The Truth About China's Generative AI Revolution Nobody Talks About with Grace Shao

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
China's approach is very pragmatic.
People have been saying, oh, Deep Sea did out of necessity.
There's obviously GPU constraint, a hardware
constraint in China, and that's not the thing they're hiding.
In many ways, the engineering genius and the engineering kind
of innovation is really what setDeep Sea apart.
And for sure it challenged the global narrative around to get
better AI, we need more GPS, we need more money.

(00:20):
It was just like chucking capital at it.
And then definitely was a different approach because the
capital ecosystem in China itself is also very different.
And a lot of people talk about this like prove to concept.
You have to prove your concept first in China to get even
money. So for a lot of these start-ups,
they really were not getting a lot of funding frankly before
BPC moment. And to your point again,

(00:42):
actually it's because no one really knew it was going to have
a strong ROI. So own the BET that had money
and understood the technology orbacking it.
Welcome to Analyse Asia, the premier podcast dedicated to
dissecting the pulse of business, technology and media
in Asia. I'm Bernard Liang and today we
are diving into the deep rise ofgenerative AI in China.

(01:05):
With me is Greg Shao, founder ofthe AI Premium newsletter, a
sharp observer of the China's tech scene and also a former
technology reporter with CNBCCGTN.
And also she has the done crisiscommunications even for some of
the major tech companies out there.
And today we're going to think about what are the deep insights

(01:26):
on the interplay of innovation and geopolitics in the China AI
ecosystem. So, Grace, welcome to the show.
Hi, Bernard. Thank you so much for having me
today. Yes.
And it's pretty interesting because you've been covering the
China taxing and the rise of allthe generative AI companies, but
I think it all dates back three years ago after ChatGPT has

(01:47):
rise. And then everybody was asking,
where's the China AI ChatGPT? Why did they miss that?
So before we get to that, I wantto start off, of course, the
your origin story. How did your career begin and
what led you to the world of journalism and eventually now
doing AI analysis and commentaryon tax scenes?
I don't usually advertise this, but I studied finance from

(02:08):
undergrad. So I actually started my career
in finance. So it kind of brings everything
together now where, you know, I'm a business analyst.
I think I don't position myself as journalists anymore.
And I provide analysis and insights on the happenings of a
lot of the big tech as well as start up AI system ecosystem in
China and how they're commercializing and monetizing
in China. So like I said, I studied

(02:30):
finance, but then when I actually joined a hedge fund at
the TMT Analyst as an intern wayback, I realized I just wasn't
that good with numbers, frankly.And I was definitely a people
person and I love storytelling. That was just, I would, I always
been very extroverted and love talking to all walks of life.
So I actually studied financial journalism and moved to Beijing.

(02:50):
My parents are from Beijing and I wanted to kind of understand
that their background, the history and the culture.
And I got my first kind of trialthere and it was great.
I actually worked for CGTN in Beijing, which is actually the
biggest broadcaster in China andallowed me to travel across
China, see the country from a complete different lens because
I was able to even travel to like third tier, 4th year

(03:12):
cities, rural areas and it really just made me fall in love
with this part of the world and I stayed.
So after that, like you mentioned, I did join CNBC in
Singapore, mainly focusing on the APAC business and tech
sector, but again with the Chinafocus, just given how relevant
and present it is for this part of the world.
A couple years ago, I started working for Alibaba.

(03:34):
So I was in their strategic communications department,
mostly managing their like corporate PR as well as Christ
management, etcetera, actually during the most tumultuous
years. And I joined a consultancy and
advised mostly tech firms. So you think companies you think
of are like PayPal, Lenovo, Kaishol and financial etcetera.
So and that's like my solid 10 years and those 10 years really

(03:58):
just kind of shaped my interestsand kind of where I'm at now.
And it brings I bring everythingfrom working in house and
corporate to journalism to analyzing companies from
investment perspective to my newsletter now.
So you have a very diverse background.
I think you have done the insurance and outs of being on
the CNBC side, the Alibaba side,and then now you're with your

(04:20):
current setup. So what are the biggest
takeaways you have taken from observing the tech sector,
specifically in China, from yourviewpoint?
Honestly everything just moves so fast and it's ever evolving.
I think just a person anecdote. Like during COVID I was mostly
in Hong Kong. I didn't go up to mainland for a
couple years. When I got there for the first

(04:41):
time, I think in three years, I was trying to order food.
And then I was asking the waitress being like, excuse me,
could I order food? She just looked at me like, you
bizarre human, why are you speaking to me?
Because at that point, everything was already digital.
You're supposed to just scan your QR code on your table,
order everything down. The only thing they were
supposed to do was that you bring the food to you.
And this is at a time when robots waitresses weren't even

(05:04):
implemented yet. So I think things are just so
fast, you know like digital payments to omnipres and like
every time I which is every couple months like I'm a stock
buying new technology or some kind of evolution in digital
ecosystem. Or my friend Patrick Maggie, the
current author of Apple in China, said that is China speed.
Yeah. So what inspired you to start up

(05:27):
AI program and how does your work now help to bridge the
understanding of AI in China with the rest of the world?
Yeah, I think like I kind of mentioned, I've been on the
different sides and, you know, I've just watched and worked in
these tech companies for so longand I felt like there was so
much that was being overlooked. And given that actually since

(05:47):
2018, the trade war broke out, right, Like a lot of stories, I
find that we're kind of hijackedby the high level geopolitical
narrative. And for someone who's so
passionate about telling the stories of the businesses of the
innovation, and for someone on apersonal level, I think I kind
of straddled the two worlds of the West and the east of China

(06:07):
and the USI just felt like therewas more I could do to help
bridge that knowledge gap. That's why I started a a prom
and really just at first writingabout what interested me and
what interested the people around me mostly were academics
and investors, and it kind of just took off on its own.
I think same here. I think even though I'm
Singaporean but culturally I'm Chinese, and then trying to

