0:00

This episode is supported

0:02

by OFX Money Transfers. Currency

0:04

exchange can feel complicated. At OFX,

0:07

they keep it simple. With over 20 plus

0:09

years in business, OFX provides fast,

0:11

secure, and reliable money transfers. They're

0:14

OFX-perts, a stay on top of global

0:16

markets 24-7 so you don't have to,

0:18

giving you more time to focus on your business and

0:20

not on exchange rates. OFX

0:23

is a U.S. licensed money transmitter and

0:25

registered money service business. Reduce

0:27

your FX costs today. Go to OFX.com

0:29

to get started.

0:35

Support

0:38

for On with Kara Swisher comes from another

0:41

podcast. It's called The Science of Scaling,

0:43

hosted by Mark Roberge, a senior

0:45

lecturer at Harvard Business School.

0:47

Each week, he interviews some of the most successful

0:50

sales leaders in tech. They get tactical

0:52

about how to grow a business. Because Roberge

0:54

will tell you whether a company is valued at 1 million, 10

0:56

million, or a hundred million dollars,

0:59

the failure rate is the same. The

1:01

Science of Scaling wants to change that. Search

1:03

for The Science of Scaling wherever you get

1:05

your podcasts.

1:07

That's The Science of Scaling.

1:26

Hi everyone, from New York Magazine and the Vox

1:28

Media Podcast Network, this is Kendall

1:30

Roy pitching you the X Everything

1:32

app. Just kidding, this is On with Kara Swisher

1:35

and I'm Kara Swisher. And I'm Naima Reza

1:37

and I'm back. Yes, you are. So much has

1:39

happened in the tech world. There's a lot. There's always something

1:41

happening. No cage match. No cage... well,

1:44

that was never gonna happen. Besides you and Vinod

1:46

Khosla, which we'll get to in a better episode today.

1:48

Yeah. The rebranding

1:50

of Twitter to X. Yeah. A rebrand

1:52

more mocked than Meta's. The announcements

1:56

around

1:56

Elon's separate company XAI. And also

1:58

increasing regulatory

2:00

scrutiny from Washington on a couple of fronts, on monopolies

2:03

and on AI. Yeah, absolutely. Washington's

2:06

getting it together on AI more

2:08

quickly than before than anything else, and there is

2:10

a series of legislation happening around

2:13

tech. We'll see if it goes anywhere, it never has before,

2:15

but it's encouraging that it's early, and

2:18

there's a series of lawsuits, whether

2:20

it's Sarah Silverman, and there's more to come. People

2:23

like Barry Diller, I think, threatened to sue AI companies.

2:26

So I think people are thinking about not letting

2:28

it happen again when it happened

2:29

with the first round of tech taking

2:32

over everything. Yes, and the Sarah Silverman case is really

2:34

interesting because her argument is basically like, look,

2:36

my jokes are being ripped off. Copyright. It's a copyright

2:39

case, and it's not un-similar to

2:41

when Viacom sued YouTube, but

2:44

in this case, it's more significant because they're mashing

2:46

up other people's... It's plagiarism. I don't

2:48

know how else to put it. I keep saying that. Remember, I was talking

2:50

about copyright being the most important legal

2:52

thing, and I think that there'll be a lot of that, and these companies,

2:56

OpenAI just licensed AP stuff.

2:58

So there's a lot going

2:59

to be happening here. This is the very interesting thing

3:02

because OpenAI, having gotten out ahead of

3:04

the gate and having a first mover advantage, means that they

3:06

have actual customers' revenues, et

3:08

cetera, that can support their paying

3:10

licenses. Yeah, subscription. Yes.

3:13

As a result, they're able to build a bit of a mode

3:15

around them. They're able to afford these license fees. It'll

3:17

become very hard for smaller companies to

3:19

do so. And that, the ability of a company

3:21

like OpenAI to get so big begs the

3:23

question of bigness, which is this unlikely

3:26

alliance of Elizabeth Warren and Lindsey Graham.

3:28

Yeah, I know. The pair, the

3:29

two pair you want to go out to dinner with to go have

3:32

some fun. No, they're

3:34

getting together to try to do a number of things,

3:36

including create an agency around tech, which

3:38

I think is an interesting idea. It's something

3:41

I've thought of a lot. Every other industry,

3:43

whether it's Wall Street or whoever has agencies

3:46

around them, FDA, FTC,

3:49

every other major industry has agencies

3:51

attached to it, and tech does not. So the question is, should

3:53

there be a separate agency dealing with algorithms

3:56

and AI and privacy and

3:58

things like that? Right now the FTC

4:01

does that and the Justice Department, to an extent,

4:03

but it's all based on enforcement

4:06

and lawsuits. And so maybe there

4:08

should be someone who's more regulatory

4:10

on a day-to-day basis.

4:12

We'll see if it passes. I don't think there's-

4:14

Tech is really, in many ways, like when I think about your reporting

4:16

over decades and you talk about the

4:18

bigness and the power and the

4:21

lobbying power, the power over consumers

4:23

that companies have, strikes me that it's

4:26

a history of capitalism in America. But do

4:28

you think it's something very endemic to tack that

4:30

it requires something

4:31

to- No, they're just like regular capitalists

4:33

and they deserve regular regulators, that's all. They're

4:36

not special by any means and they're just always

4:38

trying to make money and they have

4:40

more power than, well, I

4:42

mean, the railroads had power right and then they didn't. You

4:45

know, the television industry

4:47

was a very, they're all consolidated.

4:50

So there's all kinds of versions of this, but

4:53

they're more powerful. And we live in an age where

4:55

just the distribution mechanism is that they are celebrated

4:58

so much. We've lionized a lot of these

5:01

founders and also a lot of these venture capitalists

5:03

over the years. And Twitter has been

5:05

a big part of that or X, whatever it is, but

5:08

has been a big part of it, boosting the voices of

5:10

these people. Yeah, Senators Warren

5:12

Graham did have a piece in the New York Times last

5:14

week saying, when it comes to big, tough, enough

5:16

is enough. I think that's probably a right thing. I just

5:18

don't think it's gonna pass. That's all. You know,

5:21

there's not a taste for new agencies, but

5:24

I think they're right in this one line. Nobody

5:27

elected big tech executives to govern anything,

5:29

let alone the entire digital world. If democracy means

5:31

anything, it means that leaders on both sides of the aisle must

5:33

take responsibility for protecting the free, marital people, ever-changing

5:36

whims of these powerful

5:37

companies and their unaccountable CEOs. I think that's

5:39

absolutely right. And Barack Obama said

5:41

a similar thing. Tech is not gonna solve our

5:43

problems or the really hard ones. They just leave

5:45

us the hard ones and then they take all the good stuff.

5:48

But with President Obama, one of the critiques that

5:50

he's had for many people, including yourself and

5:52

myself, it's like, well, we

5:54

let it get too big. Well, at the end, he

5:57

changed his tune quite a bit. He's changed his tune. So

5:59

let him do it.

5:59

So our guest today is a longtime venture

6:02

capitalist, Vinod Khosla, someone who's been in the

6:05

Valley for ever, I don't know, 40 years.

6:07

I don't know, a long time. He was a co-founder of Sun

6:10

Microsystems in the 1980s, a

6:12

former Kleiner Perkins partner in the 1990s,

6:14

and then started Coastal Ventures, his own fund

6:16

in this millennium. And he's also one

6:18

of your favorite people to battle

6:21

with on Twitter, Kara. Explain. Yeah,

6:23

no, I don't battle with him as much. I battle with much different

6:25

deals. But he's very, you know, I've known for years,

6:27

you know, he's been at the forefront of a lot of stuff.

