Video summary
Chinese robots are currently making headlines by shattering world records at the World Humanoid Robot Games in Beijing, where over 2,000 units competed in various sporting events. In a display of impressive hardware capabilities, one robot managed to run the 100-meter dash in 9.39 seconds, surpassing Usain Bolt's previous record of 9.58 seconds. Another machine cleared a height of 2.88 meters, exceeding the human high jump record, although the nature of this feat was questioned as it involved jumping vertically without clearing a bar. While there were also striking robotic dogs performing dressage and other unusual incidents, such as a robot crashing into a wall and catching fire, these events highlighted both the rapid advancements in physical mechanics and the current limitations in software control and safety protocols.
The primary purpose behind organizing these games is to showcase China's evolution from a manufacturer of cheap goods to a dominant force in future technologies like robotics and artificial intelligence. This initiative comes after a massive surge in investment and the establishment of numerous robotics companies across Chinese cities, driven by government support starting around 2023. The technology powering these machines has also evolved significantly; while early models relied on Large Language Models trained on text data, modern robots utilize Visual Language Models that process vast amounts of visual and spatial information. However, unlike language AI which was trained on the entire internet, robots still face a data scarcity challenge because there is no equivalent "internet" of visual human experiences, necessitating expensive data collection efforts where humans wear sensors or remotely control robots to gather training data for specific tasks like dishwashing.
A critical debate surrounding this technological revolution concerns why humanoid shapes are being prioritized over other designs and how different societies will handle the inevitable displacement of jobs. Founders often justify the humanoid form by arguing that since human environments are built for our bodies, robots should mimic them to navigate existing spaces effectively, though cynically, this choice is largely a reaction to Elon Musk's ambitious vision of general-purpose robots. Furthermore, there is a stark contrast in how China and the United States approach the social impact of automation; Chinese society appears more accepting of technological disruption due to a culture of "learned helplessness" regarding inevitable change, whereas Western nations face significant political resistance and public backlash against data centers and job losses. Ultimately, while the US relies on democratic processes to debate AI ownership and distribution, China utilizes top-down policies to preempt social unrest, raising the question of whether authoritarian efficiency or democratic deliberation will better manage the challenges of an automated future.
Read the full video transcript
Chinese robots are smashing world
records and on more than one occasion
smashing themselves. Um this is footage
from the world humanoid robot games.
It's an event in Beijing which has seen
2,000 humanoid robots compete in various
sporting competitions. Um this robot
beat Usain Bolt's 100-m world record. Um
Bolt did it in 9.58 seconds. A robot
managed it in 9.39 seconds. Um a robot
also beat the human high jump record. Um
so the record for a human man is 2.45 m.
The robot managed 2.88 m. Um it should
be noted it didn't have to clear a bar.
It just jumped up and down. So I'm not
actually sure it should count. But um
somewhat impressive nonetheless. Um
there are also giant robotic dogs.
Um they look frightening. I'm not sure
what they were doing in terms of the
sporting competition.
Uh but striking.
>> Dressage?
>> Um
something like that. Um dressage I
suppose dressage. Yeah, it's doing
dressage.
Then it would be a horse not a dog and I
saw it described as a dog. But I suppose
it could be anything. It looks a bit
like Katy Perry in the Super Bowl
halftime show. Did you see that where
she's Oh yeah. on the tiger for roar.
Um anyway, in favor of team human when
it comes to football, the evidence from
Beijing suggests we do still have a
while left as the dominant species at
the pinnacle of the game. They're not
very good um at football at all and they
do seem um
sort of very vulnerable to tackles. Um
and also actually I mean this is
probably the most viral footage from the
games.
>> It's like hurricanes.
>> Which is um another moment where robotic
athletes proved just as fragile
um as their fleshy brethren.
Uh so this is one of those very fast
robots who goes straight into the wall
um bounces back and sets on fire.
Um it's all pretty captivating footage.
Um and these are videos that have just
been all over my timeline this weekend.
But what's the point of it all? Well,
Chang Che is a writer focusing on China
and technology. Earlier this year, he
wrote a long read for the Guardian on
China's robotics revolution. Um, for
that piece, Chang Che talked to
executives from 11 robotics companies
across five Chinese cities. And when I
caught up with him earlier, I asked him
what China might be trying to prove with
the humanoid robot games.
>> I think I think that China's trying to
say that they are
you know, here. Um, they have been
working to be a technological power
since the 1990s when um, they started on
this market takeoff. And since then, you
know, they started with
um, you know, cheap goods like shoes and
toys. And now they are dominating in
some of the leading technologies of the
future like robots and AI
and semiconductors and all sorts of
different kinds of technologies. And um,
this is just an example of one
of the primary ways that China is trying
to showcase
the country is
an is now a technology technological
power as well as a uh, you know, a
military economic power.
>> And I suppose taking a step back, what
are these games proving? Because you
know, obviously these are humanoid
robots which are running much faster
than they've ever run before and jumping
higher. Um, and I mean, they're less
convincingly playing football. But I
mean, clearly there's some kind of
advance here. Um,
at the same time,
do we need a fast humanoid robot? What
what is being proved?