(06:28):
think about how do we breach theunderstanding.
I think there's a lot of misunderstanding, but let's pass
away the politics, but just focus on the technology and what
is really going on in China. Which also comes before I get to
the main subject of the I still have one more little question.
So what would be the key lessonsfrom your career journey that
you can share with my audience? Oh man, I mean, I'm still in a

(06:51):
very humble beginning of starting AI Pro.
I'm like, you know, Bernardi just saw I post on LinkedIn
yesterday. I've just hit my one year mark
with this, right? So I think there's still so much
for me to learn. But I think what's been so
pleasantly surprising on this journey and compared to working
for a lot of big platforms or big companies, it's just that
when you keep an open mind and frankly be a bit shameless or,

(07:13):
you know, if you have thick skin, say yes to everything out
there, a lot of surprises come to you.
I think I my advice would be just don't be scared to try new
things. Say yes to everything.
Well, not everything, but at least say yes to when people
invite you to coffee chats, to, you know, to conversation.
You just never know what you canlearn or what will lead to.
And even, you know, our conversation now is because you

(07:34):
dropped a message to me on LinkedIn and I frankly didn't
know too much about your podcastat first, but I found this world
of interesting interviews and, you know, extremely high caliber
people that you've talked to andI've learned so much from your
interview. So I would say definitely just
be open minded and say yes. Thank you.
And that's a great lesson to to remember.

(07:55):
So let's get to the main subjectof the I want to talk about the
rise of generative AI in China. Can you help me paint a picture?
What does the generative AI landscape in China look like
right now? That's such a big question.
Yeah, We take that different because we are just right in the
week before the big AI week in Shanghai, right?
Yeah, yeah, yeah, yeah. So I mean, just maybe we start

(08:17):
off, maybe think about the largelanguage models, Let's talk
about the foundation models. Everybody hears about deep sea,
but maybe let's help me to dive a little bit with that picture
with deep sea, the foundation model sites, not just deep sea
alone, Yeah. Yeah, and I think we can kind of
talk about the different playersthat are involved in the LLM
space. So first obviously there's the
start-ups like you mentioned, Deep sea and Moon shot came out

(08:38):
with Kimmy to K2 last two weeks ago that really kind of shook
the world again. There's other LLM player
start-ups, especially the original 4 Tigers.
These are by Chuan Jupu, Mini Max and a Moon shot.
And then we have the big tech players similar to how the US is
set up. You know, we have like the bats.
So in this case, usually it's really referring to Bite Dance,

(09:00):
Alibaba and 10 Cent. Oh, Baidu also not involved now.
It's involved. It's of all, I just think it's
at this point not as you know Frontier compared to others in
terms of their models and their kind of applications.
So first we kind of see the ecosystem like that and then we
have obviously the also the smaller research labs that are
still kind of Incognito mode anda lot of them spinning out of
universities that you meet here and there, but they're still

(09:22):
not, you know, on the pick global map, right?
So you kind of see that. So in many ways it's similar to
us. You have the startups, you have
the big tech, but I think a really big difference is that
actually out of the service, aside from deep sea, almost all
of them have some kind of financial or partnership, like
financial tie or partnership with one of the Bats.
A lot of them been, you know, financially backed by Alibaba or

(09:45):
Tencent, who are the more aggressive investors and
incubators in the Internet spaceas well as AI space since 2023.
So there's a lot of that. And then I think if you really
look at the practicality of likehow people kind of see how it's
different from the US, if you have to compare that way, is
that Chinese people fundamentally are very
pragmatic, I think. And I think, you know, when you

(10:06):
talk about these LLM research labs or whatnot, like there
seems to be a more like a biggerpush on them to try to
commercialize or go to market quicker.
You know, the idea of AGI is obviously like the Holy Grail
for everyone in this space rightnow.
But I think in some ways when I speak to people practicing in

(10:29):
the space in like Silicon Valleyversus China, is that the focus
of, you know, the philosophical,you know, approach AGI in the US
is a much bigger driver versus in China.
It's like, how do we come up best model?
How do we come up with most efficient?
How do we come up with the most cost effective one?
And then how do we commercialize?
So there's definitely that pragmatic approach.

(10:50):
I think that's a very high level.
And and then I think another fewthings you can kind of touch on
is like just like the open source and open weight kind of
approach and I know you want to talk about that.
Yeah. No, So I'm quite curious, right.
So like for example, the objective looks very different.
It's not as in the US when they talk about winning the AI race,

(11:11):
it's not just about like gettinginto AJI.
We don't want China to reach us,you know, because you know, they
might be able to beat us, but there wasn't that kind of race.
There's a lot in in China is actually specifically getting
the AI out, getting into AGI. But how do we commercialize this
AGI? How do we get into the hands of
the common people? How do we make different

(11:33):
applications? But maybe I will flip the
question a little bit. Since we talked first the
foundation models, maybe how about the applications layer?
I mean, in the US, you can thinkof code development platforms
like Cursor Cognition Labs, which the guys behind Devon or
Windsurf or maybe some of the other ones that like they have
specifically a little bit more I'm trying to just record like

(11:53):
Lovable Bolt, which create websites, gamma presentations.
How is it like in China from theapplications layer for
generative AI? Yeah, I think we can touch on
the US versus China narrative later, but honestly, I think
it's such an unproductive likewise.
Yeah. Correct.
I know I'm like feel quite strongly about that, but I.

