0:04

Conversations

0:04

with Tyler is produced by

0:06

the Mercatus Center at George Mason University,

0:09

bridging the gap between academic ideas

0:12

and real-world problems. Learn more

0:14

at mercatus.org. For

0:16

a full transcript of every conversation, enhanced

0:19

with helpful links, visit conversationswithtyler.com.

0:26

Hello everyone, and welcome back to Conversations

0:29

with Tyler. Today I'm here with Paul Graham.

0:31

Paul, welcome. Thank you. You've

0:34

written several times that your wife, Jessica

0:36

Livingstone, is a better judge of character

0:38

than you are. What other areas

0:40

of talent judgment is she better or

0:42

much better than you?

0:44

Practically everything to do with people. She's

0:47

a real expert on people and like social

0:49

protocol, what to wear at events,

0:52

what to say to someone, anything like

0:54

that. But you have all this data,

0:56

so why can't you learn from the data and from

0:59

her to do as well as she can? What's

1:01

the binding constraint here in limiting someone

1:03

as a judge of human affairs?

1:05

Partly I just don't have as much

1:07

natural ability, and partly I just don't care

1:10

as much about these things. The

1:12

conversations we have are not, Paul,

1:14

you can't wear that. They're like, Paul,

1:16

you can't wear that! And I'm like, really?

1:19

Why not? What's wrong with this? I

1:21

just don't care as much. Say when you're judging talent,

1:24

Sam Altman, Patrick Hollis, and you're judging people.

1:26

Of course it's the business plan, but much

1:28

of it is the talent. Oh, it's all the people.

1:31

The earlier you're judging startups, the

1:33

more you're just judging the founders. So

1:35

it's location

1:35

for real estate, right? The

1:38

founders are like that. So yeah, sure,

1:40

which is why the earlier

1:43

stage you invest, the more you want to

1:45

have these people who are good judges of people.

1:47

So if she's better than you at

1:49

human affairs, comparing you to other people,

1:52

what exactly, in the smallest number of dimensions,

1:55

are you better at than the other people?

1:57

What am I better at? Yes, what are you better

1:59

at? at this, right? You

2:01

mean to do with startups? To do with

2:03

people, how well they will do with their startups. What's

2:06

the exact nature of your comparative advantage?

2:09

I can tell if people know what

2:11

they're talking about. When they come and

2:13

talk about some idea, especially some technical

2:15

idea, I can tell if they know what, you

2:18

know, if they actually understand the idea

2:20

or if they just have a sort of reading

2:22

the newspaper's level of understanding

2:24

of the thing. And

2:26

how well can you judge determination?

2:29

Determination. It's

2:32

hard to judge by looking at the person. The

2:34

way you would judge determination, because people

2:37

act determined, people think they're supposed to act

2:39

determined, no one walks into their Y Combinator

2:41

interview thinking, I need to seem diffident. Nobody

2:44

thinks that they all think they're supposed to seem tough,

2:46

which is actually really painful and stressful.

2:49

It'd be better if they just were themselves,

2:52

because someone trying to act tough is just, it's

2:54

so painful to watch. But the

2:56

way you tell determination is not so much

2:58

from talking to them as

3:01

from

3:02

asking them stories about things that have happened

3:04

to them. That's where you can see determination. And

3:07

it's how they tell the story or it's what happened? No, no,

3:09

no. It's what they did in the story. What they did in the story. That something

3:11

went wrong and instead of giving up, they persevered.

3:14

Why are there so few great founders in their 20s

3:17

today? Is

3:18

that true? Is that true though?

3:20

I've heard about that. 10, 15 years

3:22

ago, you have a large crop of people who obviously

3:25

have become massive successes. Today

3:27

it's less clear companies seem to be smaller.

3:31

The important people seem to be older. Sam

3:33

Altman was important early on, but now he's very

3:35

important. He's what, 37, 38? Something

3:38

like that. Seems like more of a synthetic set

3:40

of abilities we need.

3:44

It could be because, I'm

3:46

not even sure this is true. It could

3:48

be that it's not actually true and that there's

3:50

something else going on. Like for example,

3:53

people manage.

3:55

If you look at like athletes, for example,

3:58

athletes learned how to stay good. for longer.

4:00

And so you have like people who are really international

4:03

quality soccer players at age 34,

4:06

right? And that didn't used to happen. Back in

4:08

the days of Johann Kreuf smoking cigarettes

4:11

between halves, they didn't last till 34.

4:14

And so it might have seemed like all the good football players

4:16

are 34 and it's just an artifact

4:18

of them lasting longer. But the 20 year old

4:20

star seemed to have vanished. There's not an ex Patrick

4:23

Collison. Patrick Collison probably

4:25

didn't seem to most people to be the big star.

4:27

And you can prove this because if he had,

4:29

they would have invested in early stripe rounds and they

4:32

didn't. So anybody who

4:34

claims they knew early on that Patrick Collison

4:36

was going to be a big star, show

4:39

me the equity.

4:40

Okay. But there was Peter Thiel, there was Sequoia,

4:43

there was yourself. A bunch of people knew in

4:45

fact, or strongly suspected. Sure. Yeah, yeah.

4:48

Some people did, but certainly not everyone.

4:51

So I don't know if this is true. I don't

4:53

know if this is true. I'll investigate this summer. We're

4:55

going to go back to YC and talk to some

4:57

founders and I'll talk to the partners and I'll

5:00

see what's going on. Because I had a mental note

5:02

to check and see if this was actually true.

5:04

There's a Paul Graham worldview in your earlier

5:06

essays where you want to look for the great hackers and

5:09

a lot of the most important companies come from

5:11

intriguing side projects. Oh yeah.

5:13

But if we want older people who are somewhat synthetic

5:16

in their abilities, are those other principles still

5:18

true?

5:19

No, no, not necessarily.

5:22

Not as true. Part of the reason

5:25

you want to find young founders who've done stuff

5:27

from side projects is that it guarantees

5:29

the idea is not bullsh**t. Because if

5:32

young founders sit down and try to think of a startup

5:34

idea, it's more likely to be bullsh**t because they

5:36

don't have any experience of the world. Older

5:39

founders can do things like start supersonic

5:42

aircraft companies or build something

5:45

for geriatric care and

5:48

actually get it right. Younger

5:49

founders are likely to get things wrong if they try

5:51

and do stuff like that. So it's just a heuristic.

5:54

It's a heuristic for finding matches

5:57

between young founders and ideas.

5:59

Why is venture capital such a small part

6:02

of capital markets? So it's big in tech, it's

6:04

somewhat big in biotech, but most other

6:06

areas of the economy.

6:08

You finance with debt or retained earnings or some

6:10

other method. What determines where VC

6:12

works and doesn't work? Growth.

6:14

You've got to have high growth rates because

6:16

VCs, it's so risky

6:19

investing in startups, investing in these

6:21

early stage companies that maybe don't even have any

6:23

revenues. The only thing that

6:26

can counterbalance the fact that half the

6:28

companies completely fail is

6:30

that the ones that don't fail sometimes have astonishing

6:33

returns.

6:34

Can you imagine a future where there's a

6:36

lot more tech in terms of the level but

6:38

slower rates of growth and VC quite

6:40

dwindles because the growth rates aren't there

6:42

or VC goes to some other area?