6:29

He's been very early. He's a genuine techie

6:32

and a genuine business person and really

6:34

does think way ahead. He also is also

6:37

willing to debate, which many of these people, they get to a

6:39

certain point and they just don't want to be bothered by irritating

6:42

reporters who may know as much as

6:44

they do. He's recently been doing a

6:46

lot more. And we have been talking about doing this interview for a

6:48

while, but you are a little reticent. I really

6:50

wanted to do it. You did not want to do it. Explain.

6:53

I have come around. I thought it was a really great interview.

6:56

And I think he's I wanted

6:58

to do it sort of I

6:59

would rather do a startup person, a new startup

7:02

person in AI and lean forward. But that's just

7:04

my tendency. You know, I just think he's

7:06

he's a he's a legendary venture capitalist. If we

7:08

did John Dewar, he's in that range of

7:11

John Dewar and him and several others.

7:13

Reed Hoffman, I'd say so we might as well do

7:15

Vinot. Yeah, I thought he was important for two reasons.

7:18

And one is his early work on climate and AI,

7:20

which he started writing about AI as

7:22

early as 2011. He did. And

7:24

he's writing it as the next platform after mobile, which

7:26

was early to that game. And then the

7:28

second reason I think he's a really

7:29

interesting character is because of his

7:32

global view. I having spent time in Silicon

7:34

Valley, Stanford, it

7:36

in at Santel Road, I would say that

7:39

people like there's a there's less

7:41

globalism in Silicon Valley than I would expect

7:44

for what has been the frontier

7:46

creator of technologies.

7:49

And there's been this kind of like build and they will

7:51

come philosophy. Someone

7:53

like Vinot has an appreciation, I

7:55

think of the geopolitics and the role

7:58

of China and India.

7:59

giants and the competition from them as

8:02

well. Yep, he does. He definitely has much more

8:04

of a global, most people in Silicon Valley just are

8:06

American, American, American. So, you know, and

8:08

he is too, he's invested in mostly

8:11

US companies, but he certainly has

8:13

a global perspective. All right, we'll hear that global

8:15

perspective when we get back. Let's take a quick

8:17

break and we'll be back with Fenod Kosa.

8:21

Support for ON with Kara Swisher comes from Unisys.

8:24

Sometimes you hear people say, get your head out

8:26

of the clouds. Well, Unisys disagrees.

8:28

They say, keep your head in the clouds,

8:30

keep dreaming and breaking through by getting

8:33

your business the best cloud you can. Unisys

8:35

offers app modernization that helps companies move

8:37

their apps to the cloud, build new

8:40

cloud-needed apps, and support for on with Kara Swisher is

8:43

a way to do that. Unisys is a

8:45

global company that's committed to helping

8:47

you get your business move their apps to the cloud, build

8:49

new cloud native apps, and unlock powerful

8:51

insights and performance. With Unisys

8:53

cloud solutions, the sky is not the limit. It's

8:55

just the beginning of unlimited potential. That's

8:58

the place where one breakthrough leads to another.

9:00

Unisys, keep breaking through.

9:06

This episode is brought to you by Shopify.

9:09

That's the sound of switching your business to Shopify,

9:12

the global commerce platform that supercharges

9:14

your selling. Harness the best converting checkout

9:16

and same intuitive features, trusted

9:19

apps, and powerful analytics used by

9:21

the world's leading brands. Stop

9:23

leaving sales on the table. Discover

9:25

why millions trust Shopify to build,

9:27

grow, and run their business. Sign

9:29

up today for your $1 per month trial period

9:32

at shopify.com slash tech23.

9:36

It is on. All

9:38

right. How long have we known

9:40

each other? Long, long time. Many

9:43

decades, more than- 25 years. Yeah,

9:46

I think I met you in the 90s when I got out of here

9:48

in 1996. I think you're one of the first people

9:50

I called because you were obviously well-known for

9:52

creating all kinds of companies, but then you were in venture

9:55

capital at the time. I visited two people, you and

9:57

John Doerr here on Sand Hill Road, where you

9:59

still are.

9:59

Still on Sand Hill Road. Yeah, we are still on Sand Hill Road.

10:02

Where were you at the time? What firm were you at? I

10:04

was a Kleiner. You were a Kleiner. I was

10:06

a Kleiner from after Sun when I joined

10:08

the world. Sun Microsystems. Till 2004 when

10:10

we started Coastal Ventures. But

10:13

actually we started it within Kleiner Perkins.

10:15

So the first two years we stayed

10:17

inside their building and then got

10:20

to where.

10:20

Which was just down the road and I visited the two of you. I

10:22

remember that. A couple of different people

10:24

I visited, but you were among the first. But I'm

10:26

going to start talking about, we're going to talk about a lot of things. We're going

10:29

to talk about AI, geopolitics, climate

10:31

tech, your tweets. I'm very interested in your latest

10:33

tweets if you know it. I'm not sure what's going on. I'd like to hear

10:35

about it. But we'll get to that at the end. But

10:37

you were very early to the topic writing about AI

10:40

for TechCrunch in 2012. Mustafa

10:43

Suleiman told me in a recent interview

10:45

that if you look at the pace of change in

10:47

AI, it's actually been quite incremental. We've been

10:49

working on this for

10:50

decades. It's data and computing power that have grown

10:52

exponentially in the last few years. I'd love

10:55

your thoughts about why you started

10:57

to write about it so early.

11:00

Are we about to see exponential improvement in

11:02

AI in the years to come? Let

11:05

me start with an analogy. When we first

11:07

met in 1996, that's the year

11:09

we started Juniper.

11:10

Explain what Juniper does for those who don't know.

11:12

Juniper is a TCP IP router

11:15

company or a router for the internet. It

11:18

was a time when nobody in the world

11:20

believed that TCP IP would be the

11:22

protocol for the internet. Cisco

11:24

had gotten essentially all

11:26

TCP IP plans off their docket and

11:29

bought a company called Stratacom. The

11:32

reason why that's relevant, though, is

11:36

Sun adapted TCP IP in 1982. It

11:39

was puttering along at a slow pace like

11:42

AI was puttering around for a while. Then

11:45

in 1996, when we started Juniper, it

11:47

saw an inflection. When

11:50

we started Juniper, not a single customer

11:52

said they would buy it. We built it anyway,

11:55

very much in the field of dreams kind

11:57

of vision of things.

11:59

clear then that

12:01

this puttering low end of

12:04

the exponential was

12:06

already happening.

12:07

The reason I bring that up and

12:10

why Juniper was such

12:12

a large payback for Kleiner, about

12:15

$7 billion in distributed prospects on a few

12:17

million dollar investment, was

12:20

we guessed right on the exponential.

12:24

For 10 years, I've watched

12:26

AI do the same thing. Right. Putter

12:29

around incrementally. When

12:31

I first wrote my piece on Do

12:33

We Need Doctors? Do We Need Teachers? in 2011.

12:37

It was December 2011 when I

12:39

wrote it, January 2012 that I published

12:42

it. It was clear

12:44

AI was going to be large

12:47

to me. To you. And at

12:49

some point, capability would explode

12:51

for two reasons. One, there was effort

12:54

being put in, but the best minds were moving

12:56

to AI. And the potential

12:58

existed for very large progress.

13:01

Right.

13:01

So let's talk about that, the future impact.

13:03

First, health care. In 10 years, you

13:05

have talked about this and you caused

13:08

quite a bit of controversy when you said it. And I think you

13:10

were correct at the time of how A will change the experience

13:12

of medical care.

13:13

Well, it's very,

13:15

very clear that expertise

13:17

will be very close to free in

13:20

a GPT-5, GPT-6

13:23

world.

