0:00

Come back to Jiang Yuming early

0:03

like Weibo posts. Back when

0:05

Weibo wasn't all that censored. And,

0:07

you know, he was making all these comments about,

0:10

like, look, you know, criticizing

0:12

the government saying that there should be more freedom

0:14

of expression in China. And, you know,

0:16

you can find those posts from basically

0:18

every Chinese CEO of

0:21

his era, where

0:23

these folks They worked at Western

0:26

technology companies. They a lot of them

0:28

actually even spent time in the US or in Silicon

0:30

Valley. And the sort

0:32

of ideological mindset

0:35

of the hope of the sort of future

0:37

that they envisioned for China was

0:39

much more akin to the sort of peaceful

0:42

evolution that the US

0:44

was hoping. And and, you know, these

0:47

innovators were the

0:49

closest I think that America has

0:51

really gotten to sort of creating

0:54

a a know class

0:56

of individuals in society which

0:58

deeply internalized a

1:01

lot of the liberal values that America holds

1:03

most dear. And for

1:05

us to get to a point where we are

1:07

now, where we're telling these folks No.

1:10

You can't do this because your country

1:13

has gone so off the rails that

1:15

there's no way we could ever trust you.

1:17

It's a real shame that

1:19

this is where we are today. Because these guys

1:22

were probably like the best hope

1:24

that America had for,

1:26

you know, potentially bending China in

1:28

a different direction. And and, you know, I think she

1:30

recognized that, which is why

1:33

we've had all these tech crackdowns over past

1:35

few years because he understood that the Jiangxi

1:37

Ming's and and and Mayyons of the world

1:39

were not he's sort of onboard with

1:42

his mission of of how he wanted

1:44

to define, you know, Chinese greatness.

1:46

And, unfortunately, we don't

1:48

have chairman chairman Jack Ma running

1:50

the shots China. We have Xi Jinping. And

1:54

that's the sort of reality that

1:56

innovators as well as American policymakers

1:58

are gonna be having to grapple with for years

2:00

and years to Hello, and welcome

2:03

to China Talk. Today, I have on the great

2:05

Kevin Shoe of the interconnected substack.

2:07

He also works at GitHub.

2:10

And we're gonna be talking about today's

2:12

TikTok hearing. We're recording this. We

2:14

started during the half time

2:16

intermission of the the

2:18

house's long awaited chance to grill TikTok's

2:22

CEO. We're gonna be getting into the

2:24

sort of dynamics that led to

2:26

the hearing, do a little bit of

2:28

Monday morning quarterbacking on TikTok's lobbying

2:31

performance over the past few years as well as

2:33

talk about more broadly what this

2:36

moment and the Restrict Act

2:38

means for the future of the US

2:40

and Chinese technological

2:41

ecosystems. Kevin? Thanks so

2:43

much for coming on China. Absolutely.

2:45

Thank you for having me back, Jordan.

2:48

And just a quick disclaimer for everybody

2:50

who was listening live or in

2:52

the recording do not speak for GitHub

2:54

or Microsoft. I am here on my personal

2:56

capacity expressing personal opinions

2:59

only.

3:01

So

3:01

Kevin, what do you think of this hearing? First

3:04

of all, I think it's clear and

3:06

unequivocal at this point that TikTok

3:08

as a company or as a product. Because

3:11

of his Chinese ownership is being cowed

3:13

to a completely different standard

3:16

or playing field than all his other peer

3:18

company here in the US, whether that's meta,

3:20

you know, Snapchat

3:24

or other sort of social media company. That

3:26

is something that I think coming

3:28

back to how TikTok

3:30

has prepared for this hearing.

3:33

They might have tried to actually argue

3:35

that everybody else is doing the same thing

3:37

that we're doing, and therefore whatever we're

3:39

doing as TikTok is, you know, just part of

3:41

an industry problem. But at least on

3:44

the Republican side, it's pretty clear

3:46

now, which has been implicit before, but I

3:48

think much more on the record, that

3:50

TikTok because of this ownership is gonna be

3:52

held to a completely different standard.

3:55

Whatever that standard is is still to

3:57

be determined. But I think

3:59

the argument that's, you know,

4:01

just because we are doing it.

4:04

But Facebook's also doing it. We should be talking

4:06

about it altogether. That argument

4:08

from TikTok is falling flat

4:10

as evident by this hearing

4:13

at least so far.

4:15

Yeah. I mean, what's what's remarkable to

4:17

me is that it

4:20

took this long. We're here

4:22

in March twenty twenty three. Trump

4:25

almost banned the app in twenty

4:27

twenty. And the sort

4:29

of groundswell of

4:32

sort of political consensus making. Like,

4:34

all the facts have been there for

4:37

years now. But the sort

4:39

of metabolism of

4:41

the, you know, policy media

4:44

ecosystem to what

4:46

TikTok is, what constraints

4:48

it operates under, and

4:50

the, you know, power and influence that this platform

4:52

has. Watching that

4:55

sort of move forward and

4:57

fits and starts over you

4:59

know, a multi year process

5:01

has been really fascinating and

5:03

a and a really interesting civics lesson for

5:05

me in, you know,

5:07

what it takes to go from

5:10

sort of an idea to

5:12

consensus to action. And and, you know,

5:14

just over the past week, we finally

5:17

seen that the Biden administration has made decision

5:19

that, you know, project Texas, which

5:21

they've been which TikTok has been pushing

5:24

for the past few years, is not gonna cut it. And

5:26

I think this has to,

5:28

in my opinion, with also the legal

5:32

sort of justification of

5:34

either Bany or forcing a sale of TikTok,

5:37

which actually gets into the restrict

5:39

act that was proposed by the it. Right?

5:41

But let me first kinda talk about the project

5:43

Texas Engle to I think that's pretty interesting,

5:46

especially coming from listening

5:48

to this hearing is still going on right

5:50

now. But one of the talking points that

5:53

is very clear that

5:55

Joel showed which is the Chinese

5:57

pronunciation of the guy's, you

5:59

know, name is falling

6:02

back on is project Texas. Right? He's

6:04

like, right now, we're still

6:06

basically not being able to guarantee

6:08

that there's no access to American data.