(12:14):
Think it's that. I think from the application and
you know, in, in some cases it'sso similar, in some cases it's
not right. So we take a step back and go
back to like just looking at theplayers.
So on the BAT level, Alibaba itself is taking a much more
like Microsoft like kind of approach where, you know, it has
cloud infrastructure as like thecore actually, if that's like

(12:35):
what they're trying to sell, youknow, at the end of the day,
Alibaba Cloud has seen like double digit growth in the over
the last few quarters of earnings reports.
I think, you know, they're trying to sell to enterprise.
They're trying to sell, you know, their cloud services as
well as their enterprise software, like, you know, the
add-ons, the AI, you know, AI enhanced efficiency to vendors
and stores and merchants. So that's a very enterprise

(12:56):
focused approach. Their consumer side is actually
not like doing the maze. It's fine.
They're I don't think they're putting that much resource on
it, to be honest. Obviously they have the Quinn
series, the open source slash open weight model that's really,
you know, taking up this like, you know, like the limelight as
well as you really been right behind the Hugging face for like

(13:19):
rankings. So that's Alibaba.
I think in terms of Biden, Tencent, they're taking a very
different approach. So we can kind of break it down
a little bit. Biden has really been diligently
trying to still compete on the frontier model and they were
kind of low key for a while and pretty DeepSeek.
If anything, they actually were kind of the best chatbot or like

(13:42):
pre Kimmy, they were the best chatbot, but they had a, they
have an app called Dobao and they hit I believe more than 400
Mau after aggressive, aggressivemarketing.
The issue with it is they, I think there's a lack of
functional adjacency to their core product.
What I mean by that is if you think about it like you're

(14:03):
slapping on Dolian, which is theequivalent of TikTok, right?
You're not exactly interacting and saying, hey, chatbot, like,
tell me this, do this, you know,like it's more just like a pure,
yeah. It's more like a video service
and it doesn't rely a lot of this kind of social
interactions. Exactly.
It's more like you're consuming,right, You're not interacting.

(14:24):
So they definitely also pushed out various AI like enhancement
products within DOI Yin to kind of help with like like the algo
with their e-commerce etcetera. But in terms of Dhoba, the
chatbot itself did OK, and that's a aggressive pushing.
Now Tencent took a complete different approach, which was
really interesting. And I do think that was a pretty

(14:45):
pivotal moment for China's kind of consumer application.
And what happened was they had their own, they have their own
LLM. But Tencent notoriously is known
to work in silo. Like anyone who knows China
ecosystem is like they're so competitive internally, like
each team works on their own things, right?
So basically WeChat team just decided to integrate DeepSeek

(15:07):
into their chat bot and then they just skipped over young
bot, which is their LLM. And it's pretty crazy because
what happened was like it did work in their favor because it
goes back to they do have the functional Jason and they do
have the walled garden mode. Because what I mean by that is
if you think about a WeChat, I mean, it's obviously more than a
WhatsApp, but when you're chatting with a friend on

(15:28):
WhatsApp, it's essentially a chat box.
And it's so natural for you to just open up then a chat box
with your ChatGPT like bot to ask questions, use a search
function to do it, whatever, right?
And on top of that, people in the West often don't understand
or realize that WeChat is not just a social media app.
It's literally a content platform.

(15:50):
So people have run blogs, post videos like it's this crazy
place, it's kind of like YouTubeplus, like Medium plus subset
everything plus all together in one.
You have all this content available and this data
available to Tencent that other people can't access.
Because when you like Baidu search stuff on WeChat, it
doesn't come out really. You have to search within

(16:11):
WeChat. So they have all this huge
advantage. So immediately they were able to
tap into 1.2 billion users or something crazy like that.
And it it like, you know, it, itwas kind of game over on the
consumer end, right, In that sense.
So they've done really well, I think in terms of consumer
application end. But for sure they're not like
leading in frontier model development compared to Alibaba

(16:33):
or Bytedance, because Bytedance just recently came out with
another series of dough ball LMSthat's very focused on text to
image and text to video. That's right.
If you look at it, it's like whohas better video and image like
data than by dance right? Can I ask you if am I
interpreted this correctly? But by dance strength is in

(16:54):
video, audio and of course not just though in itself but also
cap cut. I think a lot of people
underestimate how much content creators are using cap cut from
like some taking a long form content, cut it down into reels
and post it into TikTok and all that kind of parts of it.
But it seems to me that a bike dance is actually specializing
towards the multimodal large language models where it's much

(17:17):
better in audio or video. Am I getting that kind of
specialization? And then somehow text wise
they're not so good as compared to say your deep sea, your Quan.
I think there seems to be also very little understanding from
the rest of the world on the Quan models.
I think deep sea heavily distilled not just from CLOD or
Open AI but also from the CRAN models and that's why their

(17:40):
accuracy is so high. Yeah, I think Bindance seems
like they made a pivot quite recently with the recent like
April launch. But I think before that they
were still trying to compete on with Dhoba on the tech side.
They think, you know, so Qui Show is like another major video
short video player in China. It's not as prevalent in like
North America, but it's actuallyquite big in LATAM and some of

(18:02):
the EMEA markets. But anyway, my point on Qui Show
is that they actually came out with some amazing products
called Cling that were like supposedly comparable to mid
journey. I think that was maybe a
triggering moment for bike dance.
And he did see them kind of go abit low key for a few months and
then push out this amazing series of, like you said, image

(18:22):
and video multi model model likeformat.
Yeah. So so I think, you know, there's
definitely interest playing intotheir own strength and they
obviously have different leverages and I think unique
walled garden databases because I know the day it's a data, it's
a data competition as well. And then they also have unique
distribution. So I think if we take a step out

(18:43):
again to talk about the kind of start up versus the big tech, I
think right now, if you're, whatyou're seeing is that the
start-ups in China are struggling with their
distribution and reach because you have to actually onboard new
users on the consumer end versus, you know, a lot of these
big tech are just have kind of capture this massive ecosystem.
And to kind of give you some context, as I read that, like,

(19:05):
you know, for a average Chinese consumer, you open less than 10
applications a month because youjust you can use like literally
one hourly pay or WeChat to do basically everything in your
life. Like you can chat, you can pay
your bills, you can get transportation, you can order
food, you can do whatever. But in the West, most like or at
least in let's say North Americaand Europe, you're looking at

(19:26):
anywhere upwards of 30 apps to finish your day-to-day.
Like utility bill is 1 app, right?
Like credit card bill is 1 app. Like this bill is 1 app.
And like in that sense, like your, your ecosystem is like
completely already owned by likeone of these bats in China.
So you just have to capture yourvalue in that.
Yeah, and that comes to this point that I always help to the

(19:47):
my Western counterparts when they asked me about how the
Super ad works. And I tell them there's no
chance for US to have a super app because your kind of company
culture is context towards building this one thing and one
thing very well, very perfect. Like your Google search is
perfect, your Facebook is a social network.