13:24

And if you imagine that, then

13:26

a physician's expertise will

13:29

be in an AI. Definitely in the

13:31

next five years, whether

13:33

you're talking about a primary care doctor,

13:36

an oncologist, so specialists,

13:39

or a mental health, a therapist,

13:42

a psychiatrist, those will

13:44

be in AI. Whether we allow them

13:46

to practice or not is a different regulatory

13:49

question. We can get to that. But

13:51

the capability will be there and many

13:54

countries around the world will be using that capability.

13:56

So a doctor that's not a person. A

13:59

doctor that's not a doctor. a person having

14:01

more knowledge than the median

14:03

doctor or 90%

14:06

of doctors probably where I would set the benchmark

14:08

in 5E. Wow. So

14:12

everybody can get a quality doctor. And

14:15

it'll be in conjunction with a doctor, possibly

14:17

for the human element of care.

14:19

Well, she's looking in your eyes and telling you that things. Looking

14:21

in your eyes, giving you a hug when they

14:24

have to tell you you have cancer. But

14:27

otherwise, an AI will know more and have

14:29

better intuition or whatever

14:32

you want to call it. It will have a broader range

14:34

of knowledge, better knowledge of state

14:36

of the art, research, and

14:38

outcome. The best therapies,

14:40

every possible interaction. Do you know

14:42

drug interaction kills more people than breast

14:44

cancer in this country this year? And

14:47

so you can avoid those, those avoidable

14:50

errors. And that should happen

14:53

if the medical establishment lets

14:55

it happen.

14:55

But 90% of doctors, meaning almost

14:58

all of them. Almost all of

15:00

their expertise. Right. So

15:02

what's their use of the 90% that gets

15:05

their expertise useless? The number I've used

15:07

in the past is 80%

15:08

of what doctors do will

15:10

be done by an AI. But

15:13

the 20% will provide much more of the human

15:15

element of care, which they don't have time for today.

15:18

Right. You know, to make you feel like they care. So

15:20

they're still useful. In other words, they're still useful.

15:22

I think most AI systems will be used in

15:24

conjunction with humans and

15:26

humans who care about what they're doing.

15:28

Okay. One of the things you

15:30

told Sam before in 25 years, 80% of all

15:33

jobs are capable of being done by AI. You

15:35

called this quote, the opportunity to free humanity

15:38

from the need to work, though you admit

15:40

it is, quote, terrible to be the disrupted

15:42

one on the way to utopia. Talk

15:45

about that. How is that level of job loss utopia?

15:48

Now I have said a similar thing. Like there's

15:50

no reason for many people to be doing

15:52

what they're doing.

15:53

So let me be precise. What I said is 80%

15:55

of 80% of all jobs. So 64%

15:58

of all jobs. OK,

16:00

yeah, still a lot to know. Yeah, a

16:02

lot of jobs will be displaced. Now,

16:05

we've gone through that displacement before. You

16:07

know, in 1900s, agriculture was 50% or

16:10

so of all jobs. By 1980,

16:12

it was 4% of all jobs.

16:15

But I do think this time it's different because

16:17

we've exceeded the capability of

16:20

human beings, if I'm right, in the next

16:22

five years. And most

16:24

of these jobs will not need to be done

16:26

by a

16:27

human.

16:29

And humans will have other jobs

16:31

to do or other things to do. More

16:34

importantly, there'll be enough abundance. And

16:36

I wrote a piece about 2014, where

16:39

AI will cause great abundance,

16:42

great productivity growth, great GDP

16:44

growth, almost everything economists

16:47

measure, and increasing

16:49

income disparity.

16:50

Because people don't have jobs, right? People

16:53

don't have jobs. So what do we do with all those people?

16:55

Now, the good news is the abundance

16:57

will allow for redistribution. If

17:00

you take a simple metric like GDP growth for

17:02

the next 50 years in this country, and you

17:04

assume 2% a year, GDP

17:07

will go from 70, 75,000 to 175,000, that 2%

17:11

per year real growth over 50 years. If

17:15

AI accelerates that, turbocharges

17:17

it to 4%,

17:19

the GDP growth will be $475,000.

17:23

If that's the case, there's ample room

17:25

for redistribution, something

17:27

the Republicans hate. But I do

17:29

think it'll become essential. And there

17:32

will be enough to afford the

17:34

minimum standard of living for everybody

17:37

in 2004. To pay them.

17:39

Pay them to live. Yeah, pay them to live

17:42

or do things that are useful, but not in

17:44

today's jobs. And so when you think about

17:47

it that way, that's a profound shift

17:49

in the way we live, which the

17:51

US particularly has not been open to. How

17:54

do you make that shift without it being the

17:57

farming change was very, we

17:59

don't remember it.

17:59

but it was quite difficult from a social perspective

18:02

and people were very dissatisfied

18:05

to say the least. The same thing's gonna happen here. We're

18:07

already in a politically polarized.

18:09

You know, it's romantic to talk about disruption.

18:12

Like I said before, it's not fun to

18:14

be disrupted. Right,

18:15

so what has to happen then

18:17

if people, like you don't need to exist

18:20

anymore in the work sense

18:22

at least. You know, most people

18:24

don't want the work they want. Do you wanna work on

18:26

the same assembly line, assembling the same

18:28

car for 30 years in a row? No.

18:31

Nobody wants those jobs. They have

18:34

a need to work and I

18:36

think the need part will go away. I'd

18:39

be quite happy maintaining a hiking trail

18:41

out there. That's a great

18:43

job for me.

18:44

Right, which AI cannot do. Which

18:46

AI probably can't do. A robot could.

18:48

Or I may still prefer to do it even if an AI

18:51

can do it. There are jobs people will want

18:53

to do.

18:54

I love my job. You love

18:56

your job. Most people don't work

18:58

in a job they would do for free if they

19:00

didn't need to.

19:02

So would AI replace venture capitalists?

19:04

Could they figure out investments better than you? You

19:07

know, can AI do investments? Possibly.

19:12

You're pausing. Let's get rid of the doctors.

19:14

But venture capitalists are a special group of hummingbirds.

19:16

No, I don't subscribe to this.

19:19

My job is special. Right. You

19:21

know, a sufficiently advanced intelligence

19:24

will make good investments. Even

19:26

if we don't need to make investments. The

19:29

pause was in the world

19:31

of great abundance,

19:33

capitalism needs to be redefined. Capitalism,

19:36

as we've seen it, is great

19:39

for economic efficiency.

19:42

People compete, they have to get better,

19:44

things become more efficient.

19:47

Capitalism works

19:50

when there is no abundance. When resources

19:52

are scarce and you have to make the most of them.

19:56

When in fact, abundance becomes easy.

19:59

Then you have to...

19:59

rethink the role of capitalism.

20:01

I'm not

20:02

saying we need to go to socialism, I'm

20:05

a total capitalist. But

20:07

how we think about the role of capitalism

20:10

will transition over the next 30, 50 years, probably

20:13

within our lifetime.

20:14

Within our lifetimes. All right, let's talk about OpenAI.

20:16

You were the first VC to invest. People don't realize

20:19

when the company switched to its so-called capped profit

20:21

model in 2019, Reed

20:24

and Elon Musk and Reed Hoffman were

20:26

already in there who invested in it as a nonprofit.

20:29

Why did you wait to invest in it when

20:31

you were writing papers about it as early as 2012? What

20:35

was the wait for your perspective?

20:37

Well, let me be clear. We were

20:39

investing in the best AI we knew

20:42

how to invest. The picture of how it emerges

20:44

wasn't clear. It was pretty fuzzy. I

20:46

don't believe in big vision.

20:48

I'm sort of like you engage

20:51

and you learn.

20:52

Now we invest in a deep learning company called

20:54

Jetpack which failed, which would

20:56

classify Instagram images and

20:59

do a pretty good job of saying, I need

21:01

X, give me some pictures of that. The

21:03

company failed. It was acquired by

21:05

Google and that's fine.