6:11

But once project Texas has done,

6:13

we will be we're spending billions of dollars

6:15

on it. No one else in our industry

6:17

is doing it. So he's trying to

6:19

say that once we finish project Texas,

6:23

we'll be good. And, you know, the the

6:25

the question here is, one, why

6:27

is it taking is a lot for project Texas

6:29

to be finished since that, you know, it's like at

6:31

least a year and a half project at this point

6:33

and also if that is

6:35

gonna be the lynchpin to solve

6:39

TikTok's problem that just say, why

6:41

go up at the hearing now and way

6:43

forward to, you know, what either you finish

6:46

and then report to the American public

6:48

or to Congress or give them what's

6:50

much more precise progress reporting

6:52

Right? Like, it's sixty percent done. It's eighty

6:55

percent done. I think with the other commitment

6:57

that he made, that was maybe marginally

6:59

in use making is that TikTok is also

7:02

planning to delete all American user

7:04

data that's currently being stored in this Virginia

7:06

data center that's not Oracle run, therefore

7:09

not Projekt Texas. And the

7:11

the backup in the Singapore data

7:13

center. So it's an ongoing project

7:15

which makes the TikTok

7:18

CEO's job today when

7:20

frankly, way more difficult than

7:22

it probably could have been. Yeah.

7:25

I mean, this is this is sort of the weirdest

7:27

thing of all of it. Is

7:29

like they caught a real break

7:32

by having Trump sort of

7:34

get distracted and, you

7:36

know, Larry Ellison, whisper in his ear

7:38

that, like, he had a solution. And they

7:40

had three years to

7:42

put in place a better sort

7:45

of, you know, better fact pattern to

7:47

be able to go to these sorts of debates

7:50

and, you know, testify in front of congress

7:52

and and give a more convincing pitch

7:54

that look we get it and we're doing something

7:56

about it. And, you know, maybe they just

7:59

thought they could be cute and, you

8:01

know, be really quiet and

8:03

and and eventually sort of

8:05

end up at a place where this

8:07

this fell off the head and maybe, you know, if

8:10

US China relations had had developed

8:12

in a different way than it had over the past few

8:14

years, that might have been a more feasible strategy.

8:16

Or maybe, you know, this whatever

8:18

project Texas is is so

8:20

many billions of dollars that the

8:23

sort of utility

8:25

and having a better talking point to

8:27

congressional staffers wasn't

8:29

worth the expected of

8:31

value you know, the expected value of that wasn't

8:33

worth the, you know, five, ten billion

8:35

dollars or whatever that it would take to hire

8:37

all these new American engineers and poured everything

8:40

out of China. But the fact is that we're sitting

8:42

here in March of twenty twenty three with

8:45

the CEO being unable to definitively

8:47

answer questions about, you

8:49

know, engineers based on the mainland having

8:51

access to the algorithm and data that

8:54

these congresspeople are so focused on

8:56

and it didn't have to be that. You

8:58

know, Kevin, you you've been talking a lot

9:00

over the past three years and we've done shows about this

9:02

of the idea that they could have decided

9:04

to kinda do what Elon is ostensibly

9:07

gonna be doing with TikTok, which is open sourcing

9:09

the entire algorithm. So talk me

9:11

through a little bit about how that you

9:13

know, what what what that road not

9:15

taken could have looked like and why that might have

9:17

been a potential solution for by

9:20

chance to ameliorate some of these

9:21

concerns? I think one of the things,

9:23

and I've heard this from a lot of

9:25

Chinese entrepreneurs and business

9:27

people, not just, you know, at the TikTok

9:30

level. When Trump lost the twenty

9:32

twenty election event became president.

9:34

There was this false sense of comfort

9:36

that US China relations will get

9:39

can imagine magically better,

9:41

right, overnight. And I think that might

9:43

have allowed lot of the TikTok and Biden's

9:45

leadership into thinking that this is something

9:47

that they can negotiate a

9:51

way. Right? Which or there is technically

9:53

still a pending negotiation with the Biden administration

9:55

to talk about this. But think

9:58

it's very clear that if,

10:00

you know, the administration, the by administration could

10:03

have banned this app, they probably

10:05

would have done it by now. But because

10:07

the Trump attempt to ban the

10:09

app was struck down by not one

10:11

but two federal judges, for

10:14

kind of abuse of the, you know,

10:16

economic emergency powers

10:18

act, right, which is sort of his justification.

10:21

For how he as president could

10:23

do this and it clearly didn't work. Obviously,

10:25

the Biden minister wouldn't try to do it again

10:27

because only if only because you avoid

10:30

the embarrassment that that that, you know, their order

10:32

could also be struck down by the court.

10:34

And now that we have a proposal like

10:37

the act, which came out of the senate

10:39

a few weeks ago, that more

10:41

explicitly gives the

10:44

president and the executive branch the

10:46

power to do these sort of things And

10:48

if that, you know, bill gets passed,

10:50

then it's just a much easier

10:52

legal process to

10:54

ban, not just TikTok, but any

10:57

a kind of Chinese tech company and

10:59

their product in the United States. If

11:01

there is enough justification of a

11:03

natural security concern to

11:06

push them out. And

11:08

that's sort of the game that I think

11:10

is so d c inside baseball.

11:13

Right? Like, how could the president just be able to

11:15

strike out by the court? Oh, we have a new friendlier

11:17

president potentially than Trump? And life

11:19

is gonna be okay. I think that

11:22

was what happened in the last

11:24

two to three years, and it's now coming

11:26

to light again that was a false

11:29

almost a fantasy that a lot

11:31

of these folks were living in and

11:34

is coming to a head right now with a TikTok

11:36

hearing. So, Kevin, what

11:38

is what's in the district act and

11:39

what struck out to you about it? So,

11:41

the district act is a proposal from

11:43

the And

11:46

introduced by senator

11:48

Mark Warner from the Democratic

11:50

side to the senator John from

11:52

the Republican side. It also already has

11:55

the enthusiastic endorsement of

11:57

senator John mentioned, which depending

11:59

on the day, you don't really know if he's Republican or Democrat

12:02

or whatever he feels like he wants to be.

12:04

But it's a act. It's a

12:06

proposal. It's a bill. It's not a law.

12:08

It's a bill that basically gives

12:11

the president and the executive branch,

12:13

namely the Department of Commerce,

12:16

very wide range

12:19

of power to

12:21

determine what

12:24

kind of technology products or

12:26

services from America's forward

12:28

foreign adversary. Right? That's a very specific

12:30

set of definition that includes the People's

12:32

Republic of Also Hong Kong and

12:35

Macau, Russia, Cuba, on

12:37

North Korea and Venezuela under Maduro.

12:40

Anything from those countries that

12:42

that the president now has the power to

12:44

review and figure out whether

12:47

they post any national security threat

12:49

and to have the power to deal

12:51

with it, you know, accordingly. Right?

12:53

Which could be a cell. It could be a band.

12:56

It could be whatever the executive branch

12:58

sees fit. And we're really struck out

13:01

really was interesting to me from

13:03

this bill is that the

13:06

definition, the list of

13:08

things that could fall under the cover

13:10

technology areas for this bill

13:12

is extremely comprehensive.

