0:00

Casey Newton, welcome to Search Engine. CASEY NEWTON Hi! Also,

0:02

is this another week where you're supposed to be on

0:04

vacation? Uh, not really. I mean,

0:06

like, today is a work day for me. I

0:09

am supposed to be off starting tomorrow, but

0:11

I fully expect I'll be making between three

0:13

and seven emergency podcasts in the next week.

0:16

Who invented the emergency podcast? CASEY

0:18

NEWTON You know what? It, like...

0:21

An emergency podcast is, like, a stupid, like, self-aggrandizing

0:23

name, but the point of a podcast is it's,

0:26

like, people that you hang out with, like,

0:28

during these moments in your life. So,

0:30

when something happens in a world that you care about, you

0:32

actually want to hang out with your friends so you talk

0:34

about that stuff with. Yeah, and I actually, I get a

0:36

real thrill when I see an emergency podcast. I do have

0:38

this joke, which is, like, there's certain things that if you

0:40

put them in front of another word, it negates the meaning

0:42

of the word, and podcast is one of them. Like, podcast

0:45

famous, you're not famous. Emergency

0:47

podcast, it's not an emergency. But

0:50

I do get the adrenal thrill of an emergency podcast.

0:52

Do you think podcast cancels out more words than,

0:54

like, most other words? Yes. I

0:56

think if you've had to put podcast in front of

0:58

it, it's not that thing anymore. And podcast successful. Uh...

1:02

-...ha-ha-ha-ha-ha-ha-ha. This

1:05

week on Search Engine, an emergency podcast. Can

1:08

it be an emergency podcast two weeks after the news event? Sound

1:10

off in the comments. But this week, our

1:12

urgent question, who should actually

1:15

be in charge of artificial intelligence? That's

1:17

after some ads. How

1:19

did you actually sleep last night? If.

1:22

It wasn't an absolute dream. Then you

1:24

need to upgrade to the softest, most

1:27

luxurious sheets from bowl and branch. They're

1:29

made from talks and for you organic

1:31

cotton and get softer with every single

1:34

wash. Millions of Americans are sleeping better

1:36

and their signature sheets and right now

1:38

you can take twenty percent off so

1:40

hurried A B O L L and

1:43

branch.com and use Code Odyssey for twenty

1:45

percent off today. Exclusion Supply see. Side

1:47

for details. All

2:15

right, here's a question to make

2:17

pretty much any room uncomfortable. Get

2:19

everybody's attention and ask, who

2:21

do we think should be in charge here?

2:24

Do we all agree that the right person is

2:26

running things right now? Who

2:29

should get to make the final decision in your

2:31

family, in your workplace? Should

2:34

one person really be in charge? Should power be

2:36

shared? Sharing sounds good. Okay.

2:40

With who? How much? According

2:42

to what criteria? Look,

2:45

sometimes we ask the fun questions on this

2:47

show about toxic airplane coffee or the ethics

2:49

of cannibalism, but these questions about

2:51

power, I don't think these are

2:53

the cute ones. These are the

2:55

questions that start revolutions. These

2:58

are the questions that transform places or

3:00

sometimes destroy them. Who

3:03

should be in charge? Our

3:07

country was founded as an answer to that

3:09

question. We're told by the

3:11

third grade that America is a democracy.

3:13

The people are in charge. At

3:16

junior high, they walk that back a little. They

3:18

tell us it's a representative democracy, which is

3:21

a bit different, much less exciting. Just

3:24

because our country is a representative democracy, it

3:26

doesn't mean every institution in our country

3:28

will be one. There's

3:30

this word governance, which is so boring

3:32

your brain can't even save it. But

3:35

ironically, it refers to the most interesting thing in

3:37

the world. Who is in

3:39

charge of you? Most

3:43

American businesses have somewhat funky governance structures,

3:45

which we stole from the British. The

3:48

typical corporate governance structure goes like this. There's

3:51

a boss, CEO, with most of the power.