21:08

We invested in Caption Health that started

21:10

to read, essentially guide ultrasound.

21:14

And we invested in a company that was

21:16

self-driving for MRI

21:18

machines. So you could do a cardiac MRI

21:21

without an MRI technician. So

21:23

we were investing for a while

21:25

in these areas. When the

21:27

OpenAI came around, there really wasn't

21:30

a mechanism to invest in sort

21:32

of the core platform. It didn't exist.

21:35

But as soon as I remember Sam

21:37

calling and asking if we were interested,

21:40

and Sam wasn't gonna take venture capital firms

21:43

because it wasn't clear where this was

21:45

going or what would emerge. He wanted individuals

21:47

who cared about AI and

21:50

its trajectory. Right.

21:52

So explain how a small cap structure

21:54

works very briefly. To me, that's a

21:56

technical detail. It's a 100X, you

21:59

get up to...

21:59

100x your investment, not more, and

22:02

the rest goes for societal good.

22:04

And Sam called us because he thought we

22:06

cared about societal good as

22:08

much as we cared about sort of the profits.

22:11

100x is pretty good. 100x I'd

22:13

be happy with any day. What

22:15

really was the risk was where would

22:17

AI go? Would it ever generate any

22:20

profits? What would be the capability?

22:23

And would this be the way to get to

22:25

scale in AGI? In

22:27

order to have the money to do so. Now, Elon,

22:29

who was the early investor too in the nonprofit,

22:31

tweeted that OpenAI had become quote, a closed

22:33

source maximum profit company effectively

22:36

controlled by Microsoft. Not what I intended

22:38

at all. Basically saying it was a bait and switch.

22:41

He's now today started his own competitor. He's

22:43

talked about it and made a lot of noise about it. Why

22:46

is it not a bait and switch to go from nonprofit to

22:48

problem even if capped? And

22:51

what do you think of his new effort?

22:53

Well,

22:54

so I haven't asked Elon

22:56

about his effort. You're in a better position to ask

22:59

him.

22:59

No, he's not speaking to me, but go ahead. But

23:03

at one level, he's calling for a pause

23:05

in research. And the other he's accelerating

23:07

his efforts. It

23:11

speaks volume on the importance of AI. Whether

23:15

it's his self-driving cars or

23:18

Twitter or other mechanisms. I

23:20

think he believes just like I do,

23:23

like Sam does, everybody does on the

23:25

strong power of AI to be a powerful

23:27

economic force.

23:29

And so, you know, it behooves him and

23:32

others to do the most they can.

23:34

What I would say is when there's such a strong

23:36

economic force, lots of people will play and

23:38

that's a healthy thing. And so when he

23:40

says that, how do you respond? Well, too bad.

23:43

Now it's too big, right? Is that essentially now

23:45

it's a thing? Well,

23:46

I don't think he quite imagined

23:48

it would be what it is. So in some

23:50

ways, these are, I think, sour

23:53

grapes, you know, and

23:55

he's trying to catch up.

23:56

He's trying to catch up. I was interested

23:58

as someone asked me, I said, well, he was... early and then he's late.

24:01

Now he's late, which is kind of an interesting thing. What's

24:04

your relationship with OpenAI now?

24:06

You're just one of their investors,

24:08

correct? Well, I've always had a great

24:10

relationship with Sam. With Sam Altman. We

24:13

share a similar vision on

24:15

the role of technology in

24:17

society. I think that's

24:19

why he invited us and probably the

24:22

only venture firm he invited to invest in

24:24

OpenAI because we had

24:26

aligned goals

24:27

with what was the original purpose of

24:30

OpenAI, which is maximum

24:32

good, societal good. He says he's

24:34

earning and he told me in an interview, an immaterial

24:36

amount from his initial investment through Y Combinator.

24:39

Is that a fair statement? Is he not going to be

24:41

benefiting from this upsurge?

24:44

I think Sam is very direct and very honest,

24:47

so you can take his word for what he's saying. At

24:49

least I believe him when he

24:51

says something. He's

24:53

not benefiting from the success

24:55

of OpenAI, so he can be true to the original

24:58

objective.

24:58

Let's talk about government's role in

25:01

regulating AI. It's something that's been much discussed

25:03

around the world. There's efforts in China,

25:05

everywhere else. The US is a little slower, as always.

25:08

Sam told me there needs to be a global regulatory

25:11

body like the International Atomic

25:13

Energy Commission. Reed Hoffman

25:15

told me he would do the equivalent of the Blue Ribbon Commission

25:17

to declare a set of outcomes to prioritize

25:19

and a set to avoid. What's your idea

25:21

for regulation?

25:23

I think any international treaty

25:25

is a really bad idea. Okay, why?

25:28

It would be the equivalent of saying I'll trust

25:31

Xi and Putin with the future

25:33

of this planet.

25:34

Nuclear was very verifiable

25:37

when used. Biowarfare was

25:39

verifiable when used. These

25:41

things are verifiable. AI is not

25:44

verifiable. I'd be

25:46

shocked if China doesn't

25:48

have an effort to have bots talk

25:50

to every US voter in

25:53

the next election to the best

25:55

of its ability, one-on-one conversations

25:58

manipulating their views.

26:01

They'll definitely do it in Taiwan.

26:02

Certainly. So,

26:05

you can't track it. Yeah. So

26:07

the question you have to ask with regulation is what

26:09

danger are you trying to eliminate?

26:11

You know, there's sentient AI, which some people

26:14

talked about. I think... Killer

26:16

robots. Killer robots.

26:19

You know, it's silly to imagine today

26:21

a robot could get away from us

26:24

and really kill all humanity you would want to

26:26

and all those things.

26:27

What's the current plot of Mission Impossible in case you're

26:29

interested? It's great for fiction. Yeah.

26:33

But

26:34

that danger, I would venture to guess,

26:36

is a lower probability than the risk

26:38

of an asteroid hitting planet Earth in

26:40

the next 50 years or 100 years. So

26:45

it's a risk. I'm not saying it's not a risk.

26:48

It's evolving rapidly. So we have to keep

26:51

very close eyes on it.

26:53

But far greater risk than sentient

26:56

AI is

26:59

the risk that China

27:02

uses it to influence global

27:04

economics and hence global politics. They've

27:07

done a lot to

27:11

build influence globally outside

27:13

of China, which is new under President Xi. Yeah.

27:17

The Belt Road Initiative. Minerals, everything.

27:19

Minerals. They

27:21

have a trillion dollars in loans outstanding

27:24

to nations.

27:25

They own the place. They own the place.

27:28

So they have the physical resources, which are essential

27:30

to economic development. And then they offer

27:33

AI globally

27:34

for good.

27:37

But it'll come with the Chinese style

27:39

of political system. And I want Western values

27:42

to win over the Chinese political system.

27:44

And by the way, it's a reasonable assertion

27:47

for them to say most people will be much better

27:49

off under China's rule, under China's

27:51

political system. And they genuinely

27:54

believe it's a better system for society,

27:56

not just for President Xi.

28:00

They believe it's a superior system. I happen

28:02

to prefer our system.

28:03

So that's the doomsday scenario, that it's

28:05

China's century, in other words. Yeah.

28:08

I'm not as bullish on China long-term, but

28:11

the fact that they get ahead of us

28:13

in the AI race,

28:14

use it for cyber warfare,

28:17

real warfare in the field, intelligent

28:20

drones, all that stuff.

28:23

But more importantly,

28:25

in the social sphere, like influencing

28:27

elections and opinions,

28:30

but also real true economic

28:33

development. So you've said that that's

28:35

why, as Musk and others have said, that AI

28:37

should take a pause. You've advocated,

28:39

because of the quote, we need to win the race against China.