6:45

VC has been

6:47

dwindling in the sense that they have smaller

6:50

and smaller percentages of companies. If

6:52

you go back and look at the stories from the

6:55

1980s, they would do these rounds where they would get 50% of

6:57

the company. And even when YC

6:59

started, they would get 30 and

7:02

now it's down to 10 in these

7:04

rounds. God knows what they're called, they keep renaming

7:06

the rounds. But for a given round size,

7:09

the amount of equity they get is much smaller than

7:11

they used

7:12

to. It at least seems we have a new dynamic

7:14

Microsoft that ships products quickly

7:17

and innovates and on the surface

7:19

appears to act like a startup again. How can

7:21

that be true? I don't know your famous essay about

7:23

Microsoft culture. How does that happen? I

7:26

did not say that they wouldn't make money.

7:29

But they've done more than make money. They've impressed us

7:31

with their speed. I meant something very specific

7:33

in that essay. Incidentally, I talk

7:36

in that essay about how I was talking to a founder

7:38

and I was talking about how Microsoft

7:41

was this threat and you could see he was clearly puzzled

7:43

like how could Microsoft possibly be threatening. I didn't

7:45

mention who the founder was but it was actually

7:47

Zuck. It was Mark Zuckerberg who was

7:49

puzzled that Microsoft could be a threat. And still

7:52

to this day, if you ask founders, are

7:54

you afraid that Microsoft might

7:56

do what you're doing?

7:58

None of them are. It's still

8:00

not a threat to startups. Yeah, it

8:02

makes more money now, but it's still not a threat

8:05

like it used to be. It doesn't matter

8:07

in the sense of factoring into anyone's

8:09

plans for the future. No startup is thinking,

8:12

I better not do that because Microsoft

8:14

might enter that and destroy me.

8:16

In the early years of Sam Altman, what did you

8:18

see in him other than determination? Because

8:21

he's not a technical guy in the sense of being- He is a technical

8:23

guy. He was a CS major.

8:26

But you're not buying the software he

8:28

programmed. There's something about what Sam Altman

8:31

does. Well, now he'd become a sort of manager,

8:33

but he knows how to program.

8:36

But that's not why his ventures have succeeded,

8:38

right? Oh no, he does. There's some ability to put

8:40

the pieces together. Yeah, yeah. But

8:43

he's not a non-technical guy. He's not just some

8:45

business guy. He's a technical guy who also

8:47

is very formidable and that's a good

8:50

combination. There's

8:52

a now famous Sam tweet where he appears to repudiate

8:54

his earlier advice of finding product

8:57

market fit early and then scaling and

8:59

saying, oh, maybe I was wrong. Saying that what OpenAI

9:01

has done is not exactly that. Do you agree?

9:05

Disagree? Maybe OpenAI is a special case because

9:07

you have to have giant warehouses

9:09

full of GPUs. You can't just mess

9:12

around and throw something out there and see if

9:14

it works. Maybe it requires some amount of

9:16

advanced planning for something on that scale. I

9:19

don't know. I don't know what he meant by

9:21

it or, and I also don't know what it's

9:23

like inside OpenAI really. So

9:26

I don't know either. Do you still take

9:28

a dim view of solo founders?

9:30

It's harder. I know I wouldn't

9:33

start a startup alone because it's

9:35

just so much weight to bear to do

9:37

something like that. But it's very hard to find

9:39

a partner as good as you are. So it's harder and it's

9:41

easier.

9:42

You can just do it. People should do some

9:44

work. People should put,

9:46

when I talk to people who are in

9:48

their teens or early twenties

9:51

about starting a startup, I tell them instead

9:53

of sitting around thinking of startup ideas, you

9:55

should be working

9:56

with other people on projects and then you'll

9:58

get a startup idea out of it. that

10:00

you probably never would have thought of, and you'll get a co-founder

10:03

too. Right? So I wish people

10:05

would do more of that. You can get co-founders,

10:08

just work with people on projects. You just can't

10:10

get co-founders instantly. You've got to have some patience.

10:13

Why is there not more ambition in the

10:15

developed world? Say we wanted to boost

10:17

ambition by 2X. What's the actual

10:19

constraint? What stands in the way?

10:21

Huh, boy, what a fabulous question.

10:24

I wish you'd asked me that an hour ago, so I

10:26

could have had some time to think about it between now and then.

10:28

But you're clearly good at boosting ambition, so you're

10:31

pulling on some lever, right? What is it you do? Oh,

10:34

okay. How do I do it? People are

10:37

for various reasons, for multiple reasons,

10:39

they're afraid to think really big. There

10:41

are multiple reasons. One, it seems

10:44

overreaching. Two, it seems like

10:46

it would be an awful lot of work. And

10:49

so as an outside person,

10:51

I'm like

10:52

an instructor in some fitness class.

10:55

And so I can tell someone who's already working as hard

10:57

as they can, all right, push harder, right? It

11:00

doesn't cost me any effort. And

11:02

surprisingly often, as in the fitness class,

11:05

they are capable of pushing harder. And

11:07

so a

11:09

lot of my secret is just being

11:11

the person who doesn't have to actually do the work

11:13

that I'm suggesting they do. How

11:16

much of what you do is reshuffling their networks?

11:18

So there's people with potential, they're in

11:20

semi-average networks. We should talk

11:22

about that some more though, because that really

11:25

is an interesting question. Imagine how

11:27

amazing it would be if all

11:29

the ambitious people could be more ambitious. That

11:31

really is an interesting question. And there's

11:33

got to be more to it than just the fact that I don't

11:35

have to do the work. I think a lot of

11:38

it's reshuffling networks. So you need someone

11:40

who can identify who should be in a better network.

11:42

You boost the total size of all networking that

11:45

goes on

11:45

and you make sure those people with potential. By

11:47

reshuffling networks, you mean introducing people

11:50

to one another. Of course. You pull them away from their

11:52

old peers who are not good enough for

11:54

them and you bring them into new circles,

11:56

which will raise their sights. Maybe,

11:59

maybe that is.

11:59

true when you read autobiographies, there's

12:02

often an effect when people go to some elite

12:05

university after growing up in the middle of the

12:07

countryside somewhere, they're suddenly become

12:09

more excited because there's a critical mass

12:11

of like-minded people around. But

12:13

I don't think that's the main thing. I mean, how does

12:16

it make sense? Think about the power of London in the 17th

12:18

century. So the Industrial Revolution happens

12:20

further north, but the idea is the science, inner

12:23

near London, maybe Cambridge,

12:24

and Oxford. And those are the networks

12:27

people are brought- Near London only in the sense that

12:29

everything in England is near everything else by American

12:31

standards. It's not really

12:34

that London was the center of ideas. There

12:36

were a lot of smart people there, but things were more spread

12:38

out. Be careful with English history there.

12:41

Back to this idea though, of how

12:43

to get people to be more ambitious. It's

12:45

not just introducing them to other ambitious people.

12:49

There is a sort of skill to

12:51

blowing up ideas, blowing up not

12:53

in the sense of destroying, making them bigger.

12:55

There is a skill to it. There is a skill

12:58

to it to take an idea and say, okay,

13:00

so here's an idea. So how

13:03

could this be bigger? There is somewhat of a skill

13:05

to it. So it's helping people see their ideas

13:07

are bigger than they thought. Yes. Oh yes. We

13:09

often do this in YC interviews. And people

13:11

say you're especially good at that. This is what the other

13:13

people say. That's why I'm mulling over what actually

13:15

goes on because there is this skill there. The

13:18

weird thing about YC

13:19

interviews is in a sense, they're a negotiation,

13:21

right? And in a negotiation, you're always saying, oh,

13:24

I'm not going to pay a lot for that. It's terrible. It's

13:26

worthless, right? And yet in

13:28

YC interviews, the founders often

13:31

walk out thinking, wow, our ideas

13:33

are a lot better than we thought just because

13:35

what we do, you know what we do in YC interviews?