(20:07):
And when you start entrenching everyone's territory.
It is uphill battle because it is like the most
state-of-the-art. Whereas in China you don't have
that situation a lot. OK, there are some things they
innovative on top, they leapfrogon top and then it's very hard
to copy. But they also have basically
integrated the whole full stack.And that's why they can have a
super app and that's why also Southeast Asia would also have a

(20:29):
super app. That's why Europe will have a
super app, but you will never see that happening in the US.
Yeah, but coming back, right? You talk about consumer
behaviour, you talk about the enterprise needs.
Maybe one question that always comes out is when Cha GBT first
launched, right? Many question that why didn't
China had launched it first? I mean, the transformer
algorithm was available. So deep sea emerged and

(20:51):
basically put all the counterparts in this new AI
race. What lessons can you observe?
Can we learn from just observingthis deep sea involvement change
in the China AI ecosystem? I mean, I don't I don't know how
why didn't really says model before open AI, right?
I wish I knew, but I think maybethere was just like legitimately

(21:14):
not ready yet. But I think in terms of this
moment, a really interesting observation and feedback I've
heard from a lot of people in the tech industry in China right
now is that it just like helped China Internet revitalize.
Because we all know 2022 to 2024, essentially China Internet
kind of hit a bit of a like a rough patch, right?
Like the darlings were all kind of hit, like all the ADR

(21:37):
dropped. I think, you know, a lot of it
was caused by domestic regulatory reasons.
There was the anti monopoly probe.
There was a lot of like data security concerns.
And then like DD also had to do like face delisting and
everything, right? So we know all these different
kind of pots of crises per SE, but all that accumulated
together meant essentially the global capital world.

(22:00):
Like was really concerned about China Internet.
And I think kind of in a way, Chinese tech companies went into
a panic mode. And I think everyone kind of
went back into their shells a little bit.
No one really said much from a peer PR perspective.
People just stopped talking, right?
They're like, OK, it's better tonot stick my head up and stick
my head up and get hit. So there was a bit of that.

(22:21):
And I think for sure there was big blocking energy.
And then domestically after COVID, we also know that there
was a lot of actually cuts just like in the West, but there were
like headcount cuts. There was a lot of talent
retention issues. There was a lot of salary caps
like just all of these issues come out of that didn't feed
into a very positive energy for the Chinese AI tech sector.

(22:44):
Now DFC definitely completely changed that narrative.
I think whether it's from the media or it's from like
companies themselves, they said it actually allowed the world to
reshift their focus. It's from China bad or China
Internet bad to finally go, oh, innovation, like business cost
efficiency, you know, and like there's all the things that were

(23:04):
kind of overlooked for 2-3 yearswere finally kind of brought to
the front and center again and companies were able to kind of
openly talk about what they've been developing.
So like for example, like Damwa Academy, that's what Alibaba
like they are in charge of all the like machine learning and LM
research and etcetera. They've been doing this for a
while, right? Like it's not like like deep sea

(23:25):
came out and the queen just suddenly like churned out like
it's team's been around for a while.
But I think riding off of like deep seeks kind of positive, I
guess, PR vibe already. Yeah, they were able to kind of
really say, hey, look, we've actually been doing this as
well. Look, a Quinn, like I think it
definitely was a pivotal moment.And when I speak to, you know,
senior executives at these tech companies, they say, look, we

(23:47):
were so relieved because all of a sudden the media was like the
Wall Street. They were able to talk about us
as a real business again, versusjust look at us as a weather
wing for China's geopolitical issues or China's like consumers
lack, lack lackluster consumer consumer consumption.
You know, that's what kind of that pivoted moment.

(24:08):
It's actually very interesting to also examine because I'm
technically trained. So I still have a couple of
friends in the West, in Europe, China and the US.
And then we are all in medical forums.
And when the Deep Sea paper first came out, even in late
November, people were talking about it.
And I can share the other end ofthat was somebody who actually
has already proposed things likecertain methods that Deep Sea

(24:30):
were using and say maybe we should try to do this at scale.
But usually an American counterpart will just come and
say, well, you don't really needto try these methods because we
just need the video chips. That's all.
Just throw money at the problem.And then it irritates everybody
and then now everybody just got like a shot in the face and say
what you really can do. There's like no people told you.

(24:50):
So it's just kind of amazing that one of the things that Deep
Sea has really done is showing that you can make this cheaper
to do. It's not really belongs to
anyone country and there's a real horse race now.
But I want to also ask, like forexample, after Deep Sea, you
start to see Kimi and Manus AI. Even yesterday Alibaba just put
out a new crime model and it wason the top of the leaderboard

(25:12):
Hugging Face and I think starting to get a lot of
attention. Maybe what sets them apart in
the China's AI ecosystem? Are they actually similar
players? I heard of a company called If I
totally butchered it, it's U. Where is it?
Another one that's pretty well known.
So what are the ones that we aremissing and what are the ones
that are those ones that are gathering attention are really

(25:33):
one step ahead than most of their rivals within China
itself? Yeah, I think first addressing
your first bit of the kind of commentary about what DeepSeek
really meant was the part where like, you know, I think to your
point and to what I said earlier, China's approach is
very pragmatic. I think people have been saying,
oh, like, you know, Deep Sea didout of necessity.

(25:53):
There's obviously GPU constraint, a hardware
constraint in China, and that's something they're hiding.
In many ways. The engineering genius and the
engineering kind of innovation is really what set Deep Sea
apart. And for sure, it challenged the
global narrative around togetherthat better AI, we need more
GPS, we need more money. It was just like chucking
capital at it. And then definitely was a
different approach because the capital ecosystem in China

(26:15):
itself is also very different. A lot of people talk about this
like proved to concept. You have to prove your concept
first in China to get even money.
So for a lot of these startups, they really were not getting a
lot of funding frankly before deep sea moment.
And to your point again, you know, actually it's because no
one really knew it was going to have a strong ROI.