>> That's a good question. I mean, so just
to take a little bit of a step back, I
mean, these games
uh, are in its second year. So it's
still pretty new.
Um, what came before that was just an
explosion of robotics companies in
China. So around 2023, 2024, you saw um,
you know, the government started to talk
about robotics as this really important
um,
direction, this technological direction.
There was this explosion in investment
in China. Different cities and different
localities started to invest there in
their own robotics champions. So, you
had this already an enormous surplus of
companies and talent and investment and
capital
moving into this space. And I think what
happened was that the Beijing government
decided that they just wanted a way of
showcasing what was going on. And so,
they started to organize this event to
kind of bring all of these companies
together because they were already sort
of making robots for all sorts of these
applications. Let's just have them in
this really neat format where we can
actually start to compare how they're
doing and have a ton of foreign
journalists come and
and show them what China's capable of.
So, I think that that's a little bit of
a backdrop. In terms of just to answer
your question directly,
I think what you can see from the past 2
years, the major difference that I've
noticed is is in hardware.
So, if you look at the race and the
sprint, there was this 100-m dash where
the robot beat Usain Bolt. That's an
improvement in hardware. Those are
hardware problems that the Chinese
engineers have solved between last year
and this year. Now, if you notice that,
you know, the story of that sprint was
that the robot wasn't able to stop at
the end of the sprint and it ran into
the wall and it basically like
self-destructed.
That's an example of something that
that's a software problem. That's a
software problem because the robot
didn't recognize that it finished the
race and it didn't have the ability to
stop.
And so, there's a little bit of a
difference, I think, between software
and hardware. I think we can say that,
in terms of hardware, China's robots are
improving very quickly.
>> And and the software, I mean, to be
clear, here we're talking about LLMs and
artificial intelligence, are we? So, I
mean, that's what's sort of powering
Well, not obviously not not the
electricity that's powering them, but
sort of what's dictating what they do is
is an AI model which is inside or like a
few chips inside these robots.
>> So, the backdrop to
the software question is that
it is like why did we have why are
robots a big thing now rather than say
10 years ago? And the reason is because
of chat GPT.
The what happened after chat GPT was
that people realized that if you smash
enormous amounts of data with enormous
amounts of computational power
and this right now the most powerful
Nvidia chips
you can create some semblance of
intelligence and we've proved that with
language and now people are think that
they can prove this with robots which
means that instead of feeding robots
language you're feeding robots something
various kinds of visual data. Um so that
can include you know spatial data you
know there's there's a lot of like
something like a GoPro level kind of
data that's been fed into these models.
And so they're not called LLMs but
they're called VLMs
video language or visual language
models.
>> And that's because and and this is I
mean your article for the Guardian you
sort of had an interesting segment on
this because obviously for for LLMs or
for for the large language models
what they did was they just fed them all
the information that had ever been
written down so sort of Reddit Wikipedia
plus scanning in books.
When it comes to
a robot and what you're sort of dealing
with is audio visual data is a bit
harder because there isn't that isn't
all sort of documented somewhere in a
big file. So they have to employ actual
humans to either wear GoPros while
they're going about daily tasks or
control these robots remotely
but all the while sort of feeding back
to train the models. Could you talk
about that a bit please?
>> Yeah yeah. So the biggest difference So
the just to ask a question back at you
why does it feel like AI has been moving
and improving at like light speed over
the past few years, but robots haven't
really progressed. Like we haven't
actually seen robots, right? I mean if
if they're both AI, the robots should be
here. Like we should have robots next to
us now, right? And the biggest reason is
because there's just isn't the data.
Like the robots don't have something to
learn from, right? So we trained
Gemini, Claude, and and uh ChatGPT on
the internet. We literally gave it the
internet and was like, "Here, here's all
of human knowledge in text form. Learn
from it." But we don't have the inner We
don't have an internet worth of visual
data for a robot to do. It's There's
it's a little bit different with um
uh
autonomous driving. Like that's
something that is currently being
uh That's probably the closest example
that of like something that we've been
trying to feed it an enormous amount of
like car visual data. But in terms of
like, you know, for a robot to wash the
dishes,
we just don't have human
data of dishwashing. And And so what you
see now is a lot of companies are trying
to
um you know, hire workers to like wear
gadgets at home and like all these
sensors while they're doing household
chores. And part of that is because they
want to collect the data for that. So in
China, you're seeing this kind of
massive buildout of data collection
warehouses and facilities right now to
basically collect the data for this kind
of work.
>> All right, even more fundamental
question now. Why humanoids? Obviously
sort of, you know, a useful
machine can be many different shapes. Um
sort of evolution gave us two arms, two
legs, and a head. Somewhat arbitrary,
right? You know, for many tasks, you can
think of a a better
shaped sort of being.
Why are these companies sort of doubling
down on humanoid robots specifically.
>> Okay. Do you want my cynical answer or
do you want the answer that founders
typically will tell you?
>> Maybe both.
So, we can compare and contrast.
>> Okay.
Okay. So, the the what the way that the
founders will typically frame this
question is that if you look at the
human world, it's built for our bodies.