13:38

We basically start YC. The first 10

13:40

minutes of YC is the interview. So

13:42

you see what it's like to work with people

13:44

by working with them for 10 minutes. And that's

13:47

enough. It turns out. So you

13:49

think the 11th minute of an interview has

13:51

very low value. I've thought a lot about

13:53

where the cutoff is. Like where's

13:56

the point? If you made a graph, what's your probability

13:58

of changing your mind, right?

13:59

After minute number n. Yeah.

14:02

And after minute number one or two, the

14:04

probability of changing your mind is pretty high.

14:07

I would say YC interviews could

14:09

actually be seven minutes instead of 10 minutes, but 10

14:12

minutes is already almost insultingly

14:14

short. And so we kept it at 10.

14:16

We could have made it seven. I think there's often a

14:19

threshold of two and then another threshold

14:21

at about seven. Yeah. And after that, it's very

14:23

tough for it to flip. Right. Yeah.

14:25

Although that doesn't mean you're always right.

14:28

But it could just be after three hours. It's

14:29

just not going to flip. Yeah. Yeah.

14:32

You're not going to change. I didn't say seven minutes

14:34

is enough to tell, notice. I

14:36

said seven minutes is the point where

14:38

you're probably not going to change your mind. If

14:41

it's going somewhat badly and the person is

14:43

flipping positive at minute six, what is

14:45

it that's happening both in the interview and

14:47

in you?

14:48

If it changes from two to seven.

14:50

Clearly from zero to one or two, they get over

14:52

nerves or they adjust the sound volume. There's

14:55

plenty of those stories. It's probably when they misunderstood

14:57

what they're working on initially.

14:59

So great idea about presenting it? No,

15:01

more like they're near some idea

15:04

that we're familiar with. And we just assume they must

15:06

be doing that idea. And they say, oh, no,

15:08

we're not doing that. We're doing this. And like, oh,

15:11

okay, thank goodness. And then they get extra points

15:13

because not only they're not doing the stupid thing, but

15:16

they understand that the stupid thing is stupid.

15:19

So now it's, they get extra

15:21

credit for what we were subtracting in

15:23

the past. Who falls through the cracks

15:25

in the YC process as you've experienced

15:27

it?

15:28

YC, one of the reasons I'm

15:30

so contemptuous of university admissions

15:32

is that I am also in the admissions business and

15:35

I

15:36

am obsessed. I not only

15:38

measure, we not only measure when

15:40

we fail, we were obsessed

15:42

with the failure cases. YC has a list

15:45

of all the companies that

15:47

we've missed that we've, that have applied to

15:50

YC and we've turned down and they've gone on to be successful.

15:53

And we spend a lot of time like,

15:55

like someone thinking about past injuries. What's

15:58

the common element though, now that you.

15:59

When there are common elements,

16:02

my God, do we act quickly to

16:04

fix that. And so I remember

16:07

earlier, there was this company

16:09

doing email. They

16:11

would, if you wanted to send mass emails, I forget what

16:13

their name was. And this was

16:15

back when we were reading the applications. If

16:18

the first reader gave it a sufficiently low grade,

16:20

it would never be seen by anybody else.

16:23

So the first reader who was Robert Morris, who's

16:25

extremely, he

16:26

was like designated wet blanket of YC,

16:29

gave this application a C and

16:32

with the comment spam company,

16:34

that was nobody ever saw it again. And

16:37

they like ended up going public. And

16:39

so after that, I changed

16:41

that thing and the software made it

16:43

so that no one would ever see something, no second

16:46

reader would see it. After that, every application

16:48

had to be seen by at least two people. How

16:51

did you get over your fear of flying?

16:53

Oh, really? Did you, did

16:55

you, did I already talk about this? There

16:57

were no, there were either two or three places

17:00

in your writing where you mentioned that you had a fear of flying,

17:02

but you use the past tense, which implies you

17:04

got over it, but you never told anyone how. Oh,

17:07

it's a bizarre strategy. I learned

17:09

how to hang glide, which sounds

17:12

crazy. That will do it though. Right. If you're afraid

17:14

of flying, how could you learn how to hang glide? But the answer

17:16

is you learn how to hang glide gradually.

17:19

You start by just running along the flat.

17:21

And if there's a headwind, maybe you feel

17:23

a little lift, right? And then you go 10 feet

17:26

up the hill and run as fast as you can.

17:27

And you reach a total altitude

17:29

above ground level of a foot, right? You're

17:32

not afraid when you're a foot above the ground. So

17:34

you go out a little further up the hill until

17:36

a month later, you're like jumping off

17:38

a cliff with a hang glider on your back.

17:41

And so after I was good at hang gliding, I

17:43

took flying lessons.

17:45

So there's this intermediate point where

17:48

I was comfortable,

17:50

totally comfortable jumping off a cliff with a hang

17:52

glider on my back, moderately comfortable

17:56

flying a Cessna 172, where

17:58

the instructor had just. turned off the

18:01

engine and said, okay, land it. Cause

18:03

the glide ratio is actually similar to a hand glider

18:06

and still afraid of getting on an airliner that shows

18:08

you how irrational these things are. But

18:10

when I finally did get on an airliner, my God,

18:13

it was like a spaceship compared

18:15

to the planes I'd been flying. It was fabulous.

18:17

And it totally worked. My fear of flying was

18:19

completely cured. And what did you learn about

18:22

startups and talent selection from that process?

18:25

Boy, if I learned anything, I haven't considered

18:27

it until this moment. But you might've learned

18:29

it quite well. Just not articulated it. I don't

18:32

know. You asked me these hard questions. I don't have

18:34

time to think about them. Did that have anything

18:36

to do with, I don't think I've ever consciously

18:39

used anything I learned there. Do you see founders

18:41

who go through a comparable process with something

18:43

other than flying? Oh, I get where you're at

18:45

and here's how you, what you need to do next. A

18:48

lot of technical type founders

18:51

hate the idea of doing sales and

18:53

going and talking to people. And so you can tell

18:55

them, look, just go do this. And you know,

18:57

after a few months, they'll be used to it. They

19:00

may never like it. I mean, I would hate doing that myself

19:02

now, but they can at least do it. Were the Medici

19:05

good venture capitalists or do you give greater

19:07

credit to the Florentine guilds? I

19:09

have no idea. I have no idea.

19:11

I need to learn more about the... The guilds

19:14

would run competitions. The Medici would just

19:16

pick the people they liked. They both have

19:18

good records in different ways, but obviously they're

19:20

competing models. You're an economist. You've

19:22

read books about this stuff. I don't know. I

19:24

don't know.

19:25

What do I know about the Medici? I think the Medici

19:27

were overrated and the guilds were more important, but that's

19:29

a debatable view. You could argue either way. They could well

19:31

be overrated. They were the ones who had all the publicity.

19:35

So I have no idea. Your time

19:37

spent at RISD and in Florence. How did

19:39

it alter or affect your thoughts on software

19:41

design and talent selection?

19:45

Man, if you asked how it affected my ideas

19:48

about painting, I could give you an answer. Florence

19:50

was all about talent selection because how many

19:52

people in quite a small area

19:55

became both great and famous.

19:58

Yeah. One thing I remember.

19:59

There was one moment I was sitting in Florence

20:03

and I realized I had gone to the wrong place.

20:06

I was studying Florentine history. All

20:08

the buildings are right around you. I

20:10

found myself thinking,

20:12

damn, Florence was New

20:14

York City in 1450. That

20:17

was 60,000 people, yeah. That's why it was

20:19

good at art. It wasn't some weird

20:22

Florentine thing about art or

20:24

some special Florentine sense of aesthetics.

