yes, because ultimately we want humanoid robots, or robots in general, to replace certain positions that we do not want to put people, human beings, in.
Mm-hmm.
Some positions might be too repetitive.
Some positions might be in an environment that is toxic to human body.
You do not want humans to continue taking the risk or putting on all the gears to protect themselves from the toxic environment, but still having to stick to it to do the job.
In order to achieve those, humanoid robots are definitely much necessary in the industrial world.
And that is also the reason that China has been working on developing these robots for quite a long time.
And this year, we are looking at relatively a breakthrough period or acceleration period.
We're talking about a rapid cycle of technology breakthroughs, capital influx and scenario lending.
And some people have even called 2025 as the mass production Starting year, meaning that from this year on, we're encouraging the technology to transform from technological prototypes to engineering, implementation and scale applications.
Yeah, you mentioned that term scenario landing.
That's sometimes called application landing too.
And it means taking a technology out of the lab and then successfully putting it to work, solving a real-world problem in a specific situation.
So they're in that stage now.
Yeah, there's a lot of money being invested.
Companies are in a race against each other.
It's like an arms race, but an arms and legs race.
And now they're in mass production, and the price is getting cheaper as well.
Yeah, robot prices, they've come down pretty quickly, actually.
They've moved from laboratory costs of...
10s or even hundreds of thousands of yuan, to about 10000 yuan or less for some particular models, which is I don't know if the price reduction was expected this quickly, but we're seeing it already.
Yes.
Exactly.
And one example is that I've seen...
One, China's first robot dealership, they call it 7S store.
Because in China, when you call it like a car dealership, people call it 4S stores.
And that stands for like sales, spare parts, service and survey.
And so that is a 7S because that's also including solution showing the robots and also school meaning training the people who can maintain and um These kind of robots.
So they have this first robot dealership store for humanoid robots and they opened on November the 11th in Wuhan.
Have you gone?
No, but it is like a brand new store and it covers and there are a lot of robots aimed there.
The cheapest robots are priced at just over 70,000 yuan.
That's about $9,800. while the most expensive model can cost you more than $500,000.
Yeah, they can get quite expensive too.
Of course, and you put them in different scenarios.
And we mentioned this.
Ultimately, we want to have them being used helping human in different types of situations.
And maybe it starts in industrial and manufacturing implications.
Perhaps we can replace some human workers from the position they do not want to stay in.
And what are some things that we expect humanoid robots to do?
Yeah, a lot of the areas has actually has already been using humanoid robots.
For example, industrial and manufacturing scenarios.
The consensus among industry experts is that Humanoid robots will first achieve large-scale deployment in these environments, and the upcoming year is projected to be a critical year for the application of humanoid robots.
And according to Leng Xiaokun, chairman of Leju Intelligent, It is predicted that next year, companies in the industrial sector will see the procurement and delivery quantities exceeding 10000 humanoid robots.
Ooh, that's a benchmark.
Yeah, they'll be working in production lines.
They'll be deployed industrially in that fashion.
One of the things that people are, maybe even for people like you and I.
The thing that people are waiting for is the services that will be available from humanoid robots in the home.
And if you watch some videos online about this, it's pretty interesting.
But it's also very, very, very complicated.
Much more complicated than I even realized.
When you look at things like Pouring water or watering the plants, or folding clothes or folding laundry or whatever, that is an extremely complicated thing to do because of all of the things that the robot needs to recognize within its environment.
And that's only talking about the technical side of moving the robot around, never mind the safety issues.
Yes.
And when we talk about having robots working in production lines yes, there are already a lot of different machines or robotic arms or robotic equipment helping different types of industrial production process.
For example, we see that in Qingdao in China They undergo training for tasks like bolt tightening, part installation handling at national level.
Smart manufacturing demo plant.
These things are happening.
They're training robots to do that.
And with a robotic arm being able to do that, you can put that arm on a robot and that humanoid robot would be having that ability on top of whatever ability it already has.
And also, Steve mentioned housework scenarios in the United States operates the well in the Walmart, Walmart's store.
We see this having robot handling orders, taking and beverage, making them independently, and they even provide a multi-language support, which was cool one or two years ago, but now you find it quite common already.
It's kind of expected now.
Kind of expected, yes.
So a lot of these functions have already been put in use in a lot of different scenarios.