13:15

It basically covers everything that any

13:17

technology company could make under

13:19

this done. Right? So it's not just kind

13:21

of a euro suspect of networking equipment

13:24

or drones or other sort of hardware

13:26

that we've come to be familiar

13:29

with as potential foreign adversary

13:31

spywares, I'd just say, that's in the United

13:33

States. But it includes certainly social media

13:35

platforms. Includes e commerce,

13:38

technology, and services, includes

13:40

a bunch of strategic areas

13:43

like AI, quantum computing, cryptography, biotech,

13:46

all sort of stuff. And it also

13:48

includes cloud services as well. So

13:50

CDN, kind of SaaS product,

13:52

and even the companies that

13:55

manage open source software,

13:57

which, you know, technically is not actually

13:59

owned by the company.

14:02

Open source is just this thing that's a

14:04

piece of code that you can access

14:07

on the Internet you'd use as you see fit

14:09

for the most part. I'm glossing over a bunch

14:11

of details on open source. But I think generally

14:13

speaking, that's how you could understand

14:15

open source in this context. Even companies

14:17

that are commercializing where utilize

14:20

the open source for their business games are

14:22

now under the coverage of this act,

14:24

which to me is basically spelled

14:26

game

14:26

over. For any Chinese tech company,

14:29

not just the big ones, but even the relatively

14:31

small ones, the startup ones

14:33

that wanna gain some market share in the US.

14:36

So I moved to China in twenty seventeen.

14:39

And the even

14:41

then, she had just changed the

14:43

constitution, the sort of the

14:45

dark clouds are on the horizon. But even

14:47

then, like, there was still

14:50

this vision that maybe

14:52

some sort of, you

14:54

know, interconnected, commercial,

14:58

technological interaction between trying

15:00

in the rest of the world could potentially kind

15:03

of create or

15:07

or build the context in

15:09

which the US and China could relate to

15:11

each other on better terms. And

15:14

this hearing and the Restrict Act

15:17

really marked to me an end of an era. You

15:20

know, not that we need anymore. You know,

15:23

maybe you could maybe you can put that at maybe

15:25

you can date it from no limits friendship

15:27

and and and and Putin in the war in Ukraine,

15:29

but the sort of

15:32

consensus that has

15:34

been manifested with this bill

15:36

and TikTok being banned, I think

15:38

are a real marker

15:41

for just how far we've gone and how

15:43

much it's going to take to

15:45

shake this consensus. Of

15:49

just the

15:52

idea that's made it into

15:54

the American political blood stream that

15:56

US Chinese technological interaction

15:59

is somewhat akin

16:01

to, like, doing deals with the enemy. And

16:04

everything, you know, any

16:06

any sort of sort of commercial or technological

16:09

engagement needs to be viewed through

16:11

that lens because of the broader

16:15

challenges that are facing the US and China.

16:17

And, you know, we've done lots of other shows on

16:19

China talk that have explored you know,

16:21

why that is? Whether or not it's justified? You

16:24

know, is this she? Is this the US?

16:26

That that really bears the brunt of

16:28

the blame here. But the consensus

16:32

that the consensus that's

16:34

developed seems very,

16:38

you know, seems strong and,

16:42

you know, it's it's it's Like,

16:46

I Like,

16:49

from a normative sense, I

16:51

I think where I am on this is

16:54

it kind of has to happen. But

16:57

I I there's something

16:59

that I find about the

17:02

moment we're in, which is so sad.

17:05

Because I don't think it's inevitable

17:07

that the US and China

17:10

had to end up where we are today.

17:12

And, you know,

17:14

I'm gonna I'm gonna put it like ninety

17:16

ten on on on

17:18

on on Xi in Beijing's shoulders for

17:21

for pushing us pushing us to this

17:23

moment. But, you know, when when I look at,

17:26

you know, the likes of John Yuming and, you know,

17:28

this this this this TikTok

17:30

CEO who seems like a really nice guy, It's

17:32

like it's we would we would

17:34

be in a much nicer world if

17:37

the two countries and the two systems

17:39

were in a place where

17:41

they could be able to

17:43

build trust in each other. And

17:46

this guy could be thinking about how to, you

17:48

know, make creators more money and,

17:51

you know, not have to spend all

17:53

his time on completely justified,

17:56

sort of concerns about data privacy and

17:58

algorithmic bias. But that's

18:01

not the world and that's not the timeline

18:03

that we're in. And there's something

18:06

I think there's something about the

18:11

the discourse in the US around

18:13

the US China relationship, which isn't quite

18:16

grappling with like the tragedy of

18:19

all of this. And, you know, it's probably

18:22

I think the the most tragic thing

18:24

is not necessarily just the US

18:26

ChinaTalk relationship. It's the trajectory of

18:29

of of Chinese governance over the past, you know,

18:31

ten, fifteen years. But it

18:33

is something that is deeply sad

18:36

that this is the this

18:39

is the this is the place that we're all in. I

18:42

think the marker of this end

18:45

of an era with the TikTok

18:47

hearing today, the Restrictive Act,

18:49

or something like it possibly becoming

18:51

a law, is not that

18:54

hyperbolic, to be honest, that this might be

18:56

because I'm more deeper in the tech industry

18:58

these days. Obviously, I had a past life in DC,

19:01

but I work with a lot of entrepreneurs not just from

19:03

from a lot of different places around the world

19:06

as well. I have held hope

19:08

that while products

19:10

like TikTok, which poses much more

19:12

obvious natural security threats, could be

19:15

dealt in one bucket. There

19:18

are a bunch of different industries,

19:20

whether it's e commerce or cloud

19:22

SaaS or what have you.

19:25

That are just in a much more innocuous part

19:27

of the tech industry where these are just

19:29

ultimately trying to make products and expand

19:31

their services, trying to make more money, should

19:34

have justified their valuation with

19:37

their VC investors. These are just

19:39

kinda normal behaviors.

19:41

Right? In business. That used to be a much

19:43

more free flowing. But now, certainly,

19:46

based on my understanding, we're reading out the restricted

19:48

act. They're all being shoved

19:51

under this much larger movement

19:53

of increasingly adversary relationship

19:56

between the two countries. That sucks up all

19:58

these normally very ordinary of

20:00

activities into this context. And

20:03

I think that is a new level

20:05

of separation that certainly I have

20:07

not seen. Not just in the past three years,

20:09

but in the past, you know, thirty years,

20:11

probably. And it's

20:14

something that I think just needs recognize

20:16

at this point, there's not a whole lot of wishful

20:18

thinking. I think the reality

20:20

is reality, and a lot of people need

20:23

to change their plans. That they might

20:25

have drummed up and dreamed up, you know, that

20:27

John Yee, dreamed up to be a global company

20:29

to ten or twelve years ago and needs

20:32

to change because that's just the world

20:34

we live in right now. And that has nothing

20:36

to do with whether TikTok is a national

20:38

security threat or not, which there's

20:40

perfectly enough evidence to suggest that

20:42

it is. I think there still could be more

20:44

validation of that claim. Again,

20:47

we could have dealt with this single verily. Right?