28:41

You've also worried about, as you've noted,

28:43

the existential threat. What does an

28:45

arms race with China

28:47

mean that we have to do? Obviously,

28:50

we have to do it. We have to fight with them

28:53

on this issue. There's no question

28:55

we are in a techno-economic war with China.

28:58

And by the way, there's a great book

29:00

I read recently called The Danger Zone,

29:03

which argues the next 10 years, say 2033,

29:06

China will peak.

29:09

And it argues that authoritarian

29:13

regimes get extremely dangerous

29:16

when they're reaching their peak power. Because

29:19

of demographics, because of slowing GDP

29:21

growth, this trillion dollars

29:24

in debt all comes due and will

29:26

not be repaid. So

29:28

it causes real consternation

29:31

internally in China.

29:33

And that makes them much more dangerous than

29:35

they would be if they were just happily growing

29:38

at 5% a year for the next 30

29:40

years.

29:41

So I do think they're particularly dangerous because

29:44

they will be peaking over the next decade

29:46

if I buy this book's thesis, which I do. And

29:50

because of that, they'll deploy their AI

29:52

to their benefit, both

29:55

for soft power and hard

29:57

power. So what does the US do then? You

29:59

say democracy. and freedom of speech is at stake, but

30:01

you're also gonna make a lot of money. Is democracy

30:04

at stake here or what should the US do?

30:06

What should the US policy be right now

30:08

on this? And I assume you're working

30:11

in concert with

30:12

government.

30:14

Well, I talk to government a lot, trying

30:16

to influence against

30:19

those who are calling for rapid

30:21

regulation or pauses in development

30:23

because the biggest danger, it's

30:25

not saying that we don't have a danger in AI.

30:28

We have a much bigger danger

30:31

if we fall behind China, especially

30:33

because of the broad economic

30:35

power it conveys to the nation that wins.

30:38

And there isn't this notion of one winner

30:41

takes all. It may not be, it might

30:43

be. I just wanna change the probability

30:45

that they win or get ahead of us.

30:48

We'll be back in a minute.

31:02

Hi, it's Kara. I know you're a curious person

31:05

who deserve answers to life's most fascinating

31:07

questions. So I have a podcast rec for

31:09

you. On Freakonomics Radio, journalist

31:11

Steven Dubner explores well everything.

31:14

It dives into the hidden state of business history,

31:17

culture and economics with the help of CEOs,

31:19

artists and Nobel laureates to reveal just

31:21

how interconnected these topics are.

31:24

Like how the CEO of Microsoft is tackling

31:26

the AI revolution or the recent

31:28

episode about ESG investors and

31:30

if they're actually helping the environment

31:32

and a new series on the economics of

31:35

whaling. That's a new fresh topic. New

31:37

episodes of Freakonomics Radio are available

31:39

every week wherever you get your podcasts.

31:50

Listen, even if you don't have

31:52

student loans yourself, you probably

31:54

know someone that does, which means you

31:57

also know the system is broken. But...

31:59

Have you ever wondered how

32:02

exactly it got this way?

32:05

There's a high demand for college because

32:07

of this huge economic shift in the

32:09

U.S. economy and the global economy. And

32:12

then you have this, like, you know,

32:14

free-flowing access to credit. And

32:17

then you had uninformed consumers. I mean,

32:19

this is just a recipe for

32:22

disaster.

32:24

Your student loan debt and then some explained.

32:27

That's This Week on the Weeds. Listen

32:29

wherever you get your podcasts.

32:34

You in March, just before TikTok CEO

32:37

testified for Congress, you hosted a dinner with Peter

32:39

Thiel in D.C. for investors and lawmakers

32:41

to press China as a techno-economic threat.

32:44

House Speaker Kevin McCarthy was there. What

32:47

was your goal? Because at that dinner,

32:49

Thiel compared TikTok to homelessness

32:51

in his speech. And he said that, quote, both are a

32:53

really obvious problem. I'd

32:55

love you to talk a little bit about that dinner and then

32:58

TikTok itself, because it's emblematic. TikTok

33:01

is sort of the shiny object. Here's

33:04

the broader perspective I

33:06

would give.

33:07

President Xi in the 14th five-year plan

33:09

in China specifically

33:11

called for winning the AI race and

33:13

the 5G race. Those

33:16

were two things he really called for. Now

33:19

ask yourself why? Because AI will

33:22

exert great economic power globally,

33:25

which is very much in their interest.

33:27

Winning the 5G race, they

33:30

already have their telecom

33:32

equipment in a

33:34

hundred countries, and that lets

33:37

them spy on nations and

33:39

individuals within nations the way they can

33:41

do it with impunity within China.

33:45

And so I think

33:46

5G race is about surveillance

33:49

of citizens everywhere. The AI

33:50

race is about economic power.

33:53

And TikTok is an extension of the

33:55

surveillance capability. And that may

33:57

not happen, but it's not a risk-free.

33:59

we can afford to have. So what do you think they should do? I'm

34:02

gonna point to TikTok only because I do think

34:04

there's a lot more important things happening like 5G

34:07

and everything else than that. But what's to be

34:09

done about something like TikTok because it's so symbolic

34:12

and emblematic of the issue. Here's

34:14

a very popular app. A couple of years

34:16

ago, about four or five years ago, I wrote a piece that

34:18

said, best product I've seen in a long

34:20

time, I'm using it on a burner phone, we gotta

34:22

get China out of here. And I got a lot of pushback

34:24

that I was anti-Chinese and anti-Asian and I

34:26

was like, no, I'm anti-Chinese Communist Party. But

34:30

now everybody's like,

34:30

oh, we should ban them in this. What is

34:32

your feeling of what to do about that particular

34:35

issue? I wouldn't have a question,

34:37

we should ban TikTok. Okay.

34:40

Just

34:41

because of potential danger of

34:43

it being controlled by the CCP,

34:46

the Chinese Communist Party. We have to do

34:48

everything we have to do for national security

34:51

and protecting our citizens and we are very

34:54

influencable. If you look

34:56

at China,

34:57

the Chinese Communist Party controls companies.

35:01

In US, it works the other way.

35:03

The companies control government. People

35:07

haven't realized how different those systems

35:09

are and what purpose these

35:11

influence vectors serve. And

35:13

we have to realize that.

35:15

You know, it's not about, I'm not saying any

35:17

of this will happen. I'm saying

35:19

there's a risk of it happening and it's

35:22

about adjusting the probability of

35:24

these things happening.

35:25

Yeah, you've criticized Apple and

35:27

Tesla's business in China. Explain

35:29

why it's a problem and what they should do differently. Well,

35:32

to be clear, I haven't criticized

35:34

Apple's business in China or Tesla's business

35:36

in China.

35:38

What I have said is they can't

35:40

be independent views on

35:43

what is the right thing for Western world

35:46

because of their interests are conflicted and

35:49

Elon Musk would get shut down in China

35:51

or in Russia.

35:52

So he can't say what he thinks. So he

35:54

can't say what he thinks. Neither can Tim Cook.

35:57

I think that's a very different statement. in

36:00

their own self-interest as capitalist

36:02

companies, they have to do

36:04

what's good for their shareholders. And that means

36:07

they can't be trusted to represent- American

36:10

interests. American interests. Should

36:12

they be there at all? Should they be there at all?

36:15

Or should they move somewhere else? Look,

36:17

in the current system, it's for every company

36:19

to decide what they do, not my choice.

36:22

Now, we don't have a China operation, and

36:25

there's a reason we've never had a China operation.

36:27

When I was at Kleiner, I'd resisted

36:30

Kleiner having a China operation. It

36:32

was only after I left in 2004 they

36:34

started the China effort. So I've

36:37

never been a fan because the Chinese

36:39

government and the Communist Party

36:41

will not let an American company win

36:44

long-term.