3:54

But they're accountable to a board above them, a

3:56

small group of people who can depose them, at

3:58

least in theory. And the

4:01

board usually represents the shareholders. Often

4:03

the shareholders even vote to elect

4:05

the board. This structure

4:07

of collective decision-making, of voting,

4:10

of elections, it

4:12

has existed and evolved since way

4:14

before American democracy. The

4:17

corporate board model comes from England in the

4:19

1500s. Back

4:21

then, England was a monarchy, but its

4:23

companies were not. They were like, not

4:26

democracies, but democracy-esque

4:29

organizations existing in a country

4:31

devoted to the rule of the king. They

4:35

represented a different answer to this who

4:37

should be in charge question. We

4:40

took that corporate structure with us when we

4:42

divorced England, and in 1811, corporations

4:45

really took off in America. That

4:47

year, New York State became the first to

4:49

make it legal for people to form a

4:51

corporation without the government's explicit permission. Over

4:54

the next 200 years, corporations have become

4:56

very powerful. And in

4:58

that time, their CEOs have learned and

5:00

taught one another how better to consolidate

5:02

power. CEOs today,

5:05

particularly the CEOs of big

5:07

tech companies, are less likely

5:09

to answer to their boards or to their

5:11

shareholders, if they even have them. These

5:14

days, in America, our country is

5:16

a democracy, and the corporations are

5:18

the exceptions. Not monarchies

5:21

exactly, but little monarchy-esque

5:23

organizations in a country devoted to

5:25

the rule of the people. Who

5:29

should be in charge? In

5:32

America, we know we don't trust kings, but

5:34

we don't always trust the people. So

5:37

for now, the people sort around the country,

5:39

and the techno kings mostly run into businesses.

5:43

But the tension about who should hold

5:45

power remains unresolved. It

5:47

crackles. Sometimes it erupts in minor

5:49

revolutions in all sorts of places.

5:52

And exactly two weeks ago, it erupted

5:55

at a technology company. Breaking

5:59

news. incredibly

8:00

world-altering powerful. So

8:03

this never resolved question, who

8:06

should exercise power and how? It

8:08

just got even more complicated. Because

8:10

now we have to decide which people

8:13

or person should be in charge of

8:15

artificial intelligence, a technology designed

8:17

to become smarter than human beings.

8:21

Well, let's take a step back. OpenAI

8:23

is the most

8:25

important company of this generation.

8:28

For the past two weeks, as the story

8:30

has unfolded, I've been talking to Casey Newton,

8:32

who publishes the excellent newsletter, Platformer. When

8:35

we spoke last week, he was reminding

8:37

me exactly how important the story of

8:39

OpenAI is, even before this latest

8:42

chapter. It is not

8:44

super young, it was founded in 2015,

8:47

but with the launch of Chat GPT

8:50

last year, it started

8:52

down a road that very few companies

8:54

get to start down, which is the

8:56

road to becoming a giant consumer platform

8:59

that you had mentioned in the same

9:01

breath as a Google or a Facebook

9:04

or a Microsoft. And

9:07

when you are seeing that in the case of

9:09

Chat GPT, you have a product that is being

9:11

used by 100 million people a week. And

9:14

you have a CEO who has become the

9:16

face of the industry. Sam Altman has become

9:18

essentially the top diplomat of the AI industry

9:21

over the past year. The

9:23

number of reasons that

9:26

you would fire that person with

9:28

no warning is just

9:30

extremely small. And

9:33

the idea that even after he was fired,

9:35

you still would not say with any specificity

9:38

what he did is even smaller. Those

9:40

are just some of the reasons why this has just been such

9:42

a crazy story. And when you saw

9:44

it, how did you get the news? I'm

9:49

happy to tell you that story. My

9:51

parents were in town and they asked if we

9:53

could have lunch. And I

9:55

thought, I'm gonna take them to a really

9:57

nice lunch in San Francisco at an institution.

10:00

called the Zuni Cafe. Zuni

10:03

Cafe is known for a roast chicken

10:05

that is so good but it does

10:08

take an hour to cook. So we

10:10

order a few snacks and my parents being my

10:13

parents said hey why don't we get a couple

10:15

cocktails and a bottle of wine and I said

10:17

guys it's 11 45 a.m.