20:50

And, you know, coming back to other stuff that we've

20:52

talked about about TikTok specifically in the last

20:55

three years, there are opportunities

20:58

where TikTok could have led

21:00

the pack or really stayed in front

21:02

of this problem as well. Whether

21:05

it's open sourcing their recommendation algorithm,

21:08

which, you know, I've talked about before with

21:10

you, whether it's open sourcing

21:12

even the access level between

21:15

their you know, Chinese engine forces and

21:17

their American engineering forces to show

21:20

a little bit more actual transparency and

21:22

not to spend money building a quote unquote

21:24

transparency center to, you know, parade

21:26

a few journalists around and watch, you know, videos

21:28

about terms and services, those would actually

21:30

be ways to technically verify you

21:33

know, these claims. And again, it

21:35

is TikTok going well beyond

21:38

the norm of the industry. Which

21:40

I can understand why they would feel

21:43

kind of maybe victimized to do

21:45

the why are we being held up to a different

21:47

standard? Why not why is Facebook as Snapchat

21:49

not doing this, but that is just a world that

21:51

we live in. It is that if you do wanna operate

21:53

here as a Chinese own company, you will

21:56

have to go well

21:58

beyond the norm,

22:01

the standard to demonstrate your

22:03

trustworthiness. And there are ways

22:06

to do that via open source, via other

22:08

ways of communicating this in a technical level,

22:10

and we're just but

22:12

fortunately, that didn't happen with TikTok. Right?

22:14

So here we are. Yeah.

22:17

Sort of one side note is interesting

22:20

sort of like she governance thing ending

22:22

up in this is, like, everyone's citing the

22:24

national security law. Right? Like, Chinese

22:27

Chinese government can basically make

22:29

make part make make private companies

22:31

do whatever they want if it's in the interest of national

22:34

security. And, you

22:36

know, she didn't need this law.

22:39

To make companies do what he wanted

22:41

them to do. But

22:43

the the sort of vision of of,

22:46

like like the like, the legalization of

22:49

Chinese governance is something that he's been very

22:51

focused on as a sort of, like, way of being

22:53

for the party. And it's not rule of law.

22:55

It's rule by law. Right, of, like,

22:57

we want to create the legal framework

23:00

to understand and justify everything

23:02

that we're doing. But, you know, now fast forward

23:04

five years, six years later, and

23:07

the saying the quiet part out loud

23:10

of the Chinese of the

23:12

of the Chinese government's influence over

23:14

every aspect of sort of

23:16

Chinese corporate and and and private

23:18

life ends up leading to the case

23:20

where you have where you have

23:22

the CEO of TikTok basically being

23:25

forced to answer our questions which

23:27

are unanswerable because at

23:29

the end of the day, you know, by

23:31

dance has no has no recourse

23:34

when it ends up

23:36

getting those sorts of getting those sorts

23:38

of requests and directives. And that's

23:40

the thing, right, Jordan. Like, coming back to

23:42

the hearing. Right? And, you know, as the live

23:44

folks tune in, the hearing is still

23:46

happening. I got Like, this

23:48

morning, when the news dropped that the Chinese

23:51

Commerce Ministry issued an official statement

23:53

about, you know, any

23:55

sort of TikTok sale or the investor divestiture

23:58

being basically dead on arrival from a Chinese

24:00

regulatory perspective, it was quite

24:02

a PR curveball thrown at

24:04

to take that CEO the morning before he

24:06

has to go to the hill. Like, not

24:09

to be able just like, wow. You're really not trying

24:11

to make this guy's job any easier

24:14

than they already. And, like,

24:16

what is the timing of that was actually a little bit

24:18

curious. I don't know if you thought much about

24:21

the timing of that particular statement,

24:23

which could have happened, frankly, any

24:25

day. And of course, a lot

24:27

of members on the hearing

24:29

committee seized on that statement.

24:32

Right? To really push him

24:34

to answer a question that is literally impossible

24:37

for him to answer and basically kind

24:39

of undermines any sort of valid

24:41

case that you could have made about the separation

24:43

between TikTok, the Communist

24:46

Party, which was said the Chinese Communist Party literally

24:48

just said

24:49

you can't be sold without

24:51

-- Yeah. -- our permission. What do you say

24:54

to that? And

24:56

what

24:56

would Yeah. And, of course, you know, he's he's

24:59

This comes back to the Monday morning quarterback thing.