36:45

They will let them strategically win short-term,

36:49

but not long-term. That's correct. Okay. So

36:52

let's move to India and its role in tech

36:54

over the next couple of decades. Sam Altman

36:56

met with the Prime Minister Modi during

36:59

a recent visit to DC and reportedly discussed

37:01

collaborating on AI. I want to know what collaboration

37:04

would look like. His visit was controversial given

37:06

human rights records there. What do you

37:08

make of him?

37:10

Well, I think Modi has

37:12

done a very good job on the economy

37:15

and getting government aligned.

37:17

I completely disagree with his

37:19

Hindu without

37:21

movement bias. I

37:24

think the biggest danger for a country

37:26

like India and continued economic development

37:29

and GDP growth is leaving a population

37:31

behind. So leaving the Muslim population behind.

37:34

Even if you were only interested in the

37:37

self-interest of the Hindu population

37:39

or his constituency, I'd

37:41

make sure they're not left behind. 200 million

37:43

people left behind is a recipe

37:46

for social unrest and disaster which

37:48

could set back the economy. So

37:50

my view is on

37:52

that front, he's being

37:54

politically opportune but causing

37:57

a long-term problem.

37:59

But of course, politicians

38:02

are okay to postpone problems by a decade

38:04

or two. But it is the single biggest risk

38:07

in India's economic development, I believe, leaving

38:10

one population behind.

38:11

And so what's to be done? Because this

38:13

is not a solvable problem. Would you think it'd be

38:15

good? You're so well known in India to

38:17

take a stronger stand there.

38:19

Well,

38:20

I don't think it's an unsolvable

38:22

problem, especially

38:24

with the power it convenes on us for

38:27

coming full circle to what

38:29

India can do with OpenAI. First,

38:32

India has the talent base to develop OpenAI

38:35

or applications that could

38:37

do a lot of good. Free doctors,

38:39

free teachers, free oncologists,

38:41

free robots. I've

38:44

had a number of conversations along that line.

38:48

So there is a lot of good to be done, and you

38:50

don't need to impoverish

38:52

one population to give to another.

38:55

We will have enough abundance

38:57

to solve the income disparity problem

39:00

or the social distribution problem. I

39:02

think the minimum standards on the planet by 2050

39:05

could be awesome, the minimum

39:07

standard.

39:08

So is it important for you to take

39:10

a political stand in some more promising like

39:12

India if you want them to be on this journey?

39:15

Yeah. You know, where I spend my

39:17

time is a whole different question

39:20

we can get to. You know, I have to

39:22

trade off between working with Bob Mungard

39:25

on Fusion. We're working on a

39:27

public transit system that should be in every

39:30

one of the 4,000 cities around the planet

39:32

instead of just the 200 cities that have

39:34

public transit today. There's

39:37

enough important things I'm working

39:39

on. So it's a question of

39:41

how effective can I be?

39:43

By getting into a beef with Modi. You

39:45

want him to develop AI versus focused

39:48

on the problems you see. Well, I

39:50

think those are independent vectors.

39:53

And I think on the economic development

39:56

front and the technology development front, they'll

39:58

probably do a reasonable job.

40:00

I could help but probably

40:02

not materially. Can

40:06

I stay in touch a little bit,

40:08

but I don't allocate a huge amount

40:10

of my time to that.

40:12

To that, doing that, okay. You just

40:14

mentioned climate. Let's talk about climate. You

40:16

mentioned a number of projects, public transportation,

40:19

fusion. At the Breakthrough Energy Summit

40:21

in October last year, you said, quote, if we try

40:23

and reduce carbon by 2030, we'll have much

40:25

worse off than set a reduction target of 2040.

40:29

Slow down, explain.

40:31

I don't mean slow down in

40:33

the research and development science, but

40:35

I do mean forcing

40:39

un-economic solutions

40:43

is the wrong approach to solving the climate

40:45

problem.

40:46

I wrote a paper about 10 years

40:48

ago, maybe 15 years ago,

40:51

in which I defined what I call the Chindia

40:53

price.

40:55

It's the price for any technology

40:57

at which it would be

40:59

broadly adapted in India and China

41:02

against its fossil competitors. I think

41:04

this was in 2010. I think

41:05

Gates calls it the green premium, right? When

41:07

does it end? Like at

41:10

that price,

41:12

it gets broadly adapted. Above

41:14

that price

41:16

is going to be just showcase,

41:18

you know, dressings on the cake, not the cake

41:21

itself. Right, and so you think doing

41:23

that too quickly is a problem. You've had this

41:25

experience, you said in 2007, the

41:28

clean tech field was ripe for VCs and you said there are easy

41:30

pluckings. That was a tough time for you, those

41:32

investments. And I know you and I have talked

41:34

about that. You had investments in biofuel,

41:37

you had solar investments. It

41:39

didn't work out at the time. Was that what you're

41:41

talking about too early? No,

41:42

that's not what I'm talking about. Solar was actually

41:45

quite profitable for us. Each

41:47

had a different trajectory. Biofuels,

41:51

some things worked out, some things didn't.

41:53

And I think for aviation, long distance

41:55

aviation, for example, biofuels

41:58

is still the right answer.

41:59

And LonserTech is a pretty popular public

42:02

company developing it.

42:03

Were you there too early? Do you feel

42:05

regretful of these investments? Or do you?

42:08

Well, actually the climate investments

42:10

for us worked out fine. Some

42:12

of the solar things worked out fine. Some

42:15

didn't, some did. Enough did to

42:17

make it profitable for us.

42:19

LonserTech is a very successful company

42:22

in biofuels. Yes, we lost

42:24

some companies, but we made more money than we

42:26

lost,

42:27

which is the key metric in venture capital. No,

42:30

I

42:30

know, but when I interviewed Dor, he was

42:32

like, it wasn't the hit we wanted

42:35

it to be. He was also early in a lot of

42:37

those things and felt it was too early. Now,

42:39

Kleiner's strategy was different

42:42

than ours, right?

42:44

And I'd left Kleiner by 2004

42:47

and working on our own. Right, but he was doing it

42:49

separately. Yeah. QuantumScape

42:52

has been a successful investment in batteries,

42:54

and we bet on batteries. At the same time,

42:56

we bet on biofuels for cars and

42:59

aviation. Biofuels for cars,

43:01

lost to batteries.

43:03

Biofuels and aviation looks like the

43:05

lead horse still.

43:07

All right, so when the biofuel company you did back,

43:09

it's CURE, is it CURE, K-I? K-I-O-R,

43:12

CURE. It went bankrupt.

43:15

Fortune actually asked, and I'd love you to comment

43:18

on this, if it was, quote, evidence that fast-moving

43:20

venture capital investors are ill-suited

43:22

to tackle such technically demanding time-consuming

43:25

endeavors.

43:26

I disagree with that. Tell me why.

43:29

You know, QuantumScape, for example,

43:31

slow-moving, long time to develop,

43:34

but in a very, very large market,

43:37

has been very successful. LancerTech,

43:40

if you measure profits today, has

43:42

been very, very profitable,

43:44

makes up for many, many cures in

43:46

terms of our losses, which

43:49

is the nature of venture capital. High-risk bets,

43:51

small percentage,

43:53

win.

43:54

They make up for all the losses. And

43:56

I think biofuels will work

43:58

out for us, as worked out for us.

43:59

with LancerTech, batteries

44:02

worked out for us. We bet on both. It's

44:04

impossible foods.

44:05

Impossible foods has been very, very profitable.