(26:36):
So own the bets that had money and understood the technology
were backing it. And that's also why that the
Bats were the first round of investors and incubators in this
space. So that's kind of to your point
there. I think to your second bit about
how Manus and Kimmy etcetera are, they're quite different.
How I see it'll be like they're I would separate.

(26:57):
I would say let's talk about these Seq and Kimmy first and
then Mana separately. I think they're very different.
So I think the deep sea moon shot situation is that like it's
very interesting because it they're not anomalies.
You're realizing that it's not like one off like, you know,
high fly, which is like this super genius that just found a
way to make more cost efficient frontier LLMS.

(27:19):
It's actually that the other research shops are keep catching
up. So I think that one big
difference comparing B to a lot of leading US startups right now
in this space, or if you want tocall them startups or not, but
like say compared to the open AIand thenthropics is that they
fully embrace open weight. So a lot of people call them
open source and you know, you'remore technical than I am.
But in the sense of technologiesof essentially open weight

(27:41):
models, I think there's been a lot of discussion around this
and there's a pragmatic reason, a philosophy philosophically, I
think honestly, you know, the Oliver has come out, he's a
media shy person, but he has come out to say he wants to
attract the best talent. It's the best way to retain top
talent. It's a perfect way to put the
most frontier technology available in China in the hands

(28:04):
of other innovators. He wants it to be massively
adopted and he wants to put Chinese innovators at the front
of the global innovation lead. Essentially.
There's a very strong, I think patriotic component to that.
And then I think there's also philosophical choice where he
just believes that, you know, and this is long debated in open
source versus closed source, like communities even in

(28:26):
technology beyond AI. Is that like, is it better to
share the technology and for more people to innovate on top
of it, right, and to be scrutinized and, you know, and
to adopt? Or is it better to kind of have
your proprietor information on data or everything close up?
So I think for him, it's a very personal philosophical choice to
be open source. There's that.

(28:47):
I think there's another thing which I found interesting is
that a lot of the founders now in the AI space are actually
quite young. They're born in the 80s or 90s.
I think they grew up in a time frame where China really was
rising economically and their soft power.
And there's a lot of culture social aspects to this too.
So I think people were like, if I put this out there, I can

(29:10):
prove that we're not copying. We are actually innovate and
that kind of like you are just acopycatter like mentality was
like there's a chip on the shoulder, right?
Because the generation before them were being called
copycatters and they're like, wewant to prove ourselves that we
are actually innovators. So, so there's always that like
psychological, cultural aspect too.
And then last but not least, which is pragmatic reason is it

(29:31):
goes back to like distribution, right?
So for them, it's like foundation models are becoming
commodities. How they view it is and value
will be accrued in the products and services built on top of
them. So open sourcing essentially in
that sense is a strategic decision.
So you open source it, you basically utilize and leverage
other people's distribution, create moats in that sense, then

(29:52):
you can have a really great model, but everyone will have to
use your model eventually if youjust could continue to lead in
their frontier model. So there there's a few different
reasons I think why they came out the way that they did.
And it's definitely been a bit different from how the American
research labs are approaching frontier models in terms of
Manus, I feel quite differently about.

(30:13):
So I actually was lucky to get access cold quite early on and I
tried it out. It's very cool, you know, like
everything was moving super fast.
Screws are going crazy and clearly from a non-technical
person, but you know, at the endof the day, I think what it was
the product itself, honestly, from my personal usage and
friends around me for research and simple task was not that

(30:36):
different from Open AI, you know, like O 4 or anything or
any of these kind of like leading agentic products.
I think what they did really interestingly is the PR
actually, I think they definitely first of all, they
had a founder that was very not media shy.
He was very eloquent in speakingin English and he did a very

(30:56):
strong PR push. This was very different from
majority of the other start up founders where a lot of them,
you know, are like nerdy, like researchers.
They get really media shy and they're not coming out and
talking about their own products.
They didn't really go found or direct in that sense.
And you know, people don't know about them as much.
So first of all, they did a veryfrom consumer marketing and PR

(31:16):
#2 they played into the scarcitykind of strategy.
And like I said, for people who know like they really weren't
giving out the access codes evento some of the leading like top
tier media and globally. People were begging for them.
They thought it was so exclusive, but really like it
was like a it's a genius marketing strategy, right?
Like so I think they definitely rode that hype a little bit in

(31:39):
the beginning. Not saying that part is not
good. I'm just saying I don't think it
was anything significant, betterthan other players.
And then I think another big thing that they've done
definitely now is that they've actually moved out to Singapore.
I don't. Know if yes I know they're in
town. I would like to meet them and
interview them. Yeah.
If you ever get, if you ever talk to them, please let them
know that there's a podcaster who actually covers a lot Asia.

(32:02):
I'm just joking. Yeah, but I want to understand a
little bit. Like I know madness also ran on
cloth 4.0 as well now, because it, the way how it executes its
agent take workflows, because I look into all the like the
computer lines that they were turning on.
And I always have this when I teach a class, when I actually
showcase Manus to the business leaders.
What I always like to do is I put on my phone with the Manus

(32:24):
app on and I have the task that's running.
I told this Class A let's just wait for a while, But I actually
had my, my mobile phone on just to see what the app was actually
doing. And everybody said, Oh, here are
the results. And everybody will have this
like prestige moment, wow kind of thing.
But how do they why, why do theydecided to move out of China?
I think that there's is probablyvery competitive within China,

(32:46):
but I think even they are actually quite in the lead in
terms of agentic AI. Yeah, I think there is a few
reasons. One is, you know, we kind of
touched on it earlier, but therewas a huge pull out of US
capital in China. So prior to 2022 I would say
like a lot of US private equity firms and venture capital firms