25:01

I think he has a better answer for that. He's like, look,

25:03

in the US, like like technology

25:06

transfer, sales, like, have to

25:08

go through sorts of inventory controls and, like, we

25:10

have the same thing in the PRC. But it

25:13

leads you to the same

25:15

point it leads you back to the same

25:18

question, which all the congresspeople have,

25:20

which is an unanswerable one, which is, look,

25:22

these algorithms are were created

25:24

by Chinese engineers who have access

25:26

to them on a daily basis, and there have been lots of

25:28

stores areas about how sort of algorithmic

25:30

tweaks happen in the US

25:33

employees, like, have no idea what is

25:35

going on under the under the hood of

25:37

their own app. And, you know, that's something, again,

25:39

that they could have done something about

25:41

over the past over the past three or four years,

25:44

like it's not in possible. But

25:47

it's for whatever reason a calculation that they

25:49

decide wasn't wasn't in TikTok's interest and now

25:51

they're having to pay the price for that. Broadening

25:53

it out a little bit, Kevin. We're

25:57

at the we're we're sort of

25:59

at this inflection point where

26:03

more and more of

26:06

US US and China, so of

26:08

technological inter

26:10

inter interconnect points are going

26:13

to be unwound. But there

26:15

are still really dramatic dependencies

26:18

on both sides of the equation, which aren't

26:20

going to be, you know,

26:22

which which neither country is gonna be able

26:24

to kind of, like, wriggle their way out of any

26:26

time. Soon. Maybe start off

26:28

talking about batteries and,

26:30

you know, what what sort of themes that illustrates

26:33

about just how reliant these two countries

26:35

are still gonna end up being on

26:37

each other to to sort of realize their national

26:39

goals? That's right. So

26:42

you know, as this larger trend

26:44

of the globalization and, you know,

26:46

interconnected points being broken

26:48

down. One, I'm rethinking the

26:50

name of a newsletter, whether it should still

26:52

be called interconnect it or interconnect it with

26:54

a question mark. But on

26:56

the battery points specifically, think

26:59

that is one of those industries. And we're

27:01

talking specifically about electric

27:03

vehicle batteries. Where

27:07

there is so much lead

27:09

that, you know, China has built

27:11

in that industry that

27:14

as the United States tried to electrify

27:16

itself with the

27:18

inflation reduction act the

27:21

subsidies that are being rolled out

27:23

as we speak to the American consumers

27:25

to buy EVs

27:27

made by domestic brands like

27:29

Ford and GM and whatnot. There

27:32

is a awkward

27:34

reliance that is hard for US

27:36

politicians recognize of how

27:38

much US automakers, frankly,

27:41

simply do not have access to

27:43

enough domestic technology to

27:46

acquire reliable

27:48

batteries for their EVs, and they must

27:50

rely on kind of global leaders like

27:52

CATL, which is a Chinese company that

27:54

is by far the largest supply. BB

27:57

batteries in the world rely on

27:59

their technology to be

28:01

able to make the American dream of electrification

28:03

come true. And the evidence, I'll actually

28:06

point to two recent evidence. One, lot

28:08

of folks might have heard about, which is

28:10

Ford has decided to

28:12

build EV battery in Michigan with

28:15

CAT house technology to

28:18

be able to simply supply enough

28:21

functioning battery for the

28:23

next generation of Mach

28:25

three and Ouzmaki, I think,

28:27

and also the EV pickup truck. The Ford

28:29

one fifty. And a more recent

28:33

news is at I think this was

28:35

just reported on today. Ford

28:37

had battery incidents,

28:39

a bunch of trucks were caught on

28:41

fire, they paused the production, and

28:44

the batteries were being supplied by the

28:46

South Korea. Battery

28:48

giant SK or the

28:50

company that you call SKOM or SKOM.

28:53

Is there a battery subsidiary? And

28:55

it's been, you know, diagnosed now

28:57

and published by the US regulators that

29:00

these batteries are defective

29:02

and is one of the causes at least of the fire.

29:04

Which kind of further pours oil

29:08

on the fire, so to speak, that even the

29:10

alliance that the US has built

29:12

around the world with South Korea,

29:15

Japan and other countries.

29:17

That isn't enough to really

29:19

catch up to the lead that China has built

29:22

in this particular industry of EV battery

29:24

making to be able to help Americans

29:26

notify. So that is

29:28

one industry where you

29:31

know, the the the practicality of

29:34

decoupling or de globalization is

29:37

about to really hit, you know, where

29:39

we see the rubber mister Road because it's very

29:42

kind of easy to talk about it in a congressional

29:44

setting. Even in a legislative setting,

29:47

whether it's giving money away, to

29:49

incentivize or to push company

29:51

out in the contact subcontractor. But

29:54

if you need a battery, you need a battery. You

29:56

don't wanna roll out trucks that's gonna set

29:58

on fire and there is this one source

30:00

of reliable battery and it just happens

30:03

to be from your foreign

30:05

adversary. So there's still a threat

30:07

of interconnectedness, but

30:10

it's sort of dangling on these very

30:13

you know, rope lines, technological

30:16

reliance, that's hard to catch up.

30:18

And it's hard to kind of really recognize

30:20

honestly too. If you are somewhat who

30:22

is in DC and trying to drum

30:26

up all these narratives about

30:28

the globalization

30:29

decoupling. Yeah.

30:31

One of the the the

30:33

most interesting one transitioning to talking

30:35

a little about AI is the

30:38

idea that Chinese firms

30:40

are going to be increasingly reliant on

30:44

US cloud providers to

30:46

have access to the latest and greatest

30:48

NVIDIA chips. That are gonna

30:50

be run from data centers in Singapore.

30:55

This given the

30:57

sort of export control restrictions, you

30:59

know, there'll be some wiggle room,

31:01

I think, with with with chips to

31:04

sort of stay around the

31:07

the capabilities that have been developed in in

31:09

twenty twenty and twenty twenty one. But

31:11

it it won't take very long for

31:13

the, like, hack that

31:16

NVIDIA has done to get its

31:19

get its chips, you know, just under the limits that

31:21

the commerce department has set. For that stuff

31:24

to just be out of date and, you know,

31:26

not nearly as performant or as energy

31:28

efficient as what you can get from the

31:31

from the latest and greatest technologies. And

31:33

the solution now is

31:36

for sort of foreign cloud providers

31:38

to allow the the Tencent and and

31:41

and Ollie's and bike dances of the world

31:43

to train their models just outside

31:45

of China. Because of course, you

31:47

know, China can't develop

31:51

a leading edge leading edge chips

31:54

and supercomputers without access to,

31:56

you know, the likes of the likes of TSMC. But

31:59

I I find it very hard to imagine

32:02

a world in which two

32:04

to three years from now when it becomes

32:06

even clearer that these

32:09

models and the inference needed to power

32:11

them on sort of nationwide scale

32:14

is something that the US is providing

32:16

ChinaTalk that's gonna be something

32:18

that is going to fly, when Jenson

32:20

Wang, inevitably in twenty twenty

32:23

four, will be called in kind of in front

32:25

of Congress to ask why he is why

32:27

he is doing stuff, which is sort of, like,

32:31

powering the the the Chinese

32:32

economy. Yeah. I think that's another

32:34

one which the interconnect, and this still

32:37

holds, but it's just kind of two layers

32:39

beneath the surface. I

32:43

shouldn't be surprised by this, but I was surprised

32:45

to read that, oh, of course, you

32:47

could just rent the fucking GPUs.

32:50

In a data center in Singapore

32:53

or in Dubai or in

32:55

wherever that is not the United States

32:57

and not in China. That is so obvious

33:00

that you just put in a credit card and AWS

33:02

or, you know, GCP will gladly take your

33:04

money to run those workloads and never ask a question.

33:07

Right? Or even not a or you could do a via

33:09

partners where they're domiciled elsewhere

33:11

outside the United States and the

33:14

US. And this is something that is very standard

33:17

in the industry of cloud computing

33:19

infrastructure, data infrastructure, or

33:21

just webs of partners

33:24

and global service ERC. You know,

33:26

these are called GCI's, you know, huge Indian

33:28

companies, Accenture. They

33:30

do this stuff for a living.

33:32

And you know, initially when

33:34

we offered about the news about video,

33:36

you know, the sanction cannot sell the

33:39

highest end GPU to China, we're like, oh, crap.