Now we're looking at integrating them into one robot and having the robot working in these environment and carrying out multiple tasks.
So what are some other things in the future that they may be able to do, like general factory operations?
What we're what we're mentioning before is that maybe one robot in one specific position and they can actually do something that is like specifically coded.
But in the future, when the robot can be developed so maturely, they can just do everything more actively by themselves and recognize the scenario and environment by themselves, and to do and follow the order that you gave them.
So loading and unloading and transferring and handling, and also automotive manufacturing like sorting and proportioning materials.
Also in areas like 3C computer communication, consumer electronics area that this robot can be very, very useful.
Also, shipbuilding, like grinding and polishing.
Also, agriculture area can also be used in a lot of areas, like precision field operations.
So also future capability improvement is one thing that a lot of these companies are working on.
Like.
Within two to three years, humanoid robots may be able to perform more complex operations within industrial settings.
Of course they will.
Because they're continuing to gather and gather and gather more data and using that data to train these and code these robots.
And it's only going to get better and better and better and better.
There was a report from UBS Securities that predicts the volume of humanoid robots in the industrial sector.
They think it's going to exceed 1.5 million units. by the year 2050.
And use in factories is just going to be commonplace, I think.
We're talking about it a lot on Roundtable and other media outlets are talking about it all the time as well.
Most recently, I think we talked about...
Robot chefs in the restaurant.
For us, this is how I feel about it.
For us, that's big news these days.
That's going to be the norm, I think, in the future.
We're just in the transition phase right now and we're not used to it and it's all new and there are so many companies investing so much money and time into this, so many resources.
We're headed towards that norm.
It sounds exciting to imagine a beautiful, clean factory with full of humanoid robots walking around carrying cases, making brand new, like new energy, cars and stuff.
But when can I expect them to join my family and work for me?
It's going to be a while because there's so many things to work out.
Let's go back to the household chores example.
So, folding laundry, the robot has to recognize, see a soft crumpled T-shirt and understand its three-dimensional shape.
Then it has to plan how to pick it up with what corner to grab, right?
The way that humans do.
We just instinctively know where to pick up the T-shirt or a towel, but the robot doesn't know that.
It also has to manipulate the flexibility aspect of the towel, right?
And that's one of the hardest problems in robotics, they say.
And then it has to know what a folded shirt even looks like.
Right.
So it doesn't.
So it knows to differentiate between a folded piece of laundry and a non folded piece of laundry.
Then you've got the dexterity in the hands.
How strong do I need to grab this thing to pick it up versus something else?
And that brings us to the safety issue.
Imagine you have a baby. or a puppy in the home, AI already hallucinates.
It's gonna happen.
It's gonna happen with humanoid robots too.
So would you feel comfortable right now having a humanoid robot in the house with your infant child?
Probably not.
Because of those issues that we talked about.
They need to know how to code the robot so that it can perform safely all day, every day in the home, alongside other human beings.
But I am relatively more optimistic because remember around five, three years ago, when we talk about automobile that can drive themselves, the smart driving, pilot driving we were talking about.
Oh okay, so if we're using this technology, does it mean that every car on the road should be using the technology so they can communicate with each other, So they can make proper judgment and decisions on the road?
So one, well, real, unfortunately, accident happened.
We know how to priority or how to make sure the responsibility falls onto which party in the collision, in the car accident.
And when something happens, should we allow human to take over the control of the car or should, in really risky situations, should we allow the machine in the car, the computer in the car, to take over from human?
We had the discussion.
We had the conversation for months. quite many times.
But look at now.
Now we're looking at cars driving themselves.
We're having this pilot mode on the road.
They're shining with blue lights.
And They are safe and they are coexisting with human drivers.
And they have figured out how to handle this handover of controlling rights in the car in a way that we didn't imagine five years ago.
So I feel like some of the problems we're talking about now might disappear instead of being solved when new technology comes up.
Oh, wow.
One million percent.
You just asked when can I. Yes.
And I said not tomorrow.
Not tomorrow.
Probably five years or seven years.
I mean, you can order a humanoid robot and have one in your house tomorrow if you want to.
It's just not going to be able to do all your laundry and dishes and other things.
And I think the technological challenges might be solved. by unexpected ways.
That was my point.