20:27

They were just the most progressive

20:30

city.

20:31

It's really

20:33

too bad that the far left has hijacked

20:36

that word. It's such a good word. Maybe we

20:38

can get it back. But

20:41

I remember thinking, I've gone to the wrong place.

20:44

I went to the place where the puck used to be

20:47

of hundreds of years ago.

20:50

But you can learn a lot from studying where the

20:52

puck used to be.

20:54

Oh, not really.

20:56

Are you sure? Famia was a pretty

20:58

crappy art school. But

21:00

being in Florence itself. Oh, I

21:02

can learn a lot from looking at the works. Yes, but

21:04

I don't have to go and live there. But

21:06

what is it you learned that was relevant for Y Combinator?

21:09

Because you're doing something comparable to what the

21:11

Florentines did. Picking a pretty small area,

21:14

making it the absolute center for talent selection,

21:17

a magnet that drew people in, and having

21:19

a lot of winners. So you copied what you

21:21

were living. I had already learned that

21:23

from Harvard.

21:24

Maybe you had to learn it twice. No,

21:26

I was already very well aware of this

21:28

phenomenon. So to

21:30

the extent YC uses anything

21:33

like that, we were definitely thinking YC was

21:35

started within

21:37

the convex hull of Harvard. Like

21:39

the places Harvard spread out through

21:42

Cambridge, but YC's original office

21:44

was within it. And so we've

21:46

consciously tried to make YC the

21:49

Harvard of startups. No question about that.

21:51

We had the model right there. Harvard's

21:54

very screwed up, as you know. You look at their admissions.

21:56

How much is like Dean's favorites and legacies.

21:59

It's not, it's not

22:02

Y Combinator, right? They're trying to build a coalition

22:06

and you're just caring about picking winners. The

22:08

thing is though, all universities

22:10

have a sort of admissions

22:12

process that's corrupt in that way,

22:14

except possibly Caltech. Caltech

22:17

might actually do undergrad

22:19

admissions properly. And so since they're

22:21

all messed up, Harvard

22:23

does still have this draw compared

22:25

to the others. Does AI

22:28

make programming even more like painting?

22:31

God, what a question. How do you make

22:33

these questions? Does AI

22:35

make programming more like painting?

22:40

I have no reason to believe that. It

22:43

might be true. It might be true. You're

22:45

piecing things together more arguably

22:48

in this new world,

22:49

post GPT world. What I was saying

22:51

when I said programming was like painting is they're

22:53

both building something, right? And

22:56

you're not building something anymore with

22:58

AI than when you were writing code by hand.

23:01

So I would guess not. I would guess

23:03

not.

23:04

How far is mid-journey from

23:06

quote unquote real art? I

23:08

don't know. Is it more decrepit modernism? Is it a fantastic

23:11

revolution or as art, put

23:13

aside the startup angle? How do you view it? I

23:15

don't, I see all these AI generated

23:19

images. I don't know which ones are from mid-journey

23:21

and which ones aren't. So I can't say

23:23

for sure about mid-journey, but I have definitely

23:26

seen some AI generated stuff

23:28

that looks amazing, that looks truly

23:30

impressive. Not just that it looks impressive.

23:32

So it's amazing as graphic design, but

23:34

do you think it's art in the same way that

23:37

Rembrandt is art? This

23:38

whole thing about like what's art

23:40

and what isn't, I think it's all a matter

23:42

of degree. Like my crap,

23:45

carnation coffee mug is

23:47

art. It's just not very good art. So

23:50

there's not like some threshold where above this

23:52

threshold it's art. Everything people make

23:54

is art, just varying degrees of goodness.

23:57

But I can tell you some of the things I've seen

23:59

the worry.

23:59

I generated, I'd be impressed if

24:02

a person made them. So

24:04

that probably is over your threshold.

24:06

If you're good at talent selection, who is

24:08

an underrated painter and why?

24:11

Ah, wow. Boy,

24:14

there is a topic I think about

24:16

a lot. So

24:17

there's a bunch of different reasons people can be underrated.

24:20

Almost all good artists are underrated.

24:22

I agree. It sounds weird, but if you

24:24

look at where the money's spent at auction,

24:27

it's almost all

24:29

fashionable, contemporary crap. Because

24:32

if you think about how prices

24:34

in very high end art

24:36

are set,

24:38

they're, they're auction prices. How many people

24:40

does it take to generate an auction price?

24:42

Two, just two. Right.

24:45

And so you have boneheaded

24:47

Russians who want to have a

24:50

Picasso on their wall. So people will think they're

24:52

legit or hedge fund

24:54

managers, wives, who've been told

24:56

to buy impressive art to hang

24:59

in their loft. So when

25:01

people come over, they'll say, Oh, look, they've

25:03

got a Damien Hurst. The

25:05

way art prices at the very high

25:07

end are set is almost entirely

25:10

by deeply

25:11

bogus people, which

25:14

is great. Actually,

25:16

when I was an artist, I used to be annoyed by this. Now

25:19

that I buy a lot of art at auction, I'm delighted

25:21

because it means there's all

25:24

this money. Andy Warhol's screen

25:26

prints selling for $90 million. Right.

25:28

Yes. And old masters can be, I wouldn't

25:30

say cheap, but I would say radically underpriced.

25:33

Couple hundred thousand. Or even less

25:35

for some good ones. Yeah. And so

25:38

I know cause I buy them, but

25:41

I used to be annoyed by this. And now I

25:43

think it's the most delightful thing in the world. Cause

25:45

there's all this loose money sloshing around

25:48

and so-called contemporary

25:51

art is like the sponge that just absorbs

25:53

all of it. There's none left. Some

25:55

of the things I buy, I'm the

25:58

only bidder. I get it for the reserve.

25:59

of price, no one else in the world wants it,

26:02

or even knows that it's being sold. So

26:04

I am delighted about this.

26:06

So the answer to your question, which artists

26:09

are undervalued? Essentially

26:11

all good artists. The intersection

26:14

between the very, very,

26:16

very famous artists, like artists famous

26:18

enough for Saudis to have heard of them. Like

26:21

Leonardo, I would say, is

26:24

probably not undervalued, except for

26:26

the artists who are

26:27

like household names, every elementary school

26:30

student knows their names. They're all undervalued.

26:32

If you think that something has gone

26:34

wrong in the history of art, and you try to explain

26:36

that in as few dimensions as possible, what's

26:39

your account of what went wrong? Oh, oh,

26:41

I can explain this very briefly.

26:44

Brand and craft

26:47

became divorced. It

26:49

used to be that the best artists were

26:52

the best crafts people. And once

26:55

art started to be reproduced

26:58

in newspapers and magazines and

27:00

things like that, you could create

27:03

a brand that wasn't based on quality.

27:06

So you think it's mass media causing the

27:08

divorce between brand and craft. It certainly

27:10

helps. And then talents responding accordingly. Fundamentally

27:13

what went wrong? You invent some shtick,

27:16

and then technically

27:18

it's called a signature style. So

27:20

you paint with this special shtick. If

27:22

someone can get some sort of ball

27:24

rolling, some speculative ball rolling,

27:27

which dealers specialize in, then

27:30

someone buys the painting with your shtick and

27:32

hangs it on the wall in their loft

27:35

in Tribeca. And people come in and say, oh my

27:37

goodness, that's a so-and-so, right?

27:40

Which they recognize because they've seen the shtick.

27:42

But say if we have modernism raging in the 1920s, in

27:45

the 20s mass

27:46

media's radio for the most part.

27:48

No, newspapers were huge. Modernism was well-

27:50

Not for showing paintings. There was no color in the papers.