33:41

This is it. But then it's like, oh, yeah. They

33:43

just live in the cloud. You rent them for the

33:45

most part anyways. If you are

33:47

a company. Now, it does still

33:50

handicap national initiatives where

33:52

you do wanna own the hardware to develop

33:54

your supercomputer. think that will get step

33:57

back a few you know,

33:59

years perhaps in But

34:02

Nvidia has a monopoly on

34:05

the entire hardware side

34:07

of, you know, AI training and

34:10

for it's generating right now. There's no question

34:12

about it. You just look at the stock. It's

34:15

going up and up. doesn't matter what the Fed

34:17

is saying. And that's

34:19

the reality of business.

34:22

And that's another thing with

34:25

I don't know if whoever's, like, doing the enforcement

34:27

in the commerce department not really aware

34:29

that you could just rent server, you don't have to buy

34:31

them, and the effect is pretty much the same. Maybe

34:34

a little bit more expensive, but that is where

34:36

we are heading to. So I guess

34:38

this would just be a business boom to all

34:40

the, you know, cloud regions in

34:43

Singapore and Southeast Asia and maybe

34:45

even

34:45

Japan. No. You know, South Korea. Yeah.

34:49

I mean, it's I think the the justification

34:51

around the regulations was, like, look, these are

34:53

national security concerns. You know,

34:55

we're we're doing this for national security. And

34:58

the PLA can't be that dumb to

35:00

run their sort of like ballistic modeling

35:03

in Singapore. Because America

35:05

will, like, find out about that, and then, like, we'll be able

35:07

to see their ballistic modeling or

35:09

whatever. But maybe not. I think that could be

35:11

right. Yeah. Like, they that part of the

35:13

modeling or that AI initiative it's

35:15

trying to make will probably never leave the

35:17

border for obvious reasons.

35:19

But it doesn't necessarily handicap your

35:22

you know, sense times of the

35:24

world. Yep. Right?

35:26

Whole And that's what we've seen is is

35:28

is is all of these all of these entity

35:30

listed companies basically

35:33

being able to access whatever NVIDIA compute

35:35

they want. Via subsidiaries or,

35:38

you know, changing the name of their company, so it's not

35:40

on the any list anymore. And

35:42

just finding a friendly friendly

35:45

cloud service provider to to run,

35:49

you know, to to to to to

35:51

run these models for them.

35:53

Yeah. And, you know, and these presents, like,

35:55

a b sanction laws are easy

35:57

to pass, and you know, climate

35:59

that we are we have and do you see it right

36:01

now? But it's a

36:04

lot of work to regulated

36:06

actually for us and implement. And

36:08

for what reason or another, you know, the

36:10

commerce department, which I did spend a little

36:12

bit of time working in past is really becoming

36:15

kind of the most popular department or agency

36:17

in BC these days. And ways that, again,

36:19

beyond my wildest expectation, we used

36:21

to be not that cool. And kind of a backwater

36:24

agency. And now, like, we've been talking about

36:26

in, like, every single, you know, hot issue

36:28

of the entire

36:29

country. So gonna

36:32

really staff up now. Yeah.

36:34

No. They I mean, they had this they

36:37

they had this new budget request for for more

36:39

money for this. But the one the one thing that really

36:41

struck out to me was this two point

36:43

four million dollar line item for

36:45

thinking about emerging technologies. And

36:48

my first response was like, okay, I'm glad

36:51

you guys are wanting

36:53

to think about emerging technologies. But if you

36:55

actually wanna control if you're gonna need a lot

36:57

more than two point four million dollars,

36:59

to do the analysis to understand exactly

37:03

how these industries work and the and the trend

37:05

lines that you may want to bend

37:07

to make sure that you know,

37:09

you're advancing a US national

37:11

and and and economic interests. And

37:14

yeah. I don't think I don't think two point four million

37:17

dollars would even get halfway there

37:19

just on just on this cloud compute question

37:21

which we've been talking

37:22

about. Yeah. And we're only talking

37:24

about defense. Right? Jordan, is the

37:26

offensive side of this relationship with

37:29

the ChinaTalk, which you've written

37:31

a bunch of about and post even

37:33

job descriptions to what we have in the commerce

37:35

department to help them recruit, which is that

37:37

they're still the office department. We still need to

37:39

build our advanced manufacturing

37:41

capacity in the United States. Yes,

37:43

the law has been passed. The money hasn't

37:45

allocated anything. But how are we

37:48

actually going to properly

37:50

and intelligently allocate, you

37:52

know, the fifty two billion dollars or the thirty

37:54

nine point six billion dollars that go directly to

37:56

manufacturing subsidies. This requires

37:59

really smart people that we actually have not

38:01

staffed up. And that that's another part

38:03

that I'm a little bit surprised by. Again, that shouldn't

38:05

be. Which is that, oh, we actually pass the

38:07

allocation first, and then we

38:09

hire the people to be

38:12

allocating. Maybe that's

38:14

is just how we're gonna have to work on

38:16

these things these

38:16

days. The pace of bills and regulations

38:19

are outpacing the number of people who can't really

38:21

do the job.

38:26

Any other points you wanna hit, Kevin? What else

38:28

is on your mind? I know this

38:31

it's a personal issue in a sense, right,

38:33

for both of us. You've worked in Chinese

38:35

tech tech place in the past. I've worked with

38:37

a lot of them as well. And

38:39

it's always an assumption that there's at least

38:42

a certain part of the economy between

38:44

these two countries. That could just stay

38:46

benign. Yeah.

38:48

But I don't think that is

38:50

the reality that we are walking

38:53

into

38:54

whether we like it or not, and that's just something

38:56

that we need to recognize and operate

38:58

under. Come back

39:00

to Jong Un early

39:02

like Weibo posts. Back

39:05

when Weibo wasn't all that censored. And,

39:08

you know, he was making all these comments about,

39:10

like, look, you know, criticizing

39:12

the government saying that there should be more freedom

39:14

of expression in China. And, you know,

39:17

this is this, you can find those posts

39:19

from basically every Chinese CEO

39:22

of his era, where

39:25

these folks They they

39:27

worked at Western technology companies.

39:29

They a lot of them actually even spent time in

39:32

the US or in Silicon Valley. And

39:34

the sort of ideological mindset

39:39

of the hope of the sort of future

39:41

that they envisioned for China is

39:43

was not it was was

39:45

much more akin to the sort of peaceful

39:48

evolution that the US was

39:50

was hoping and and, you know, these

39:53

these these innovators

39:56

were the closest I think

39:58

that America has really gotten to

40:00

sort of creating a a

40:03

class of individuals in society,

40:06

which deeply internalized a

40:09

lot of the liberal values that America holds

40:11

most dear. And for

40:13

us to get to a point where we are

40:15

now, where we're telling these folks.