44:08

Remember, we bought 50% of the company

44:10

for $3 million. So it's been, even

44:12

at depressed prices, it's a very

44:15

profitable effort for us. And

44:17

that's the nature of venture capital.

44:18

So you don't feel that you

44:20

were early to those or is this the time now

44:22

to really invest? One of the things, you brought up nuclear

44:25

fusion. This is the tech you've invested

44:27

in. Sam Altman just announced this back. Bezos,

44:30

Peter Thiel, Bill Gates. Talk about

44:32

fusion, for example, which Mark

44:34

Benioff has called the holy grail. Many

44:37

people think we're decades away from functioning fusion

44:39

power in plants. There are efforts to make small

44:41

modular nuclear devices. People

44:44

are thinking about nuclear, nuclear energy

44:46

plants to restart them. Where are

44:48

you

44:48

looking at in climate? So my

44:50

view is, and

44:53

I tweeted about this recently,

44:55

that fission will take longer to permit

44:58

than fusion will take to develop

45:00

from scratch. So

45:03

I connected with Bob Mungard

45:06

on starting Commonwealth Fusion when

45:08

he was a postdoc at the MIT

45:11

Plasma Fusion Lab.

45:12

So you think that's faster than any political?

45:15

My bet is that technology will be developed

45:18

faster than you could permit one

45:20

nuclear fission plant here

45:23

in the US

45:24

with all the risks associated

45:27

with it. So

45:29

I do believe that's a large solution.

45:32

It's not the only solution we are working in.

45:35

We have some investments in solar

45:37

still, but for dispatchable

45:39

power, which is reliable power when the

45:41

sun isn't shining and the wind isn't blowing.

45:43

Right, no, renewables aren't enough. The wind and solar

45:45

are great. Wind and solar are great. Wind and water are not enough.

45:47

We've invested in those, not a lot in wind.

45:50

Right. You need reliable dispatchable

45:52

power and for which there's two

45:55

technologies. One is fusion

45:57

and the other is geo-t

45:59

term often ignored, but

46:02

we've invested both and both seem to be

46:04

working out. And that I

46:06

think either path would provide great

46:09

amount of power economically. And

46:11

you're leaving out nuclear, small modular

46:14

devices, all these. So we've not invested,

46:17

I used to be bullish on nuclear. I'm

46:20

less bullish, so

46:21

because of the permitting cycles, the

46:24

social objections, the extension

46:26

and permitting, I think you could develop fusion

46:29

faster than you can permit one new

46:31

nuclear

46:32

plant. And there's no way to overcome that, because many

46:34

people think it should be. I don't think we need

46:36

it. I do think some of the risks are

46:38

real.

46:39

I do think we can address risks,

46:41

and we were investors in terra power

46:43

too. But I think fusion is close

46:46

enough

46:47

that it will far exceed

46:50

the scaling capacity of any

46:53

other technology. All right. Let's

46:55

talk, you did talk about tweet. That was a funny tweet, but

46:58

you're kind of active on Twitter these days. You're

47:00

getting a little trolly there, Vinod. What's happening?

47:03

What tweet are you talking about? I'm not even asking about your beach controversy.

47:06

How is the beach, by the way? What tweet are you talking

47:08

about? Oh, you have a lot. Let me read. You

47:10

want me to read you a couple? I really enjoyed your English major tweet.

47:13

You said most English majors lack a goal

47:15

or purpose. Why else would you major in your native

47:17

language? Justine Musk, ex-wife

47:20

of Elon Musk said most English majors are women.

47:22

So tech bros are basically saying that women who go

47:24

to college and study literature are lacking in personalizing

47:27

goals. No wonder we love

47:28

them so. Why

47:31

do this, Vinod? I'm just curious. It's been

47:33

a curious development. It was half

47:35

and just, but not really. Yeah, I know. I

47:37

do think when

47:38

you spend $100,000 or $200,000 for college in the US, you'd better get an

47:43

employable skill. Okay.

47:46

Now it won't be important once we...

47:48

Yeah, but you were saying we're not going to be working. Yeah,

47:50

if we aren't working and reduce the need for

47:53

employment, then we'll be fine. In

47:55

fact, I've said that more recently.

47:59

I've also... As I said, by the way, I'm

48:01

not against English in graduate degrees.

48:05

Just undergrad, purposeless,

48:07

people end

48:09

up with a lot of debt and a job invading

48:11

at a restaurant.

48:12

So pointless. So pointless jobs. You

48:15

don't like them just because... You know, English majors are not grammar.

48:17

It's about literature, but okay. All right. Why?

48:21

You know, and I'm not saying nobody should

48:23

read literature. And it's not everybody.

48:27

Half the people who do English will do great

48:29

because their knowledge, education,

48:32

diploma, Stanford,

48:35

stamp, whatever. But

48:37

I'm talking about the downside case of

48:39

the people who won't get jobs because they don't

48:41

have a skill. Now,

48:43

this will become less important in the age of

48:45

abundance in the eye if you buy my thesis.

48:48

But... So go

48:50

ahead and read Dickens. Go for it. It

48:53

just doesn't matter. What you enjoy if humanity

48:56

has the leisure of not needing to

48:58

work. Need to do it. But I was curious why you did

49:00

that. I'm fascinated by it. It was

49:02

unusual for me to see it because you

49:05

don't strike me as a dunker and

49:07

a meme lord or anything like that, a

49:09

dank meme lord. You don't have that reputation.

49:12

I'm

49:12

trying to encourage more kids

49:14

to do... Think about the right things. Okay. All

49:17

right. So it's not sort of the current trend of venture capitalists

49:19

opining on everything from Ukraine

49:22

to whatever they feel like talking

49:24

about that day. Look,

49:26

you can tweet because you're trying

49:28

to get the most number of followers. I don't try

49:30

and do that. When I have

49:32

time, I will tweet

49:35

to transmit some insight

49:37

that I have. Right.

49:38

Right. So you find it useful.

49:41

You find it useful. I find it useful. Sometimes

49:43

it's funny and that's okay. I think I

49:45

had a tweet around Elon and Marx.

49:48

Oh, fight. What's your feeling

49:51

on that? Seems juvenile. It's

49:53

funny. Okay. You know, I don't think

49:55

it's important enough for us to worry about it. If

49:57

it makes a few people laugh, then that's a good

49:59

time.

49:59

A few smiles is a good contribution.

50:02

All right,

50:02

okay. I think it's slightly toxic,

50:04

a lot of it. I don't spend a whole lot of

50:06

time on Twitter. No, I know, you recently said you probably joined threads.

50:08

Have you done it? What do you think about the competition?

50:11

You know, to be honest, I did join threads because

50:14

I think it could be an interesting

50:16

platform. I haven't done much

50:18

on threads.

50:18

Okay, so my last question, you've been in venture

50:20

capital for so long. How do you look

50:22

at venture capitalists now and their

50:25

use? Here we are in Sand Hill Road, the famous place.

50:27

It's changed drastically. Has

50:30

it changed for the good or bad? Do

50:32

you see innovation in your own industry?

50:35

Look, as industries mature,

50:37

they diversify. Venture capital isn't one

50:39

thing. My view is at different

50:42

stages in your life, you worry about different

50:44

things.

50:45

I'm most worried about maximum impact

50:48

and a better society.

50:51

And I think venture capital is huge

50:53

leverage on that.

50:55

Much better climate technologies, much better

50:57

transportation, much better medicine,

51:00

much better education. All those are really,

51:02

really exciting things to do. They

51:04

won't happen on their own. If you leave

51:07

the world to electric cars, then

51:10

it'll be on the GM schedule by 2035. I

51:13

think the forecast was we'd have 50,000 vehicles

51:16

or some ridiculous number. Then a

51:19

instigator of change like Elon Musk

51:21

comes along and causes everybody

51:23

to rethink.