27:53

You have to be- The 1920s were

27:55

good enough to make painting

27:58

like Cezanne fashionable.

28:00

So it was just about getting going in

28:02

the 1920s. Why

28:05

can't we build good British country homes anymore?

28:08

Or do you think we can? There's nothing stopping

28:11

you. But it doesn't happen. I mean, except

28:13

they're planning people. You can't. Do you know anything

28:15

about building houses in England? You just cannot build.

28:17

It's impossibly difficult. It's one of the worst countries.

28:20

It's the reason it looks so nice here. Yes.

28:23

If they had America's zoning rules here, the

28:25

entire countryside would be plastered with houses.

28:28

There's not any fields left at all because

28:30

the difference in value between agricultural

28:32

and building land, it's like, God,

28:35

100X or something like that. And this

28:37

place is small. Yeah, but there are new buildings,

28:39

say, in Cambridge, many other parts of Britain,

28:42

and the new buildings are not country homes. Yet

28:44

everyone on average is wealthier.

28:47

And why the change? Welder than where?

28:49

Oh, in the old days. Welder than

28:50

Britain in the old days. So even before the

28:52

railway, you have large numbers of country, most

28:54

country homes being built when living standards

28:57

were pathetically low.

28:58

A country house has got to have a certain

29:01

amount of land around it. You can't just have

29:03

them plastered down a street. It would look wrong.

29:06

You want a different style of architecture for something like that.

29:08

And there is a place where they build big houses.

29:11

You know what it is?

29:12

It's like when the Macintosh appeared and you

29:15

could have whatever font you wanted, right? Most

29:17

people have bad taste. In the

29:19

old days, you had to have a classical

29:21

looking house because that was the only way to build houses.

29:24

In the Victorian period, actually, was when things went

29:26

wrong. You could have an Italian villa or wait,

29:28

no, it could be a Greek temple or something

29:30

that looked like it was from the Tudor period. Take

29:33

your pick, mix them together. Greek temple

29:35

with Tudor bits. And it was all over from

29:37

that point on, really. In which

29:40

way, if any? It's not like they build good houses in America

29:42

either.

29:42

No, and neighborhoods are worse. Yeah, there

29:45

are nice individual homes.

29:46

But is there any truly beautiful neighborhood

29:48

built after 1950?

29:51

Where would it be? In any country?

29:53

No, in Palo Alto, there's a

29:56

neighborhood of houses built by this guy called

29:58

Eichler, this developer

29:59

hired some of the best mid-century

30:03

architects. And arguably those neighborhoods

30:05

are good. Although they messed up the trees.

30:08

When were those built?

30:10

50s and 60s. Okay. But that's 70 years

30:12

ago now. You haven't done anything

30:14

in 70 years. I'm sure

30:17

someone is building some good houses somewhere.

30:19

The point is you don't really need to. And

30:21

so it only happens by accident.

30:23

Right? Developers mostly

30:26

are thinking we need to just turn this

30:28

land into houses as soon as possible. The buyer buyers

30:31

don't have any taste. We don't need to sweat that. So

30:33

we'll just build random houses that look

30:35

big and have large master bedrooms. People

30:37

will buy them. We'll get our

30:39

capital back and go on and do the next one. They

30:41

don't need to be good. Houses don't need to be good.

30:44

They didn't need to be beautiful in the old days

30:46

either. It was just technology was so

30:48

constrained. You didn't have any choice.

30:50

You get to go back in time. Your health is guaranteed

30:53

and you know, whichever languages you might need. And

30:55

you spend six months somewhere safe.

30:58

Your safety is guaranteed. Where do you choose?

31:01

I often think about this question. You

31:03

seem like someone who often thinks about this question.

31:05

Yeah. Sadly. Well, the way

31:08

I think about it is I keep trying to escape from the obvious

31:10

answer and I don't manage to because the obvious

31:12

answer that Athens, which is that's

31:14

where everyone would pick. It's not what I would pick,

31:16

but what's your choice number two then? I'll tell

31:19

you mine. I think that there are things where.

31:23

I am obsessed with the mystery of what

31:25

the hell was going on in Dark Age Europe. I'm

31:27

deliberately using the term Dark Age because they're

31:29

trying to outlaw it. But if

31:32

the term Dark Ages hadn't been invented

31:34

already, that would be a fabulous invention

31:37

to describe that period. Yeah, sure.

31:39

New things were getting invented. Some people were

31:41

doing good things just like always happens, but

31:44

it was as bad as things have been and

31:46

there's so few records. Nobody knows how

31:49

things happened. Like I would be really

31:51

interested to know when

31:53

barbarians infilled the Roman Empire,

31:55

but there were still big Roman

31:57

landholders and they had to give a large percentage

31:59

of.

31:59

their estates to these new barbarians who

32:02

were in a sense running things and yet they

32:05

were in the cockpit, but they didn't know what the buttons

32:07

did. So you want to go to Northumbria

32:10

in Northern France? No, no, no, no. Or do you want

32:12

to be Dark Ages?

32:13

You've picked the arrow, now give us the spot. Like Provence.

32:16

Provence. Provence in 600. What

32:22

was going on? What was going on in

32:24

Provence in 600? I would be very

32:26

interested to see that. But almost

32:29

like morbid curiosity. I wouldn't

32:31

learn as much as I would from going

32:33

to Athens, but I just really

32:36

want to know what was going on. I'm a big fan

32:38

of morbid curiosity, by the way. It's the

32:40

best kind of curiosity in many cases. I can

32:42

tell. I would pick the Aztec Empire

32:44

before the Spanish arrived.

32:46

So I feel I have vague glimpses of what ancient

32:48

Athens was like. There are still Greek ruins, but

32:52

the Europeans marveled at the cities they

32:54

saw and they burnt all the books. And

32:57

I think I would learn more by somewhere more strange.

32:59

Maybe. The reason

33:01

I'm interested in medieval Europe is

33:03

it's because where our world came from, right?

33:07

The clocks we used, the writing system, all

33:09

the clothes eventually evolved from that. So

33:11

the reason I'm interested, I

33:13

long ago realized that the

33:16

medieval period wasn't a dip. It wasn't

33:18

like there was this high level of

33:20

civilization and then it dropped down for a while and

33:22

then it rose back up again. It was more like there

33:24

was one civilization that was high and went down

33:27

and another civilization based in the north

33:29

that rose up. And so

33:31

in a sense, it was the beginning of everything,

33:33

right? That's

33:35

why I want to know what the hell was going on in 600 or 700. Did

33:38

Rome have to fall or can you imagine a path

33:41

where Rome has an industrial revolution and

33:43

we save ourselves 700 or however

33:45

many centuries of time?

33:47

Like you need the fragmentation to get the competition

33:50

for

33:51

Britain to become significant

33:53

and maybe, maybe, I don't know. Ah, 2000

33:57

years is a long time or even 1700

33:59

years.

33:59

years, right? Because China never falls. Actually,

34:02

Rome was half the decent until 200, so 1500 years. But 1500

34:06

years, things could have changed a lot. But

34:08

did they in China? Chinese empire never collapses.

34:11

It takes many forms, many dynasties, but

34:13

it ends up stalling.

34:15

So maybe the collapse of the Roman empire was

34:17

one of the best things that could have happened for Europe

34:19

at least. Oh, people. I'm sure people didn't

34:21

think so at the time. I'm sure people were thinking at the time,

34:23

it didn't have to be this bad. I don't

34:26

know. Neither of us knows about this kind of

34:28

thing.