40:17

No. You can't do this because

40:20

your country has been ruled Because

40:22

the country of your leadership has gone so

40:24

off the rails that there's no way we

40:26

could ever trust you. It's

40:29

just a it's a it's

40:31

a real it's a real

40:33

shame that that this

40:35

is where we this is where we are today. Because

40:37

these guys were probably

40:39

like the best hope that America had

40:42

for potentially

40:45

bending China in a different direction. And and, you

40:47

know, I think she recognized that, which

40:50

is why we've had all these tech

40:52

crackdowns over past few years because he understood

40:54

that these people, you know, that that the Jong

40:57

Unings and and and myons of the world

40:59

were not sort of onboard with

41:01

his mission of of how he wanted

41:03

to define, you know, Chinese

41:06

Chinese greatness. And, you know,

41:08

unfortunately, we don't have chairman

41:11

chairman Jack Ma running the shots China. We

41:13

have Xi Jinping. And that's

41:15

the sort of reality that

41:18

as that that sort of innovators

41:21

as well as American American policy

41:23

makers are gonna be having to grapple with for

41:25

years and years to

41:25

come. Yeah. The folks like Jack

41:28

king and others

41:32

would have the the best

41:34

ally of the United

41:36

States in its way of dealing

41:38

with US China relations. But

41:40

instead, we are also in

41:43

our best to push them away.

41:46

All of all two Singapore, apparently. And,

41:49

you know, they are kind of unappreciate

41:52

it, I'd just say. Right? Like, in both

41:54

their home and their target market

41:56

with that are just unfortunate. So

41:59

with that, you know, thanks for having

42:01

me back in the show Jordan. I wish it good to

42:03

hang out with you to riff on some

42:05

of this stuff.

42:08

Yeah. I mean, last thing, let's just talk about, like, a

42:10

talk about AI for a second. I don't know. You're at GitHub.

42:12

You guys created, like, the first real product

42:16

that people are using. You

42:19

know, any thoughts, reflections on

42:21

on GPU t four and what it means for

42:24

maybe not just US China relations but like

42:26

humanity? Or US China relations?

42:28

don't know. Again, you know, I'm just, you know, my

42:30

own personal fear about anything

42:32

and everything and not really to GitHub or Microsoft.

42:34

But I will share one thing with

42:37

you that I think is relevant to the audience

42:39

is that so you know how I write this new

42:41

that is actually bilingual. Right? So

42:43

Connected has every post that

42:46

is both the English, Chinese that are certain

42:48

by myself. And one new workflow

42:50

I have incorporated recently with chat GPT4

42:52

in particular is to ask

42:55

it to translate for a

42:57

draft from English or Chinese for me

42:59

before I kinda edit and publish.

43:01

And I found the result I

43:04

wouldn't say absolutely better than

43:06

alternatives, but surprisingly more

43:09

fun to do.

43:10

Yeah. And also, some of the expressions are

43:13

a little bit more colloquial in Chinese,

43:16

which, you know, makes sense for my

43:18

purpose. Because I'm writing a personal newsletter.

43:21

I'm now trying to write an academic paper

43:23

or newsletter or news article

43:25

or something more formal. And I did

43:28

find using a chat GPU

43:30

before to help me, you know, produce the first

43:32

rap in Chinese has really boost my

43:35

either productivity level or at least my

43:37

happiness level when I'm due. Oh,

43:39

this is actually cool with it. Say it. Oh, this

43:41

sounds about right for me and

43:43

my taste. So I really enjoyed

43:45

that on the personal side of that's been

43:47

the most kind of integrated

43:50

habit that have had with

43:52

the the GPT four chatbot so

43:54

far. Yeah. It is

43:56

really interesting how it how it translates from

43:58

English to Chinese because, like like,

44:01

you never get broken sentences.

44:04

Like like the English and the Chinese,

44:06

it's never like it like, sometimes

44:08

it doesn't capture the meaning as well,

44:10

but it's always like a sentence that a real

44:13

person would write, which is not something

44:15

that you get from Google translate or

44:17

Baidu translate or Deepel or whatever all

44:19

these other ones that I've tried and been, you know,

44:21

success absolutely disappointed by. And I think that's

44:23

just like, it's it's a really

44:25

remarkable thing and

44:28

it's kind of incredible just imagining

44:30

where we're gonna be six months or

44:32

a year from now because I think we're

44:35

we're definitely in sight of

44:37

getting to, like, ninety nine

44:39

point five percent where,

44:42

you know, maybe maybe pretty soon

44:44

the only stuff that you don't end up needing

44:46

to go over a second time is

44:49

like, you know, let

44:51

me put it this way. Like, I can envision

44:53

a world where all that's left

44:55

that really needs humans to look at

44:58

is like literature and poetry. That

45:00

really has those like double, triple meanings

45:02

you're gonna really wanna capture that maybe the computer

45:04

might struggle with. But sort

45:06

of the the stuff that I do on trying to talk

45:08

of, like, yeah, I'm translating some WeChat

45:11

article or whatever. And, like, getting

45:13

the gist is much important, then much more

45:15

important than, like, the deeper subtlety. We're

45:17

very close to not needing a human in

45:19

the loop in that --

45:20

Yeah. -- at all. That's right. And I think the two

45:22

points I'll make that Jordan, one is

45:24

that I think the cool oil coaguleness

45:27

right, that we're talking about with the Chinese translation

45:29

of chat CPD4 really speaks to the

45:31

unique power of large language models,

45:34

more than any other approach

45:37

AI or merchant that we've done

45:39

so far. Right? Because the large language

45:41

model is all about basically mimicking the

45:43

most quality data

45:45

that's put it into the model

45:48

as training and then output something that is

45:50

closest to whatever was modeled

45:52

before. And because a lot of

45:54

the training data is, you know,

45:56

quality, I guess, Internet content or whatnot,

45:58

it just kinda sounds like Internet writing.

46:01

Right? At least to me. Yeah. And I think that's

46:03

just really fits my need in particular because

46:05

all I'm doing with Internet writing. Right?

46:08

You're doing it's mostly Internet writing.