10:19

but you know what let's

10:21

do it. So bottle of

10:23

wine comes the cocktails come we have

10:25

our snacks and we're waiting for the

10:27

chicken and I think I'm

10:30

gonna use the restroom and I get up

10:32

to use the restroom and look at my

10:34

phone and I see the

10:36

news because 78 people have been texting me saying

10:41

holy motherfucking shit what

10:44

is happening? And

10:47

so I go back to the

10:50

table and explain to my parents

10:52

everything that I have to about opening eyes, demoing

10:54

and everything and then I walk outside and I

10:56

get on a Google Meet with my podcast post

10:59

because of course we're gonna need to do an

11:01

emergency episode and I just stare at my parents

11:03

through the window and watch the chicken arrive at

11:05

the table and them start to eat it. So

11:07

you never got to eat the chicken? I did

11:10

well eventually the Google Meet ended and I got

11:12

to have some chicken and it was delicious but

11:15

there was a while there where I was quite hungry and

11:17

jealous of them. And so you

11:19

guys the initial thing is just like holy

11:21

crap this was nuts and like

11:23

was your instinct oh there's

11:26

going to be like like the board is

11:28

gonna come forward and say like hey he's

11:31

done something awful like were you waiting for

11:33

a shoe to drop? Absolutely

11:35

because there again

11:38

the number of reasons why the board would

11:40

have fired him is just very small right

11:43

when I saw it my thought was it's

11:45

always either money or sex is

11:48

why a high-profile person loses their

11:50

position right? And the

11:52

board's description didn't really lean one

11:54

way or another in that direction.

11:56

I started to you know people just started

11:58

to speculate for a while theories at me.

12:02

But again, because this was such a

12:04

consequential move, the expectation was always that

12:06

even if the board wouldn't say it

12:09

in their blog post, they would at

12:11

least tell their top business

12:13

partners, they would tell the top executives at

12:15

OpenAI, and then it would just sort of

12:18

filter out to the rest of us what

12:20

actually happened. But days later, that was still

12:22

not the case. Even after the company was

12:24

in open revolt with 95% plus of the

12:27

company threatening to lock out the door if

12:29

the situation wasn't reversed, the board still wouldn't say

12:31

what happened. Have you ever

12:33

seen anything like that before? Um,

12:37

well, I mean, look, CEOs get fired.

12:39

There's actually an argument that CEOs don't

12:41

get fired enough, right? Like we live

12:43

in the Silicon Valley bubble where we

12:46

have a cult of the founder, and

12:48

there is a very strong feeling that

12:50

the founder should almost never be removed

12:52

because the company cannot survive without them.

12:54

And so it's always very dramatic when

12:56

a founder gets removed, right? Like probably

12:58

the biggest founder drama I can remember

13:00

before this one was the removal of

13:02

Travis Tallinnick from Uber. The difference

13:04

there was that Uber had been involved in a lot

13:07

of public wrongdoing before

13:09

he was removed. And so there was

13:11

kind of a steady drum beat of

13:13

stories and people calling for him to resign before

13:16

that happened. But even then, his board members

13:19

turned on him. And in Silicon Valley, that

13:21

is a taboo for for someone that you

13:23

appoint to your board and you say, be

13:25

a good steward of my company. The expectation

13:27

is you are never going to remove the

13:29

founder. And in fact, we have other Silicon

13:31

Valley companies where the founders have insulated themselves

13:33

against this by just designing the board differently.

13:35

So Mark Zuckerberg has a board that

13:37

cannot remove him. Evan Spiegel at Snap

13:40

has a board that cannot remove them.

13:42

So again, that's just kind of the

13:44

way things operate here. And

13:46

how does a founder choose their board

13:48

members? So the most

13:50

common way is that if UPJ run a venture

13:52

capital firm, which I do think you should, they

13:55

need to talk to you about that. So I

13:57

come to you and I want to. I

14:00

want to get some of your money. You say,

14:02

okay, I will buy this percentage of your company

14:04

for this amount, but I want to take a

14:06

seat on your board. And the

14:09

idea is, hey, if I would have a

14:11

lot of money locked up in your company,

14:13

I want to be able to have a

14:15

say in what happens there. I see. And

14:17

normally speaking, normal company, Facebook, whatever, you've

14:20

got a board, they have a little

14:22

bit of a say because it's their money,

14:24

but a powerful founder of a powerful company

14:26

will set it up so that they don't

14:28

have much of a say. Yeah,

14:31

basically, they create a different

14:33

kind of stock, and

14:35

they will control the majority of that

14:37

stock. And that stock has some sort

14:39

of super voting powers. So when the

14:41

board goes to vote on something, their

14:43

votes will never exceed the number of

14:45

votes cast by the founder. The

14:50

OpenAI board was set up very differently, which

14:52

I'm sure we'll talk about. And so it

14:54

made this sort of thing possible, but absolutely

14:56

nobody saw coming. After

15:02

the break, the strange origin story

15:05

of OpenAI. And how it

15:07

led to the events of this month. Hey,

15:41

it's Ryan Reynolds, owner and user of MidMobile,

15:43

and I am recording this message on my

15:46

phone. I'm literally on my mid phone. Why?