34:29

Looking forward, how optimistic are you about

34:31

the future of the UK? No real wage growth

34:33

since 2008, no real productivity growth

34:36

for as many years. What's

34:38

up and what's the path out of that?

34:40

I am optimistic because

34:43

they still have a gear that they

34:45

haven't shifted into. I suppose

34:47

I'll really have gotten native when I say we instead

34:49

of they. Because you grew up in Pittsburgh, right?

34:52

Yeah. Here Pittsburgh. Yeah. But

34:55

I'm British by birth. Does that count? Really

34:57

British or like your parents were diplomats here? No,

35:01

yes. So you're really British? Yes. Okay.

35:04

Yes. Just ask HMRC as far as they're concerned, I never

35:06

left. HMRC is the

35:09

British IRS. Okay. The

35:11

reason I'm optimistic about Britain, I was just

35:13

thinking about this morning, is because

35:16

people here are not slack. They're

35:18

not lazy and they're not stupid. And

35:20

that's the most important thing.

35:22

So eventually non-lazy,

35:25

non-stupid people will prevail. And

35:27

I've funded a couple startups here. You

35:30

can see like when you introduce these

35:32

people to the idea of trying

35:34

to make something grow really fast and have these

35:36

really big ambitions,

35:38

it's like teaching them a foreign language.

35:41

But they do learn it. They do

35:43

learn it. It's not like English people

35:45

are somehow genetically inferior to

35:47

Americans. So I think

35:50

they have all that potential still to go. I'm astonished

35:52

when I see statistics like I think GDP

35:55

per capita in the UK is only two thirds

35:58

of what it is in America.

35:59

as Mississippi. It's preposterous,

36:02

I know. So imagine the potential there.

36:04

Imagine the potential. But you don't

36:06

feel that way about Mississippi necessarily.

36:09

No, I think Mississippi

36:11

is probably already up close

36:13

to its full potential. I don't know. Never

36:15

actually been there. I shouldn't say things like that. But

36:18

why and how did so many things here end up

36:20

under capitalized? So the water utilities, the

36:22

NHS, it seems to be a consistent

36:24

pattern. That could make us more pessimistic

36:26

because the flow numbers don't reflect

36:29

the fact that capital maintenance is even worse than

36:31

we had thought. I don't know. You're an economist.

36:33

I don't know what the term capital maintenance means. But you're

36:35

a British person,

36:36

I'm told. Under capitalized? I don't

36:38

know what these things mean. I

36:41

mean, the NHS is run by the government. So

36:43

things run by governments are often bad. Although

36:45

the NHS seems to be pretty good, even though people

36:48

attack it. It's a lot more civilized than the American

36:50

system. But how long it takes an ambulance

36:52

to arrive if you fall down? Well, that only recently got bad.

36:54

That's just in the last couple of years. That's what under capitalization

36:56

means. You keep on borrowing against the future. You

36:59

don't plow resources back in. So

37:01

eventually, you don't have any more. When things get bad

37:03

enough, they fix things. This place is not

37:05

run by the kind

37:06

of yahoos that America is. It may

37:08

be a small country, but people running

37:10

things, you know, they're not just like

37:12

boneheaded political appointees. Like

37:16

when things are wrong, they notice they're wrong and

37:18

they fix them. This is a very old country.

37:21

That's another reason. It's not going to tank.

37:23

They've been through some bad stuff before.

37:25

There have been ups and downs. Are

37:28

you an optimist about the city of San Francisco?

37:30

Not the area, the city. Tell us why.

37:33

I can't tell you because there are

37:35

all sorts of things happening behind the scenes

37:37

to fix the problem.

37:39

In politics, you mean, or in tech

37:41

startups?

37:42

No, no, no. Politics, politics. The problems

37:44

with San Francisco are entirely due to

37:46

a small number of terrible politicians. It's

37:48

all because Ed Lee died. The mayor, Ed

37:51

Lee, was a reasonable person.

37:53

And up till the point where Ed Lee died, San Francisco

37:56

seemed like a utopia. But it was like

37:58

when Gates left Microsoft.

37:59

things rapidly reverted

38:02

to the mean. Although in San Francisco's

38:04

case, way below the mean. And

38:06

so, didn't take that much to ruin

38:08

San Francisco. It's really, if you could,

38:11

if you just replaced about five supervisors,

38:14

San Francisco would be instantly a fabulously

38:17

better city. But isn't it the voters you need

38:19

to replace? Those people got elected, reelected?

38:22

The reason San Francisco fundamentally is so

38:24

broken is that the supervisors

38:26

have so much power and supervisor

38:29

elections

38:29

you can win by a couple hundred votes. So

38:32

all you need to do is have this hardcore

38:35

of crazy left-wing supporters

38:37

who will absolutely support you no matter what

38:39

and turn out to vote. And everybody else says,

38:42

oh, local election doesn't matter. I'm not going to bother.

38:44

It's a uniquely weird situation that wasn't

38:46

really visible. It was always there, but it

38:48

wasn't visible until Ed Lee died. And

38:51

so now we've reverted to what that situation

38:53

produces, which is a disaster. Now

38:56

we're in 2023, say two or three years

38:58

from now, what do you think the regulation

38:59

of AI will look like? Oh,

39:02

God knows. Oh, you keep asking me questions

39:04

I have no idea about. You have plenty of idea.

39:06

You know as much as anyone, I suspect. No, that's not

39:09

true. That is not true. I

39:11

really have no idea what AI regulation

39:13

will look like or even should look like, which

39:16

is an easier question. Here's

39:18

an easier question yet. So there's a broadly

39:20

a tech community in the Bay Area. Like you said, I

39:22

had to make up the regulations for AI

39:24

this afternoon. It would be really

39:26

hard. I agree. And so that's

39:28

how far away I am from being

39:29

able to answer that question on the spot. I

39:32

couldn't even figure it out in a day. Like when it should

39:34

be modular or when there should be a regulation

39:37

on AI as a thing that itself is

39:39

intractable. I'll tell you one meta fact

39:41

though. There was a guy on Twitter. I think his

39:43

name was Rob Miles, who said

39:45

that trying to make safe

39:48

AI will be like trying to make a secure

39:50

operating system. And that is absolutely

39:53

true. And therefore frightening

39:55

because the way you make a secure operating system

39:58

is not by sitting down and thinking.

39:59

at a table with a piece of paper

40:02

about the principles for making a secure

40:04

operating system, more like you

40:06

try and make such principles and then someone

40:08

hacks your operating system, and then you think,

40:11

oh, okay, sorry, and you patch it,

40:13

and then they hack that. Making

40:15

a secure operating system is like making a

40:17

fraud-proof tax code. It's basically

40:20

a series of patches that were based

40:22

on hacks, successful hacks, right?

40:24

Is it the Mellon? They say the US tax

40:27

code is basically a series of responses to

40:29

things Mellon did. Some

40:29

parts of it, yes. Yeah, so secure

40:32

operating systems are like that. And so I'm

40:35

worried about AI because

40:38

whatever the regulations are, they'll be wrong.

40:40

I'll tell you that. They'll

40:42

be over-regulated in some ways, and

40:44

miss and have huge holes in others. There's

40:47

a Bay Area tech community, at least in the not

40:49

too distant past. They agreed about many things,

40:52

but very recently it seems on AI, there's quite

40:54

a divergence of views. People who are very worried,

40:56

oh, it's going to kill us all in the world. People

40:59

who say, oh, there's problems, but this is going to be great.

41:02

In as few dimensions as possible, what accounts

41:04

for that difference in perspective from

41:06

people with broadly similar backgrounds and

41:08

who used to agree on many things?