51:24

I think that's possible in every area, from

51:27

housing to cars to transportation

51:30

to medicine, entertainment in every area.

51:32

And I'm very, very excited. Frankly,

51:35

it's much more fun than playing golf for-

51:37

So you're not retiring? I'm not retiring.

51:40

I, as long as health permits with

51:43

that one caveat, I'm not playing

51:45

golf, I'm not going sailing. I have

51:47

much more fun uses of my time

51:50

that really excite me.

51:51

Okay. All right, Vinod, thank you

51:53

so much. I really appreciate it. Thanks.

52:05

So you and Beno Kostla won't be going golfing

52:07

or sailing anytime soon, it seems.

52:10

We never would have, anyway. It doesn't matter. So

52:12

we're just continuing to do what we don't do together.

52:14

That's true. Maybe I'll do another interview 20 years from

52:16

now, since both of you will still be working. No,

52:18

I shall not. You will not be in the 80% of the 80%

52:20

whose jobs are taken. No,

52:23

never actually. I thought it was interesting

52:26

that he attributed the jump in AI not

52:28

to the transformer paper and not

52:30

to the increasing scale of computer

52:32

power, but to the idea that people are moving there.

52:35

That was a leading indicator for him, was that humans were moving

52:38

there. Yeah, I think he's right. It's a very VC

52:40

way to look at it. Well, that's how they watch

52:42

things, like where people are going, whether it's crypto

52:44

or anything else. And I do think people

52:46

do things in their self-interest, and so they're watching

52:48

self-interest happening. And that's one of the things that

52:50

drives Silicon Valley. Everyone rushes to the internet

52:53

or rushes to crypto or rushes to whatever. But

52:55

Kara, it's not all

52:56

benevolence? No, nothing. None

52:58

of it. Shocking. Self-aggrandizement

53:01

fuels everything in Silicon Valley. So

53:04

that makes sense for me to say

53:06

that. And of course, those underlying trends, the transformer paper,

53:08

the scaling of power is what also those people are watching,

53:10

not just their self-interest, but where are the

53:13

jobs? What can they build? Yeah,

53:15

but I don't even think they read. You don't think they read?

53:17

No, I don't. Where can

53:20

they get money? Oh my gosh. I don't know. I

53:22

think engineers are... So I've met so many

53:24

earnest engineers who are just excited about fixing the next

53:27

problem. Well, they get bored. They get bored and they

53:29

move on to other things. They have very hummingbird-like

53:32

personalities in a lot of ways. They shift to the next

53:34

thing. They do. He outlined this vision

53:37

of AI doctors. It's all

53:39

kind of dystopian. And in a way, it kind

53:41

of makes sense, like this 80% of 80%, because I've heard doctor friends

53:44

say to me, look, I should be able to look

53:46

at numbers for

53:48

most of my patients and then spend more time

53:50

with a smaller set of patients that's needing

53:52

intervention. So

53:55

that makes sense.

53:56

But don't you think that's a bit dystopian or you think that is

53:58

what the world will look like? I don't know if it's

54:00

dystopian. It's happened to farmers.

54:03

It's happened to manufacturing people. It's

54:05

just because it's happening to richer people.

54:07

I think, look, doctors should

54:09

spend more time with patients and not doing forms and

54:12

other things that they're not as good at. I don't think there's

54:14

anything wrong with replacing rote work,

54:16

wherever it may be, with digital

54:19

versions of that that's better. There's not

54:21

any great talent

54:23

in reading a radiology like an x-ray. I

54:27

think there are, of course, jobs that come

54:29

mute with technology.

54:32

That's definitely true. But I think the challenge

54:34

for me was actually, he said, one, humans

54:36

will have other jobs to do. And you've said this

54:39

before with drivers and cars.

54:41

But my big question is, what jobs? I

54:43

don't know. I don't know. I don't know. Nobody

54:45

imagined the internet would create Uber. Nobody

54:48

imagined all kinds of things.

54:50

I don't know something else, but none of us are

54:52

felling trees with axes anymore. We're

54:54

not doing it. Yeah. I think

54:56

the question, the consideration of like, okay, well, what are

54:58

those jobs? And the second problem is, well, he

55:00

says, abundance will allow for redistribution.

55:03

Sure,

55:03

abundance will allow for redistribution of the math,

55:06

but people's politics and mindset

55:08

in this country of, well, we don't really want to

55:10

pay for people who don't work to have health insurance

55:12

or basic income. Those things are hard

55:15

to change. No, they're not. That's

55:17

how life happens. I mean, they didn't have gay marriage 30

55:20

years ago, and now they do. One out of five

55:22

children in the US live in poverty and they haven't found

55:25

a way to fix that. So it's not like just by people

55:27

suffering, you're going to change something. What

55:29

has happened is that a lot more, we've got to come

55:32

up with new jobs. And that's the challenge

55:33

rather than saying, oh, no, oh, no, on everything

55:36

is let's get together with smart people

55:38

working together to figure out what the

55:40

next jobs, how we should reform things. Maybe

55:43

we won't be able to do it, but saying

55:45

it's not going to happen is not really particularly

55:47

helpful. I think neither thing is helpful saying it's not going

55:50

to happen is not helpful, but saying that it's assuredly

55:52

going to happen is also not helpful. It is a very

55:54

big problem. But it is. It is. It's like

55:56

saying cars aren't coming. It's like saying that redistribution

55:59

will happen or.

55:59

humans will have other jobs. Well, yeah, but that's his, it's his

56:02

opinion. It's not easy, no one says

56:04

it's easy. I think it's just, this is, he's just

56:06

stating what I think is probably accurate

56:08

is that this is gonna happen and so this

56:11

is my idea. Yeah, and which isn't dissimilar

56:13

to most of Slaemon's idea, I think the

56:15

question is, is it going to be an idea that

56:17

countries can get behind? In Europe, I think

56:20

there's a lot more inkling and desire

56:22

for that than in the United States, which is just built

56:24

on a more, a different ethos. Well,

56:26

that's an old, that's an old idea

56:28

of the US.

56:29

That has been dying for years and so I

56:32

do think we've changed a lot more than you realize and

56:35

I think there is an openness to

56:37

a lot more creative solutions

56:40

for people. I think people have reached a limit where

56:42

selfishness is not necessarily helping

56:44

us. It's not quite as powerful. or

56:46

not people believe in big government or people believe,

56:49

I think actually government's being gutted more and more and

56:52

people, whether people wanna get behind social

56:54

security programs and other programs and

56:56

there's healthcare, all kinds of things, free education,

56:59

look at the pushback, Biden's getting on loan forgiveness,

57:01

which by the way, has problems. Okay,

57:04

we suck, so I don't know what to say. I either have to

57:06

not suck or not, we will have to do something. I

57:08

think that this is like the hopeful way to look at

57:10

this and the way that I'm trying to look at it is,

57:12

okay, this is going to be the push that we need to

57:14

get to a more

57:17

collective society and a society

57:20

that addresses some of these huge underlying

57:22

problems, chasms that we haven't been able to cross,

57:24

but we were able to cross, for example, in the pandemic,

57:26

there was a coming together and I'm hopeful

57:29

with AI that there will be a coming together to deal

57:31

with some of these things.

57:33

Well, we'll have to deal, again, I think

57:35

there's, most things spin forward,

57:38

not everything, and we go backwards, but kids

57:40

used to work in factories, just like things

57:42

change and we didn't think they were gonna change and

57:44

they do. And so this is an opportunity

57:46

to change in a good way and hopefully, I

57:49

am am heartened by government's movement

57:51

so far, I don't know if it's gonna go anywhere, but to say, oh,

57:54

they'll do nothing again. I hope they

57:56

don't, they have an opportunity.