15:48

Because fancy recording studios cost money. And if

15:50

we spent money on things like that, we

15:52

couldn't offer you screaming deals. Like if you

15:54

sign up now for three months, you get

15:56

three months free on every one of your

15:58

plans, even unlimited. at mintmobile.com/Switch.

16:01

Limited time, new customer offer. Activate within 45 days. Additional

16:04

taxes fees and restrictions apply. Unlimited customers using more

16:06

than 40 gigabytes per month will experience lower speeds.

16:08

Video streams at 480p. See mintmobile.com

16:10

for details. Why is Instacart

16:12

the holiday rescue app? Because you

16:15

can get all your seasonal decor

16:17

delivered instead of having to drive

16:19

to 12 different stores. Candles and

16:22

candy canes delivered. Reats and reindeer

16:24

delivered. Lights from lows delivered. And

16:26

since I know you're going to

16:29

ask. Inflatable snowman delivered. So

16:31

this season stay in and get decked

16:33

out. Download Instacart the holiday rescue app

16:35

to get free delivery on your first

16:38

three orders. Offer valid for a limited

16:40

time. $10 minimum per order, additional terms

16:42

of pay. Thank

16:59

you all for sticking around this afternoon. We

17:01

had some great conversations and we're hoping to

17:03

have another great one. It's

17:05

the fall of 2015. Just a couple

17:07

months before OpenAI would be willed into

17:09

existence. Elon Musk and

17:11

Sam Altman are on stage together at

17:13

this conference on a panel called What

17:16

Will They Think Of Next? Questions is

17:18

about artificial intelligence. And one question they

17:21

asked is about AI. This technology that

17:23

in 2015 still felt way

17:25

off in the future. And Elon could share

17:27

with us their positive vision of

17:30

AI's impact on our coming life. Sam

17:33

Altman, who at the time is the

17:35

head of Y Combinator. He

17:37

goes first. I think there are, the science fiction

17:39

version is either that we enslave it or it

17:41

enslaves us. But there's this happy symbiotic vision, which

17:44

I don't think is the default case, but what

17:46

we should work towards. I think already. Sam's dressed

17:48

like a typical 2015 startup guy. Blaze

17:51

are colorful sneakers. What I

17:53

noticed is his eyes, which to me always

17:56

look concerned. Like someone whose car just made

17:58

a weird noise at the beginning of a- long road

18:00

trip. In 2015,

18:02

Sam Altman has a reputation as

18:05

a highly strategic, deeply ambitious person,

18:08

but also someone a bit outside of

18:10

the typical Silicon Valley founder mold. He's

18:13

made a lot of money, but says he's donated most of

18:15

it. He's very obsessed with universal

18:17

basic income. The

18:19

kind of person who tells the New Yorker that

18:21

one day he went on a day-long hike with

18:23

his friends and during it made

18:26

peace with the idea that intelligence might

18:28

not be a uniquely human trade. He

18:30

tells the magazine, quote, there are certain

18:33

advantages to being a machine. We humans

18:35

are limited by our input output rate. He

18:37

says that to a machine, we

18:40

must seem like slowed down whale

18:42

songs. But

18:44

I don't think there's any human left that

18:47

understands all of how Google search results

18:49

are ranked on that first page. It really

18:51

is. On stage, Sam's pointing out the

18:53

ways in which AI is already here.

18:55

We're already relying on machine learning algorithms we

18:57

don't entirely understand. Google search

18:59

results or the algorithms that run

19:02

dating websites. In this case, the

19:04

computer matches us and then we have babies. But

19:06

then have babies and so on in effect, you

19:08

know, you have this like machine learning

19:10

algorithm breeding humans. And so really,

19:12

I mean, you do. And

19:15

so there's this and then, you know, those people like

19:17

work on the algorithms later. And

19:19

so I think the happy vision of the future

19:21

is sort of humans and

19:24

AI in a symbiotic relationship, distributed AI,

19:26

where it sort of empowers a lot

19:28

of different individuals, not this

19:30

single AI that kind of governs everything that

19:32

we all do that's, you know,

19:35

a million times smarter, a million times smarter than any other

19:37

entity. So I guess we should work towards. Elon

19:39

goes next. I agree with

19:41

what Sam said. I mean, we

19:44

are effectively already a human

19:47

machine, collectives symbiotes.