And also I would love to welcome one to my home, maybe five years later, because they also have a lot of different skills, different specialties, that it is unfair to expect a human assistant in your home to grasp or to have all the ability.
For example, we see that in Xiangjiang, number one university in China, they have robot robots, humanoid robots equipped with deep seek model, and the robots can identify acupoints and perform acupuncture for elderly in nursing homes, and they can even provide emotional support.
They can talk to you because it has large language model built in them.
So If one day, when we welcome a humanoid robot instead of just a washing machine or a sweeping machine into our homes, they have a lot to offer.
Yeah, but now, I mean, you and I were talking yesterday about this.
They're kind of just like a...
I don't know.
I don't mean to be dismissive or condescending when I say this, but they're just kind of like a party trick now almost.
Yeah, exactly.
That's why I think, when you were saying that when we're looking back, like five or 10 years ago, cars have not been developed like this.
So that's, I think, is the trend or path a lot of these industries will go through and step by step.
Um, so currently, or maybe in the future, one to two years, the these kind of human or robot are mainly still focusing on being a companion or acting as a, something that you can just do, accompany you, and then uh, some experts are saying then, maybe in three to five years, these human or robot can enter commercial service scenarios, like acting as hotel service personnel who interact with people, and is projected to to do these and to be largely and scalably employed.
And then maybe in future 10 to 8 to 10 years, you can finally expect them to be in your home.
I know what you're talking about.
Earlier this year, CMG hosted several games and competition related to robots.
Robotic dog racing, robotic dog boxing, competition and also humanoid robots marathon.
And you see the clumsiness.
Yeah, the humanoid robot world games in August.
Yeah, there was the soccer match, there was the boxing match.
And the funny thing is when it's robots, or humanoid robots, you see, they struggle in unexpected situations.
Yes, for dogs or for human beings.
When you're boxing, the hard part is find the spot to punch and to find the best posture to defense.
But for the humanoid robots, when they're boxing or when the robotic dogs are boxing, The hard part is getting up from the floor.
You have to find out your balance again and something in that regard.
Just a couple of years away for some, though, 1X's availability, that's from the United States, right?
Their pre-orders are now, you can pre-order for $20,000 or you can lease it too, about $500 a month.
They're expecting delivery by 2026 at some point.
I don't know if that's early or later in the year.
Tesla, theirs is called Optimus, right?
They put their prototype out in 2022.
And Elon Musk said during this summer that Optimus version 3 could potentially become available in early 2026.
Really good examples on top.
Well, besides the...
Xiaopeng humanoid robot who's just a bit too humanoid.
And, that being said, we've discussed the potential problems and current challenges faced by the technology of humanoid robots.
How about the entire industry?
If we want them to actually be part of our life, what else are we looking to solve?
We actually just mentioned some of these.
All of these AI and human or robot, they need data.
So they need information.
They need database to actually build up the whole system and recognition of their, let's say human, like AI, brain.
Yeah, and human behavior.
Yes.
But how can we actually collect this kind of data and information?
And that is one of the biggest challenges this industry is facing when experts are saying that we still need a lot of time to actually train these robots.
And it's not only about coding, it's also about the Like the insert of a lot of this information.
When you say that, it makes me think that when I actually welcome a robot in my home, it will have to collect data from my family too.
And making sure that data, that information is secured is also quite important.
Yeah, encryption on scale.
That's a big deal, right?
But there is another example.
I don't think we mentioned this company yet.
Figure AI, their robot's called Figure 3.
And I was watching a video on it.
It folds laundry pretty well.
It loads the dishes pretty well or the laundry pretty well, but not really perfect.
So they need to get from it can kind of do it to it can do it better than human.
No problem.
Every single time.
And the CEO, his name escapes me.
Sorry.
I didn't write it down.
But anyway, he was talking about that and he said that data solves all problems.
And when you scale up the amount of data and the computing power being used to train a neural network, then it kind of shows improvements across the board.
So what they're doing now, that particular company, they're spending a lot of money.
They're collecting videos of people performing household tasks and then using that to train the robot so that they know how to behave in certain situations.
Same strategy when it comes to robotic chefs.
The chefs are also learning from the best chefs in the world and know how to cook the best dish.
I can have one of those.
Well, I think, whether the machines end up in factories, living rooms or rescue missions, one thing is for sure their story is no longer science fiction.
It's becoming part of ours.