41:10

Is it temperament? Is it genes? Is

41:12

it like a snake bit? It's probably which

41:15

aspect of the problem they choose to focus on,

41:18

right? You could focus on either one.

41:21

And so if you focus on how

41:23

it could be good, there's all sorts of exciting

41:25

things to discover. And you discover lots of

41:27

genuine ways it could be good. And if you

41:29

focus on how it could be bad, it's

41:32

true there too. So it could be either.

41:35

I managed to keep both thoughts in my

41:37

head simultaneously. I simultaneously

41:39

think there are all kinds of good things and all kinds

41:41

of bad things and that both will be, they

41:44

will be unimaginably good and unimaginably

41:46

bad. So

41:48

it sounds like that

41:50

produces oscillation, doesn't it? Yes. That's

41:52

worrying. That in itself is worrying, right?

41:55

Why hasn't Lisp been more successful? Or

41:58

do you think it has? Closure.

41:59

is a dialect of Lisp and Clojure

42:02

is very successful. So it's

42:04

been successful in that respect. And there's another

42:06

way it's been successful. Some languages

42:08

that are not considered dialects of Lisp like JavaScript,

42:11

if you showed JavaScript to people in 1970, they

42:15

would say, this is a Lisp, except

42:17

for the syntax, this is a Lisp. So

42:19

Clojure, it's literally

42:22

successful through Clojure. It's de

42:24

facto successful through JavaScript.

42:27

But why doesn't everybody use Clojure

42:32

or some other dialect of Lisp? Because the notation

42:34

is frightening.

42:35

There's an initial hump

42:37

with the notation. And if you

42:39

give people an initial hump, like

42:42

if you put, you must know about

42:44

this, there must be names for this in economics.

42:46

If you put some sort of obstacle, right?

42:49

Like a container in front of people's

42:51

front door, they'll go off

42:53

to the left and then they won't go right back

42:55

in front of the container and resume their original path.

42:58

No, they'll take this other path that goes like miles out

43:01

of their way just because of that one block in front

43:03

of their front door. And so the syntax, the

43:05

reverse Polish notation

43:07

puts people off. So is AI generated programming

43:10

going to vindicate you on Lisp over time

43:12

or cousins of Lisp? No.

43:15

Because the AI doesn't care about the notation, right? It

43:17

does. It does. Because it's trained

43:19

based on the amount of code that's out there. But you could

43:22

train it on something more like Lisp. You have to train on actual

43:24

examples people have written.

43:26

You could write some code in kind

43:28

of perpetual motion machine, train other AIs

43:30

on that code and converge to something better,

43:32

better, better. No? There was some research

43:35

paper recently where they trained an AI on

43:37

the output of AIs and it converges on crap.

43:40

Maybe there'll be some solution because

43:42

it's a very rapidly evolving field. But I

43:45

think you have to have a large corpus

43:47

initially of

43:48

examples written in the language. Other

43:50

than hackers and hacking, what other

43:53

human activities are what you have called high

43:55

cost interruption? That is, if you're interrupted,

43:58

you lose your train of thought, have to start all over again.

43:59

again? All math, I think

44:02

must be like that. I think anything. Painting,

44:04

yes or no? Not

44:06

as much in my experience, not as

44:09

much. You don't have to think ahead

44:11

as much in painting. It's annoying

44:13

to be interrupted no matter what, but it doesn't absolutely

44:16

destroy you like it does in the middle

44:18

of writing a program or something. You

44:20

don't build a big mental model of something in

44:22

your head when you're painting. That's what it is.

44:24

It's when you've got this giant house of cards in your

44:27

head. That's when you get destroyed

44:29

by an interruption. Do you think kids

44:31

today spend too much or too little time kids today,

44:33

learning high cost interruption activities?

44:37

Do they spend too much or too little time?

44:39

Like are we under investing

44:41

in high cost interruption activities? If you

44:43

think about what it's like in schools, kids are constantly

44:46

being interrupted and they always have

44:48

been. So whatever the answer

44:50

is, it's not gonna be kids these days. It will be

44:52

a statement about school for centuries.

44:56

Kids have such short attention spans. It's

44:58

one thing they can't do as things that are big and long. I think

45:01

if you said, okay, you have five

45:03

hours to sit in a quiet room and build

45:05

something, I don't think they'd be up to it anyway. Our

45:08

last segment is what I call the Paul Graham production

45:11

function,

45:11

which is how you got to be Paul

45:13

Graham. Would you major in philosophy again

45:16

at what, Cornell? If you were doing it all over.

45:18

No, I would not major in philosophy. And why not?

45:21

Didn't it allow you to think at a very general

45:23

level? It's hard to say. It's hard to say, right?

45:26

It's good to be able to take ideas and

45:28

flip them around like a Rubik's cube and take them

45:30

apart. Notice the two parts are the

45:32

same shape or something like that. But

45:35

I don't think I actually learned that in

45:37

philosophy classes. I think you would have, I would

45:39

have learned that in classes about anything

45:41

hard. I mistakenly

45:44

thought that

45:45

you could just go and

45:48

learn the most abstract

45:50

truths, right? It sounds great

45:52

to a high school student. Why do I have

45:54

to learn all this specific crap? I'll

45:57

just learn the most general truths.

45:59

And needless to

46:02

say, that's one of those things that sounds

46:04

too good to be true, and it is. Because if you go and

46:06

look in philosophy classes, I

46:08

remember when Bill Clinton was saying,

46:10

you know, it depends what the meaning of is is. And I'm

46:12

like, hey, that's what I majored in. What

46:15

the meaning of is is, literally. Which

46:19

kinds of ideas come more naturally to you

46:21

while you're walking?

46:22

Which kind? Yes.

46:25

You're not painting ideas, but some kind of ideas.

46:28

Because you're written about how using

46:30

walking to learn ideas is

46:33

a good thing. But which kinds of ideas? This cross-sectional

46:35

variation, right?

46:37

Yeah. Well, ideas

46:39

about whatever you're thinking about. Until

46:41

a few years ago, I was working very intensely

46:44

on programming. And

46:46

so I had the problem

46:48

I was working on loaded into my head. And

46:51

whenever I was doing something

46:53

without any interruptions, I would start to think about that.

46:56

So I think it's good for whatever you happen

46:58

to be thinking about. Like mathematicians apparently

47:01

walk a lot. I find it best for

47:04

learning from what the other person knows.

47:06

Not so good for my own ideas. What? When

47:09

you're walking with someone else. With someone else, and I'm talking with them,

47:11

and I learn from them better, I

47:13

don't find that I'm very generative when

47:15

I'm walking.

47:16

Walking makes all kinds of thinking better. I've

47:18

seen images like MRI images

47:21

or something like that of brain activity. I

47:23

don't know how they do MRM, just some kind of images

47:25

of brain activity. And your

47:27

brain is definitely more active when you're walking. So

47:30

YC office hours, classic YC office

47:32

hours, were to walk down the block

47:35

and talk as you walk. Which also

47:37

has the side benefit that

47:39

you're side by side and not looking

47:41

the other person in the face, which I think may

47:44

be better.

47:45

It's certainly better. It's like confession

47:47

in the church. You don't see the priest, or

47:50

you're on a therapist's couch, probably

47:52

you're not looking right at them, or vice versa. Or

47:54

you're driving your kid back to boarding school.

47:57

And they'll say things they would not otherwise have told you.

47:59

Or talk.

47:59

at all? Yes.

48:03

What are the best environments for learning

48:05

while walking?

48:07

Urban British countryside or how

48:09

do you optimize this dose?

48:11

Okay, I actually have.