(33:08):
had like very large presence in China.
You know you, the Sequoias and the Orbic pinkest of the worlds
were all there. Unfortunately, I think due to
geopolitical pressure as well asjust, you know, the sentiment
back then, a lot of firms decided to pull out.
There was that aspect where I think Manus definitely decided
to take on US capital. So they took on benchmark,

(33:30):
actually bench, I believe they were benchmarks first China
investment. It was quite a big story.
Beyond that, obviously I think you and I are also know that
there is actually American government mandates that don't
allow or I guess not mandates, but restrictions that don't
allow U.S. investment institutions to invest in AI,
robotics and what are called a sensitive industry, so

(33:52):
semiconductors etcetera. So I think in that sense there
was definitely some kind of probably you know discussion in
the back end that made them haveto leave.
And then I think on a more like more company level it's they
really just took on the Singapore strategy.
So what I mean by that is we've seen a lot of companies over the

(34:13):
last couple years that are Chinese founded or majority
backed by either Chinese capitalor researchers or Thailand or
whatever. They have a Chinese heritage.
Let's put it this way, that moved the headquarters to
Singapore or set up shop in Singapore and their predominant
market are actually not is not China is the West.
So the reason they did that is first, obviously to kind of

(34:35):
frankly go under the radar of sensitivity checks, whatnot.
But another thing that's just a pure PRIR thing, right, For
raising capital. It's easier to say you're
Singaporean versus Chinese if you're raising capital from the
US for our PR approach. It's easier to also tell, you
know, consumers and stakeholdersthat you're actually
Singaporean. So a lot of them are doing this
and what even PR professions have called it.

(34:57):
It's called the China shedding strategy.
They try to lean away from Chinaand Singapore is kind of the
perfect place. I mean, you know it better than
me, but I lived in Singapore briefly for a year.
But you know, look, it's the APEC hub for almost all big
check and think of in the world.So everything from bite dense
Alibaba to like Meta and Netflixand you know, you know, like

(35:17):
Google. So essentially it has all the
talent in the world available, right?
It has domestic talent, expat talent.
And it has also a really strong advantage, which people again in
the West sometimes overlook is that like a lot of talent are
bilingual in Mandarin and English.
These are their native languages.
And it just makes it so much easier for you to have Singapore

(35:39):
employees to have talk to PM onesitting in Shanghai or Beijing
in their mother tongue and vice versa.
You know, you can have someone who speaks full in English to
talk to their maybe BD partners or their stakeholders in the
West in English. So there's a lot of reasons for
that and I'll tell them that Singaporean government has been
extremely welcoming and just like really trying to track

(35:59):
better talent and more companies.
And so I think there's a lot of reasons why Singapore has
benefited from this and become avery vital hub in Asia for tech
and AI talent. And remember that Bill Gurley,
who is the former Benchmark partner now, I think he's the LP
and came out in his own podcast with with in the BG 2 podcast

(36:19):
where he said something along the lines that the menace is not
in China. All the servers are outside
China, from Singapore, Dubai, and he was justifying the reason
why they're not considered Chinese.
And Benchmark shouldn't be drawninto conversations like that.
I think that whole rationalization is interesting
for me. And of course there's a lot of
rumors of some of the by dance founders living in Singapore.

(36:40):
I probably know where they do, but there's a question for
another day. OK, let's talk about the China
AIC now. Are there any lesser known but
promising companies or innovations within this whole
generative AI or engine AI spacethat we should be paying
attention to? But they are very natively
Chinese like. I will give one example from the

(37:01):
hardware space, this Unitary Robotics I when I talk to my
Western counter bus, I think they barely know what company is
this and yet it's one of the most talked about companies in
China. Yeah, I think Unitary is a
perfect example, but actually they're there's I think they're
selling more than like 50% of global humanoid robots right
now. They're if you know, you know,

(37:22):
kind of company. Yeah.
So actually on on this topic, I think we're noticing that like
Western media per SE are pickingup on the humanoid robot.
Sorry, but I think China's always had a very strong lead in
this. There's a few reasons for this,
right? Like obviously just having the
manufacturing supply chain all here for the last like 30-40

(37:42):
years. I think in China right now, you
know, for a lot of human or robots, if you break it down,
it's still the rechargeable batteries and lighter sensors.
It's the motives, it's the, it'sthe, it's like all this little
pieces that you basically had and your home appliances and the
DJI drones and the EVs, etcetera, right?
And all of a sudden, because of the type of humanoid robots

(38:04):
globally, you know, there's a lot of startups kind of trying
to lean in on this and they're trying to pull together all
these moving parts. I was actually just talking to
someone that works in supply chain yesterday was like, I
really want to get deeper into understanding the supply chain
of robotics. But right now I feel like
there's no major player that's talking about it yet because
they're so fragmented actually. But for sure like your point

(38:27):
unitary robotics like they're coming out with human or robots,
but you get robots like these old dog as.
Well. Yeah, and a lot of them are
being sold for what, third or fourth of the price of a Boston
Dynamics or a figure AI. And essentially what I think
right now is a lot of it, honestly, it is, and I would say

(38:47):
gimmicky, but it's still a very early stage.
And this is from talking to people in industry.
And to your question actually, like UB Tech is a really big one
that implementing into factories.
They have a partnership with Zeely, the car maker.
And then, you know, we have GAO BA that's really big.
You know, there's a few other players.
I was at Beyond Conference in Macau just a couple of months

(39:08):
ago, which is a big AI robotic conference.
And there were at least like 20-30 different companies doing
this, exhibiting their robots. They're somewhat the same.
The reason I say there are so early stage is because the EB
tech executive was saying, actually the reality right now
is that we're still we still just don't have enough 3D dinga.