19:50

Like this, like like a like

19:53

a giant cyborg. That's

19:56

actually what society is today. No

19:59

one in the room. cool-headed

22:00

and averting apocalypse instead of like steering

22:02

wildly into it is not a thesis

22:04

that survives modern times. Exactly.

22:08

And so then they think like well, maybe

22:10

we do it as a for-profit company, right?

22:12

Like Sam Altman at the time was running

22:14

Y Combinator, which is the most famous startup

22:16

incubator in the United States. It's responsible for

22:19

Stripe and Dropbox and a number of other

22:21

famous companies. So the obvious thought was well

22:23

why don't we just do it as a

22:25

venture-backed startup? But the more they

22:27

think about it, they think well, gosh, if we're

22:29

again building a super intelligence we don't want

22:31

to put that into the hands of one

22:34

company. We don't want to concentrate power in

22:36

that way because we think this thing could

22:38

be really beneficial. And so we want to

22:40

make sure that everyone reaches the benefits of

22:42

that. So that leaves them with a

22:44

nonprofit and that winds up being the direction to

22:46

go in. And this might be

22:48

jumping ahead but like my

22:51

guess would be like one of the

22:53

reasons as I understand it the technology

22:55

usually moves at like the fastest pace

22:57

it can instead of the most judicious

22:59

pace it can is because if you're

23:02

moving slowly someone else will move more

23:05

quickly, more faster, faster, faster.

23:07

More faster quickly. More faster

23:09

quickly. And so why did

23:11

the responsible company succeed this

23:13

time? Well,

23:17

it had some advantages. One,

23:19

it was probably the first mover

23:21

in this space that was not

23:23

connected to a giant company.

23:26

So Google, for example, already had

23:28

AI efforts underway. Facebook also had

23:30

AI efforts underway. This was

23:32

really the first serious

23:34

AI company. I

23:36

think that because it was a

23:39

startup and because it was a nonprofit,

23:41

it attracted talent that would be less

23:44

inclined to go work for a Google

23:46

or Facebook. Right. There are recruiting advantages

23:48

that come with telling people we do

23:50

not have a profit motive. We are

23:53

a research lab and our intentions are

23:55

good. And so they attracted a lot

23:57

of really smart people. They also had

24:00

the imprimatur of Elon Musk, who

24:02

was one of the co-founders, who

24:04

was a much more reliable operator

24:06

in 2015 than he is today.

24:10

And that served as a powerful recruiting

24:12

signal. And so all those people

24:14

get together and they get to work and they started

24:16

working on a bunch of things and not everything worked

24:18

out. They had a hardware division at one point, like

24:21

they were interested in robotics and it just kind of

24:23

fizzled. But then they started working on

24:25

this GPT thing and things got better for them.

24:29

According to reporting from Semaphore, in early

24:31

2018, Elon Musk makes a bid to

24:33

become OpenAI's president. The board shoots him

24:35

down. Soon after, Elon

24:37

quips OpenAI, publicly setting a conflict

24:39

of interest with Tesla. Semaphore

24:42

also reported that Musk promised to invest

24:44

$1 billion in OpenAI. When

24:47

he left, he said he would keep the promise. He

24:50

didn't. So OpenAI was short

24:52

on money, which was a problem because the

24:54

next year, 2019, the company announced

24:57

their expensive new project, GPT2,

25:00

a much more primitive ancestor to the

25:02

chat GPT likely used. Training

25:04

even this model was hugely expensive

25:07

and OpenAI realized it would not be able

25:09

to get by on donations alone. One

25:12

thing that we've learned over the past

25:14

year, as all of us have been

25:16

educating ourselves about large language models like

25:18

chat GPT, is that they're incredibly expensive

25:20

to run. I talked to a former

25:22

employee of OpenAI this weekend who described

25:25

the company to me as a money incinerator. They

25:27

don't even make podcasts? They don't even make podcasts.