So, it just makes natural sense that you
build robots that are shaped in the form
of a human because then they can be able
they can enter
redesigning spaces, all the spaces that
humans can go.
Which I think makes a lot of sense. But,
if you look at the practicality of how
we got here, I think that it's actually
the cynical answer is that
China is very much uh
reacting to and kind of thinking through
the lens of the United States. And when
you talk to the founders and I asked
them this question to the founders in
China who are building robots, I'm like,
"Why are you guys building humanoids?"
And they were like,
"Cuz Elon Musk is doing it."
>> And why is Elon Musk doing it? I mean,
for the lols, maybe.
>> Well, I think that Elon Musk has like
this really grand vision where the
robots he's imagining that these robots
can are going to be able to do
everything in the future. Like, if you
talk to him and his prediction about
robots are doing it, he'll be like,
"Yeah, you know, I think he said at some
point like years ago like, 'Oh, it's
going to be able to babysit. It's going
to mow the lawn. It's going to be able
to go to a factory.'" His imagination is
that this robot is going to be able to
build Tesla cars and then come home with
you at night and take care of your
3-year-old child. Like, it's going to
it's going to be able to do everything,
which is
extremely ambitious, right? Like, you
can have you can imagine that a
different approach that's a lot more
modest would be something like building
designing a specific robot for each
specific kind of task, but that's not
what Elon Musk is known for.
>> Okay, I've got another broad question
for you. So,
the debate going on both in in the West
and China as far as I understand it is
how quickly can this technology replace
people's jobs? Um in the United States
the focus is more on white collar jobs
because it's these chatbots that are
sort of the cutting edge of AI. In China
much more focus on physical jobs and and
and robotics. Um which society do you
think will be better prepared for and
better able to deal with the challenge
of mass job displacement?
>> You know, I don't know if I can give you
uh a
an answer of which is better than the
other right now, but I can tell I can
sort of think through it out loud with
you.
So, I think that in China um
there is just a
general acceptance
both from the leadership and the
population that technology is coming and
it's going to
come whether you like it or not. So,
often times you you have this kind of
um
like learned helplessness in China where
it's like, "Okay, you know, the
technology is coming no matter what." Um
you know,
and so when people lose their jobs, it's
not something that can kind of create a
kind of political or social resistance.
Whereas in the United States that is
extremely powerful. I think it's so in
the UK as well. I think in general just
Western countries have a little bit more
of this culture of hey, we and you can
see this right now with the data center
put back backlash, right? So, this is
like a really big difference in the
United States and China. China just
builds data centers. They don't have
They don't have to deal with the public
resistance to that. Um Americans are
extremely pissed off that they're having
to share electricity uh with, you know,
these big tech companies coming in to
their neighborhoods and building massive
data centers. Um so, I think that what
what's going to happen in the next few
years is very very much very soon I
think is
you're just going to see a lot more
political resistance to this kind of
stuff and the roll out of robots. I I
these are things that I think anybody in
the United States has to think about.
Like, okay, we're going to build a
robot, it's going to replace these
workers. What's the political
resistance? What what what do I have to
do in order to soften that? That that's
not a problem that China has to deal
with.
So, I think
who is better positioned? I mean,
I don't know. I think I think China is
in a weird kind of dark cynical way
because it's just not a society that is
used to this kind of like labor revolt.
>> I suppose the the alternate argument as
to why sort of this this could end up
better for the masses in China than it
would in the United States is because
oligarchic power is a bit more of a big
deal in the United States. Obviously,
you know, it's not because China is more
democratic, but there is potentially a
stronger state that could redistribute
the gains of any job displacing
revolution. I don't know if you think
that's sort of overly optimistic about
sort of China's
>> Yeah, well, I I think that my my like
kind of sense of the way that the US and
China is operating is that they they are
obviously they're they're completely
different systems, but these kinds of
questions because they're both
being transformed so quickly with AI,
you see these questions arise um in
China and the United States, but how
they're going to react to them is going
to be different. So, in the United
States, obviously, there's a massive
question about distribution. Bernie
Sanders wrote like an op-ed about how,
you know, we should have like ownership
over AI. The whole country should have
ownership over AI. Um these debates are
being played out in public. I think in
China, it's you know, you hear it among
economists and you hear the debate uh
within sort of policy making circles.
Like, okay, you know, at some point the
public is going to be very dissatisfied
if all this money is being concentrated
by, you know, by capitalists and and by
our our tech companies. Is there a way
that we can find a way to distribute?
But the way that they will do this is
through is through some kind of top-down
policy. Whereas in in in the United
States, it's going to come through a
kind of democratic process. So I guess
you know, the question is is the
American democratic process going to get
there first or is the authoritarian
model where there's a sort of like
anticipating they're sort of like trying
to preempt a revolt from the public? Is
that going to beat the democratic
system? And at this point I don't know.
>> That was the journalist Chang Che
speaking to me earlier today. A really
fascinating interview. One of the most
interesting questions in the world right
now, isn't it? Which system is going to
deal better with the destruction of
jobs, the Chinese or the Americans?
My money is probably on the Chinese.
But obviously it could go either way.