(39:28):
And for these robots say five years ago, a lot of their
movements are pre trained right and pre programmed.
So they can only really do, say lift glass up, right?
That and they put it down, lift it up, put it down.
They have like these single tasks or maybe Max like 3 tasks
they can do to really implement the software into the hardware,

(39:49):
integrate that. Right now it's really
challenging because LLM data is not the same kind of like LLMS
are not the same of the kind of software or model that we need
for a robot. And to create this kind of 3D,
even 40 data is extremely difficult.
Essentially, you need some narrow training and you need to
collect in the real world. But naturally we haven't been in

(40:12):
a stage in our life that there'sso much 40 data collected.
So what all they can do right now in terms of software and
hardware integration and a very sophisticated level the EB tech
guy was saying is that they can actually become companion bots
or tutors. That is the best use case for
consumers right now because actually, if you connect like a

(40:33):
model into a robot and you have them sit beside a child, help
them with like the history classor like, you know, math
tutoring, they can actually pullon the knowledge in their
brains. But in terms of you, you want
them to really like, go into your homes and like, fold the
bed sheets and go upstairs and mop the floors.
This requires way too much like data that they just don't have
right now to make them sophisticated and seamless to do

(40:55):
a series of like actions. So there's that.
But I think it's definitely a space that we can continue to
watch. And I think China definitely in
that sense have an advantage just given how the whole supply
chains in China. And even Jensen Huang, when he
came to Hong Kong last November,he was doing a Firestar chat at
HEUST. He said that, you know, some of

(41:15):
the best talent in hardware, software integration engineering
are all in the GBA area, which is called the Greater Bay Area
and that's Shenzhen, Guangdong and Hong Kong.
It's because DJI is out here. Sense Times actually $0.10 out
here, Huawei is out here. So if you think about some of
the best hardware companies are all in this area.
On top of that, a lot of the manufacturing like the factories

(41:35):
of these spaces are all in this area.
And then you have a lot of very of like, you know, sophisticated
and leading academic institutions.
So we can anticipate more there.And then in terms of other
verticals, I think it's still quite fragmented given that, you
know, we're still in a very early stage of Gen.
AI adoption. But I think again, to the point
of China being very pragmatic iswhat you're seeing is a lot of

(41:56):
startups just saying, hey, you know, I'm not going to try to
compete on LMS. I'm going to compete in vertical
integration. So you have a lot of healthcare
startups, education startups. One company I did speak to
recently was quite interesting. They do like they kind of like a
mid journey, but actually for creative.
So like they charge very high subscription fees, but they're

(42:18):
extremely sophisticated in theirvideo production that you can
actually just completely cut companies ad costs.
So like they would sell to be sothere's these kind of companies,
there's companion bots that's extremely popular in Asia right
now. So you can think about like fake
boyfriends, fake girlfriends, which again, borderline like,
you know, you need regulatory kind of control around that,

(42:38):
especially for with the kids. And then, you know, you have all
these different like legal apps,you know, etcetera.
So, so, so I think there's definitely a lot to see
anticipate, but I think like youmentioned robotic hardware side,
it's something that China potential will come up on top.
So what is the one thing you've know more about the China AI

(43:01):
Taxi now that very few people do?
I don't know. I think everyone knows what I
know of any I'm. Sure you know something that
everyone knows, but from your observation, like for example
one of the things that maybe people don't realize, I mean
robotics is 1 and any other interesting ones.
Yeah, I think, I think it's two things.

(43:21):
One is the small things in the like in the verticals is that
actually healthcare app, healthcare AI and aggregate
check is extremely advanced and it's something that the a
country as a whole top down is really encouraging as well.
So, you know, people adopting AIinto agricultural, you know,
uses like, you know, the technologies, I'm not sure, but

(43:41):
you know, it's really like advancing the data analysis and
implementing into agriculture. So to make your blueberries more
plum, make your Peaches more sweet, whatever.
It's not the sexy one that consumers know about, but it's
actually very advanced. It's there's definitely a lot of
companies operating the space. I think in a very high level
narrative is, and I wouldn't saypeople don't know, I just think
it's something that is not talked about enough in the West

(44:04):
is that this is not an arms raceend of the day.
And I think that's something I want to talk.
About I. Think I, I think end of the day,
look, we've, we've gone through globalization and technology and
knowledge and talent at this point is so transferable and the
development of, you know, the digital ecosystem cannot really

(44:26):
just be stopped. And I feel like, like in terms
of like physical boundaries and borders anymore, it's just not a
very conducive or productive narrative for I think academics
and innovators and business people.
I understand there's obviously political reasons behind this
narrative given the kind of geopolitical narrative and the
backdrop, but it's just really not really like I don't know, I

(44:50):
don't, I guess my point is I don't really get the point of it
because after you talk about it,it's still like, OK, everything
is still moving around. There's still globalization and
the framing of China's AOL development at A at a head to
head context to US really missesthe point where and then they
first of all, China's current deployment LED productivity kind
of approach is really consumer facing.

(45:13):
And the US is, you know, also doing the same thing.
I think for the business people and the tech leaders, end of the
day, sometimes it's just advancing society and creating
profitable businesses, OK, that's that.
And then on the bigger humanity level, is that like, are we not,
should we not be collaborating and thinking about how this
technology is actually potentially very dangerous for

(45:35):
our next generation, for humanity as a whole and
regulating it or understanding the safety kind of around it
better? Like, you know, putting in
restrictions and everything? Sure, it hinders development in
some ways, but it's a very shortterm approach.
So I think this whole narrative around China versus the US just
seems very unproductive for, honestly, the future

(45:57):
generations. And So what is the one question
that you wish more people would ask you about AI in China then?
I think just approaching again, just as a normal market, like I
don't think China is like as a government has like a kind of
crazy sci-fi dream that's different from the West.
You know, I mean, like some people ask like what's like the

(46:18):
Chinese government's like, you know, vision of it.
I was, I don't think it's that different from whatever the
American or the European or the whatever, you know, I think it
obviously is a technology that is extremely still in an early
stage. But at the same time it will be
extremely vital for economic development, societal kind of I
think even prosperity, even potentially forming cultural