46:10

So that really makes me happy to, like, boost

46:12

my productivity. Even though all the other,

46:14

you know, tools might be technically correct,

46:16

but they just sound so mechanical. And the

46:18

other thing too, I'll add that we can pin this for

46:20

maybe another day that I made thinking a lot about

46:22

is that Chinese

46:25

as a language, especially the online

46:27

version of Chinese, it's

46:29

much more subtle There

46:32

are, like, three, four different Tundras

46:35

and, you know, like, CAE and

46:37

an audience at a way to communicate topics

46:39

that are hard to a cupcake or maybe

46:41

a lot fun to communicate. Right? So

46:43

one thing I've been thinking about a lot is,

46:46

is there just gonna be a limit to how

46:48

good a Chinese chatbot,

46:51

whether it's early from Baidu or some other

46:54

AI model would be because the training

46:56

data itself that they can access is

46:59

just more subtle. So the output is

47:01

hard to make the output subtle or

47:04

precise rather when the training data

47:06

is so subtle. And

47:08

that's a very kind of a specifically Chinese

47:10

Internet

47:12

characteristic, if you will. Yeah.

47:14

I mean, my bet is that we're

47:17

on this weird exponential curve and

47:19

like anything. Like, that sort of

47:21

stuff is gonna get smoothed out just by

47:24

scaling what you can train

47:26

on. Though I,

47:29

you know, it it it's it's an interesting

47:31

question, you know, to what extent, like,

47:34

it's like something intrinsic about Chinese

47:37

the language itself versus you

47:41

know, versus, like, all the other challenges

47:43

we talked about about export controls or,

47:45

like, you know, not having access to the best algorithms

47:47

or the best, you know, the best computer scientist.

47:49

Or whatever. That's gonna be that's gonna be

47:52

something that may hold Chinese firms

47:54

back as they're looking to roll out similar

47:56

technologies to to

47:58

open AI.

48:00

Yeah. Yeah. Anyways, that was

48:02

another that it's

48:04

been stuck in my brain as a

48:07

unique characteristic. Especially to the Chinese

48:09

Internet per se. Right? Maybe not, like, Chinese

48:11

the language itself, but the the

48:14

Chinese Internet

48:14

culture. So, yeah, man. You know, there this has been

48:17

something about I got one more. So there

48:19

was something about ByteDance. Like,

48:22

I think this was like twenty eighteen or twenty nineteen.

48:25

Jiaoming decided that everyone in by

48:27

dance had to learn English and, like, they were

48:29

gonna try to run the the company

48:31

in English and have all the senior level meetings

48:33

in English. And, you know,

48:36

like a lot of Chinese people, even

48:38

highly educated ones, are not super comfortable

48:40

doing English language meetings.

48:43

And, you know, that was this weird

48:45

bottleneck. And I think it's something that every

48:48

global company company with global

48:50

operation struggles with is, like, you know, what language

48:52

are we gonna operate in and whatnoted? And

48:54

and watching the likes of of

48:57

of GPT four just

49:00

really push, you

49:02

know, just just going beyond what

49:04

what we thought was the state of the art with, you

49:06

know, Google translator or what have you. And

49:09

getting us very close to

49:11

the sorts of babblefishes where

49:14

we we speak something and

49:16

it will come out in another language in

49:18

a sort of, you know, the the sound will

49:20

come out in another language so that you can have, like,

49:22

real time meetings with ten different people

49:25

and ten different languages being able to all communicate

49:27

with one another. Like, we're not that far away

49:29

from from from that.

49:32

And it's a it's a real

49:34

it's a it's a real weird irony

49:37

that, like, the tools that would make US

49:39

and Chinese And

49:42

and just more global sort of international

49:44

collaboration business collaboration are gonna be

49:46

coming precisely at time

49:48

where, you know, we're we're we're leaning

49:50

towards a much more globalized world. Yeah.

49:53

I think us as synthetically live

49:56

translating, you know, from one language

49:59

to multiple language will become a pretty

50:01

solved problem with AI models

50:03

very soon. And

50:06

that will just make any executive know

50:08

if you're Chinese who wants to run the company

50:11

so much. And one thing I did

50:13

here just as a fun anecdote is that

50:15

the Alroboe who is know, the current CEO

50:17

by dance, you know, Johnny Mings

50:19

Bestie and, you know, university roommate,

50:22

I think, has made it his personal

50:24

OKR to study

50:26

English and that's well publicized within

50:28

the

50:29

company. So he is trying.

50:32

Yeah. But maybe maybe maybe

50:34

He doesn't have to try as hard anymore. Maybe you don't I

50:36

don't think he I don't think he has to try hard anymore. I

50:38

don't think he's gonna give you a bigger deal. It's so

50:40

weird. I spent so much freaking time learning Chinese

50:42

all of a sudden, I don't

50:43

like, do I even need it anymore?

50:45

Well, I think so. You don't need to be able to rip

50:47

it. Right?

50:50

Yeah. Yes. III

50:52

need to. Right? But what I am

50:54

doing is a very specialized thing,

50:57

right, of, like, trying to under stand,

51:00

like, at the, like, highest possible

51:02

level of nuance, like, US

51:04

Chinese relations and and politics

51:07

or whatever. But I think, like, for any company

51:10

where, like, that is not the game you're

51:12

playing. You're trying to sell widgets. You're trying to sell

51:14

software. You're trying to, you know, market

51:16

or this that and the other thing. Like like,

51:19

you don't need the you

51:21

don't need to like get to a hundred

51:23

percent comprehension. Like, you'll be totally

51:26

fine with ninety seven percent. Like, I

51:28

don't think you'd you'd lose any sort of like business

51:30

efficiency if you can just get to that

51:32

place. And that's how, you know, businesses around

51:34

the world have been running. Right? It's like they find they

51:36

find someone who's, like, sort of by cultural

51:38

or, like, you know, is maybe native in one language

51:41

and, like, pretty good in the other and, like, that's the person

51:43

they they have to set up their local operations. And

51:45

and that person provides a sort of like interface.

51:48

But, you know, those folks aren't linguists.

51:51

Right? They're business people who happen to

51:53

have some sort of language skills because

51:55

they studied it or because of their, you know, heritage

51:57

or what have you. And and that having

52:00

that thing just like not like

52:02

having, like, the language. I mean,

52:04

I don't know. Maybe it'll

52:06

still matter. Maybe there there there'll be enough humans

52:09

in the loop that, like, you know, to, like,

52:11

figure out what it's gonna take motivate employees

52:13

in one country or another or

52:15

or maybe subtle sort of like user

52:17

user habit things. I

52:20

think the kind of whatever

52:22

native means will still be

52:25

important. But there's a lot that won't,

52:27

which is kind of wild to think

52:29

about.

52:30

Yeah. I think we can both comfortably

52:32

say that drafting and the reading

52:36

spam or marketing code pictures will

52:38

be so much easier in any language

52:40

going forward, no matter which

52:43

country you wanna do your

52:44

business. Yeah. With that,

52:47

Thanks, Troy. This has been awesome, man. Thanks

52:49

for having me. Come

54:21

on.