25:29

That's how expensive they are. They're losing money without

25:31

even making podcasts, BJ. Can

25:33

you imagine? If

25:38

you've ever used chat GPT, you've

25:40

cost OpenAI money. Some

25:44

estimates are around 30 cents for you asking

25:46

chat GPT a question. It

25:48

has 100 million users a week. You can imagine

25:50

how much money they're losing on this thing. Is

25:52

that 30 cents computing power? It's

25:54

computing power. Yes. I

25:57

believe the technical term is an inference cost.

26:00

you type in your question to chat

26:02

CPT and then it sort of

26:04

has this large language model and

26:06

it generates a

26:08

sort of series of predictions as to what

26:10

the best answer to your question will be

26:13

and the cost of the electricity and the

26:15

computing power is about 30 cents. Got

26:17

it. So the technology is super expensive to run.

26:19

So even in the early days, they're just burning

26:22

money really quickly. Yes. And

26:24

so they have a problem, which

26:26

is that there is no billionaire,

26:29

there's no philanthropy, there's no foundation, there's no government

26:31

that is going to give them 100 to $200

26:33

billion to try to

26:36

get their project across the finish line. So

26:38

they turn back to the model that they

26:41

had rejected, the for-profit model. But

26:43

instead of just converting a nonprofit

26:45

into a for-profit, which is incredibly

26:48

difficult to do, they take a

26:50

more unusual approach, which

26:52

is that the nonprofit will

26:54

create a for-profit entity, the

26:57

nonprofit board will oversee the

26:59

for-profit entity and the

27:01

for-profit entity will be able to raise

27:03

all those billions of dollars by offering

27:05

investors the usual deal. You

27:08

give us some amount of money

27:10

in an exchange for some percentage

27:12

of the company or for our

27:14

revenues or our profits and

27:16

that will enable us to get further

27:18

faster. March 2019,

27:21

OpenAI publishes a blog post announcing

27:23

the change. The

27:25

nonprofit will now have a for-profit company attached

27:27

to it and the CEO will

27:29

be Sam Altman. He will not,

27:31

however, take any kind of ownership stake, an

27:34

almost unheard of move for a Silicon Valley

27:36

founder. The blog post

27:38

lists the names of the nonprofit board members who

27:40

will keep the for-profit in check. Sam

27:42

Altman is on the OpenAI board along

27:45

with some other executives like OpenAI's chief

27:47

scientist Ilya Sutskivar. There's some

27:49

Silicon Valley bigwigs, LinkedIn's Reid Hoffman,

27:52

Quora's Adam DiAngelo, but also importantly

27:55

there are some effective altruists like

27:57

Holden Karnofsky and scientist engineer Tasha

27:59

McCauley. If the idea

28:01

is that this board is going to be like part

28:04

of the idea is they are a hedge against AI going in

28:06

the wrong direction and they're going to try to get really

28:09

like skeptical, smart

28:11

people, like how serious are these

28:13

people as artificial intelligence thinkers? I

28:16

mean, I think they do have credibility. You

28:18

know, I don't know who in that year

28:20

would have been considered the very best thinkers

28:23

on that subject. But I would note that

28:25

in the years since Reed Hoffman left the

28:27

board to start his own AI journey with

28:29

a co-founder, it's called Inflection AI, they've been

28:32

doing good work. Holden

28:34

Karnofsky was the CEO of

28:36

Open Philanthropy, which is one

28:38

of the effective altruist organizations.

28:41

They are a funder of funders, so they

28:43

like give money to scientists to research things.

28:45

But Holden was essentially part of the group

28:47

that were some of the very first people

28:49

to worry about existential AI risk. At

28:52

a time when absolutely no one was

28:54

paying attention to this, Holden's organization was

28:56

giving researchers money to study the potential

28:59

implications of AI. So

29:01

there were people on that board who

29:03

were thinking a lot about these issues

29:05

before most other people were. And,

29:08

you know, we can debate whether they had

29:10

enough credibility, but like, certainly they

29:12

were not, you know, just

29:14

a bunch of dumb rubber stamps for Sam Altman. At

29:19

this moment in 2019, OpenAI,

29:21

the nonprofit company controlling a for-profit

29:23

subsidiary, was a little unusual, but

29:26

that unusual state of affairs would only

29:28

become truly absurd a few years later.

29:32

November 2022, OpenAI releases,

29:34

without much fanfare, without a very attractive

29:36

name, a product called

29:38

ChatGPT. Within five

29:41

days, ChatGPT has a million users.