(46:39):
norms in the future. And of course natural security
and military. But it's just as like the
Internet was right. And you can't just say let's
just close off the Internet to everyone.
I mean, it's just, it seems a bit, yeah.
Anyway, so I think people shouldjust see it more as another
competitive market and treat thebusinesses coming out of China
just as normal businesses. So my traditional closing

(47:03):
question, what would success mean to you and AI program in
terms of covering and interpreting China and Asia's
technology landscape? I shall not limit that to AI
alone. Yeah, I think it goes back to
what I talked about earlier, howI even started AI Prom on a
professional level. You know, I really hope I grow

(47:23):
more insights. I think it's a flywheel at this
point. The better coverage I write, the
better access I have. You know, I don't truly believe
this. There's been amazing people
reaching out, whether they're investors or people in these
companies. Other startups are big tech and
they provided me so much insightand I've learned so much and
opened my eyes. I think there's that and I think

(47:43):
I hope to continue to provide better write ups for my readers
and that means really nuanced understandings of these
technology development and the business development from a
personal level, I think again, Ijust really hope to kind of
bridge that gap in terms of cultural and social
understanding as well. You know, I really don't think
it's productive to create these narratives or conflating country

(48:08):
versus ethnicity versus technology versus, you know, all
these history of everything. You know, as a Chinese Canadian,
I don't want to see the rise of Sinophobia.
You know, it's not a nice thing personally.
So I think there's a lot of thatas well.
And I just hope there's more mutual understanding on both
ends. But yeah, I think that would be
what really I would consider a success.

(48:29):
There will be a good place to stop, but definitely won't be
your last to be here. And we'll probably be talking
more about the China's landscaping, not just in AI, but
also in technology as a whole. So, Grace, Many thanks for
coming on the show. Any recommendations you have
that inspire you recently? Like, yeah, I think, you know,
actually you can see I have thisbook and my good friend Karen

(48:52):
Howe wrote this and I really highly recommend it.
It's called Empire of AI. And I think, you know, she's a
veteran journalist in AI. And, you know, as much as AI Pro
and myself and my team, you focus so much on the business
strategy of a lot of these AI companies.
It's sometimes really important to be reminded of the social
consequences and environmental consequences, what this means

(49:15):
and even these like figureheads in AI right now.
So Karen Squawk definitely addresses this and I think it's
a really valuable read for anyone.
You mentioned BG 2 podcast with Bill Gurley and Brad Grinster
and I think I I really like their stuff.
I think it's in the investment space coming in the West.
I think it's very new ones, especially on that.
Actually they talk a lot about how the Chinese USAI arms race

(49:36):
narrative doesn't make any sense, but like they.
Actually, they are the only onesthat are seen.
I think they're actually talkingabout.
I think they exactly understand how the Chinese landscape works
and trying to tell everyone elsejust don't go over abroad with
this whole China versus US thing.
Yeah, I think they're just much more nuance beyond, you know,
obviously have great analysis oflike the business direction of

(49:58):
everything, but I think they're also a lot more nuance on the
geopritical take and they don't want people to be carried away
by that. I think they have a great
podcast substat. I love Professor Kyle Chan's
work. It's a lot more academic than my
writing. He's I mean, he's a professor at
Princeton. He writes a lot about China, US
industrial components, economic set up.

(50:20):
And you know, the key compares sometimes the two economies and
why they've evolved the way theyare.
I have to sometimes pause in themiddle and read it and breaking
them in half because they're very long.
But I recommend it. So yeah, I think those are
probably like top of my mind, like a few of the things I've
been really enjoying reading andconsuming.
How do my audience find you? Please feel free to tell them

(50:43):
where to subscribe to AI Prom. OK, so it's AI Pro M, please
check it out. You can follow me personally on
LinkedIn. My name is Grace Shao, last name
Shao. If you ever need to reach out to
me, I would love that to connectonline or offline.
If you're Hong Kong for coffee, whether you're an investor,
industry practitioner, academic,a enthusiast, you know, a

(51:06):
student, I just find it very fast and talking to different
people. Do reach out to me by e-mail.
It's grace@proemcommunications.com.
That's PROE mcommunications.com.Thank you, Bernard.
Yeah, Just one more last thing I'm going to ask, are you going
to be in the Shanghai AI Conference next week for the
World AI Conference? I actually.
Won't be able to make it this time but I did go last year.

(51:26):
It should be pretty exciting. I was really looking forward to
it, but actually I am about to have my second baby.
Congratulations SO. Yeah, so I'm not allowed to fly
anymore capping actually it's just not letting me fly.
So I was quite bummed about that.
But I do have quite a few sources on the ground and people
who will be sharing some insights and I will be doing
write up for stuff. Surprisingly to me, a lot of my

(51:47):
business associates, people who are not even into AI, are
actually bringing their familiesout of normal school days to go
for the AI conference to inspiretheir kids.
That is something that is reallynew to me.
And for sure, next year I have to basically get myself all the
ticket and fly to Shanghai myself to do this.
It's actually very vibrant. I went last year and this is

(52:08):
pretty DeepSeek and I was yeah, absolutely packed.
And I think this year, post DeepSeek, I've definitely heard
more and more people from aroundthe world flying in for it.
And it's a good opportunity because I think all the AI
companies and tech companies from the region or from the
country are going to be coming in to to the AIC.
So I'm very jealous of the people who can actually go on
the ground. I'm FOMO ING so hard.

(52:28):
But yeah, what I will do a subset write up.
So I will send it over to you. OK.
Catch up. We will catch up with you soon
for all those, you can definitely find us anywhere.
So thank you, Grace, and we'll continue to chat.

All Episodes

Episode Transcript

Popular Podcasts

CrimeLess: Hillbilly Heist

Crime Junkie

Stuff You Should Know

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}The Truth About China's Generative AI Revolution Nobody Talks About with Grace Shao

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}CrimeLess: Hillbilly Heist

Crime Junkie

Stuff You Should Know

All Episodes

The Truth About China's Generative AI Revolution Nobody Talks About with Grace Shao

CrimeLess: Hillbilly Heist