48:13

I think Britain is wonderful for learning while walking

48:15

because it's never too hot, so

48:18

you heat up while you're walking. It doesn't pour

48:20

with rain on you like it did today. I was drenched

48:23

today. It ends in four minutes, right?

48:25

You can bring an umbrella.

48:27

You run to the gazebo. But there is

48:29

a gazebo. There was no gazebo where I was. This

48:31

morning, I think it's good to always

48:33

walk in the same place.

48:35

You don't want to see things that distract you, right?

48:37

If you're trying to have ideas, you're not going to get

48:39

ideas from things you see. Probably not.

48:42

Not relevant ones. And so

48:44

you want to walk in the same place and

48:46

it should be something where there's no distractions.

48:48

So I would think the countryside where I go walk

48:51

is on this common, this preserved

48:54

medieval common. To an American, it

48:56

would look like a large park. And

48:58

it's the perfect thing. I just always take the same

49:00

route. There's not much on it. So

49:04

I see grass and trees. That's about it.

49:07

What's the most important thing you've learned about soundproofing?

49:10

Boy, there's a good question. I've learned a few things about

49:12

soundproofing. Can I only say one thing?

49:15

One thing is fine. You said what's the

49:17

most? Oh, you can say more than one. Okay.

49:19

One is that sound comes through

49:21

holes. And so it doesn't come through your

49:23

walls. It comes through your windows probably in most

49:25

places. And so if you fix the holes, you

49:28

fix the noise problem. The other

49:30

thing I've learned is that the solution

49:33

to sound, basically the

49:35

solution, if it's not, it's

49:37

either multiple layers in the case of windows or

49:39

simply mass.

49:40

You make some big thick

49:42

door and make it have

49:44

hinges that make it sink down. That's what

49:46

recording studios have. So when the door opens,

49:48

it rises up a little bit. And when it closes, it goes right

49:51

down onto the floor. So great big,

49:53

heavy doors, multiple

49:55

pained windows. And weirdly

49:57

enough, I've noticed I've managed to.

49:59

soundproof some places so effectively

50:02

that I've noticed this phenomenon. You only

50:04

notice with soundproofing, all kinds of

50:06

things make annoying noises. You never noticed.

50:08

That's right. It's a war of attrition of sorts. Yeah,

50:11

exactly. Exactly.

50:13

Soundproofing is worth it though. Quiet is really

50:16

good. At least for me.

50:18

There's an optimal level of fame. Do you feel you

50:20

have too much or too little?

50:22

What

50:24

is the optimal level of fame? I suppose it's when

50:27

you can get resources you need or something like

50:29

that, or if there's someone you need to talk to, they'll

50:31

talk to you. So I can now talk to

50:33

most people I need to talk to. If I want to talk

50:35

to somebody, I can find somebody who

50:37

will introduce me. So

50:39

that must be enough. But are you past

50:41

the optimum?

50:43

I don't know. I don't know. That's the

50:45

thing. This sounds very arrogant, but, but

50:47

I realized this with Y Combinator. I

50:49

realized

50:51

that Y Combinator had become

50:54

famous long after

50:56

it had become famous because as far as I was

50:58

concerned, I was just doing what we'd always been

51:00

doing. We would, every six months, we would get

51:02

all these applications. We'd have to find the needle

51:04

in the haystack, right? Get

51:06

all these startups, help them grow,

51:09

find investors for them. And then it would start again.

51:12

It didn't seem like YC was

51:14

any different. It was the same

51:16

building, the same people. We would get more

51:18

startups, right? Meanwhile, YC

51:21

was like

51:21

starting to be considered as this giant

51:24

gatekeeper for Silicon Valley or something like that.

51:26

Jessica knew it was different, right? I

51:28

don't think you're not aware. Famous

51:31

people don't know how famous they are unless they're

51:33

experts on it, like movie stars or something like

51:35

that. They're always basically taken by surprise.

51:38

And so we were especially taken by surprise

51:40

because the thing is the companies

51:43

we funded would

51:45

grow until they had thousands of employees.

51:47

But YC itself didn't grow.

51:50

Like the value of the portfolio grew with these

51:52

giant companies, but we didn't see it. We

51:54

were still just a few people doing the same thing we'd

51:56

always been doing. So how could we be famous?

51:59

But I discovered.

51:59

Like that was one of the biggest mistakes

52:02

I made with YC. I didn't

52:04

realize how many people were watching us. I

52:06

thought we were just, we could just keep doing what we were

52:08

doing and nothing really mattered. And

52:10

why was that a mistake per se? Maybe it was better

52:13

being oblivious. No, because everybody, when

52:16

basically anybody outside

52:19

Silicon Valley who wants to blame Silicon

52:21

Valley for something, who do they blame? They've

52:24

never heard of the people who are actually powerful in

52:26

Silicon Valley. They only know a handful

52:28

of people who have consumers to make new consumer brands, me

52:30

among them. So basically the

52:32

world sucks because of tech and tech

52:35

sucks because of Paul Graham. They never

52:37

heard of any of the other people. I don't seem

52:39

to get quite so much of that anymore. I

52:41

don't know. I'm glad about that. But with

52:43

YC definitely, I didn't know

52:45

how prominent

52:46

YC was becoming and how many

52:49

people would be out to get us as a result.

52:52

Very last question. In my view, a life properly

52:54

lived is learn, learn, learn all the time.

52:57

That's what Charlie Munchers. Yes. What

52:59

have you recently been learning about? Oh. Other

53:02

than soundproofing. Tintoretto. And

53:06

what are you learning? Vasari

53:08

had a very low opinion of him. Vasari

53:10

is unreliable on most things, right? I don't

53:12

know. I don't know. He way overrated his

53:14

patrons, the Medici. Yeah. You

53:17

have a thing about the Medici clearly. He

53:19

said that Tintoretto was too

53:22

independent minded. That Tintoretto

53:24

was this sort of mad genius

53:26

and that he would have been better if he

53:29

had constrained his creativity

53:31

and stayed within the limits of proper

53:33

art. You know what I mean? Very Florentine

53:36

sort of idea.

53:37

I think Tintoretto would have looked down

53:39

on Vasari as a minor league

53:42

artist. But that was interesting.

53:44

That was interesting to learn. That's how at

53:46

least some of Tintoretto's contemporaries

53:49

viewed him and Vasari

53:51

in particular. And a Y

53:53

Combinator co-founder is not gonna buy

53:55

that argument, is he?

53:57

Which was? That he was

53:59

too radical. and too off on his own. Oh, I

54:01

thought you meant Jessica wasn't going to agree

54:03

with me about Tintoretto. There's two of you

54:05

that neither of you would agree with Vasari. I

54:08

don't know. I'm now going to look.

54:10

I'm now going to look. I never thought about this, but

54:12

I was just looking at some Tintorettos. I was just

54:14

in Venice looking at the Squallagrande

54:17

di San Marco, perhaps Squallagrande or

54:19

something, where all those Tintorettos are.

54:22

And they were so dirty. It was

54:25

hard to tell what that paintings

54:27

actually looked like. But I'm going to go look.

54:29

I'm going to go look and see if they seem

54:32

freakish. Paul Graham,

54:35

thank you very much. Thank you. Boy,

54:37

that was so many hard questions. Thanks

54:42

for listening to Conversations with Tyler.

54:45

You can subscribe to the show on Apple Podcasts,

54:48

Spotify, or your favorite podcast

54:50

app. If you like this podcast,

54:52

please consider giving us a rating and leaving

54:54

a review. This helps other listeners

54:57

find the show. On Twitter, I'm at

54:59

Tyler Cowan, and the show is at

55:01

Cowan Convo's. Until next

55:04

time, please keep listening and learning.