29:43

Two months after launch, it has 100

29:46

million monthly active users. At

29:49

that point in time, it's the fastest

29:51

growing consumer app in history. There's

29:53

a new board in town and it's

29:55

staking the world by storm. ChatGPT was launched

29:57

by OpenAI on the 30th of November.

30:00

gaining popularity for its ability to

30:02

craft emails, write research papers, and

30:04

answer almost any question in a

30:06

matter of seconds. The CEO, Sam Altman,

30:08

is just 37. OpenAI

30:11

becomes the leading AI company. Sam Altman

30:13

becomes not just the face of OpenAI,

30:16

but for many people, the face

30:18

of AI itself. He's the rock

30:20

and roll star of artificial intelligence.

30:22

He's raised billions of dollars from

30:25

Microsoft, and his early back has

30:27

included Elon Musk and Reed Hoffman.

30:29

As chat GBT takes over the internet, Sam goes

30:32

on a world tour. Israel, Jordan,

30:34

the UAE, India, South Korea, Japan,

30:36

Singapore, Indonesia, and the UK. You

30:39

must be rushed off your feet here in

30:41

the middle of an enormous world tour. How

30:44

are you doing? It's been super great,

30:46

and I wasn't sure how much fun I

30:48

was gonna have. By May of this year,

30:50

AI has become important enough, fast enough, that

30:52

Sam, AI's chief diplomat,

30:55

is testifying in front of Congress. Mr.

30:57

Altman, we're gonna begin with you, if

30:59

that's okay. Thank you. Thank

31:02

you, Chairman Blumenthal, ranking member Holly,

31:04

members of the Judiciary Committee. He's

31:07

dressed in a navy suit, but now with

31:09

normal gray shoes. His eyes look

31:11

still worried. They're registering congressional levels

31:13

of worry. OpenAI was founded on

31:16

the belief that artificial intelligence has

31:18

the potential to improve nearly

31:20

every aspect of our lives, but

31:22

also that it creates serious risks we have to

31:24

work together to manage. We're

31:26

here because people love this technology. We

31:29

think it can be a printing press moment. We have to

31:31

work together to make it so. OpenAI

31:33

is an unusual company, and we set it

31:35

up that way because AI is an unusual

31:37

technology. We are governed by

31:39

a nonprofit, and our activities are driven by our

31:41

mission and our charter, which commit

31:43

us to working to ensure that the broad distribution of

31:45

the benefits of AI and to maximize

31:48

the safety of AI systems. Sam

31:50

is telling these congressmen, his likely

31:52

future regulators, but every

31:54

tech CEO has told everyone since the

31:57

invention of fire. Don't

31:59

worry. this under control. But

32:01

what is new here, what you would not

32:03

see with someone like Mark Zuckerberg in Facebook's

32:06

early years, is that Sam's also saying

32:08

he knows that the downside risk of

32:11

the thing he's creating is enormous. Casey

32:14

Newton says that this tension, that AI's

32:16

inventors are also the people who worry

32:18

about its power, that's part

32:20

of what makes this story so unusual. Usually

32:28

the way that it works in Silicon Valley

32:31

is that you have the Rara technologist going

32:33

full steam ahead and you know, sort of

32:35

ignoring all the safety warnings. And then you

32:37

have the journalists and the academics and the

32:40

regulator types who are like, hey, slow down,

32:42

that could be bad, things are the implication.

32:44

That's sort of the story we're used to.

32:46

That's the Uber story. That's the Theranos story.

32:49

What's interesting with AI is

32:51

that some of the people who are the

32:53

most worried about it also identify as techno

32:55

optimist. Okay? Like they're the sort of people

32:57

that are usually like, hey, technology is cool. Let's build

32:59

more of that. Why was that the

33:01

case? Well, they've just looked at the events

33:03

of the past couple of years. They use

33:06

GPT-3 and then they use

33:08

GPT-3.5 and then they use

33:10

GPT-4. Now they're using GPT-4

33:13

turbo. We already basically

33:15

know how to train a next generation

33:17

large language model. There are some research

33:19

questions that need to be solved, but

33:21

we can basically see our way there,

33:23

right? And what happens when

33:27

this thing gets another 80% better, 100% better?

33:30

What happens when the AI can

33:32

start improving itself or can start

33:34

doing its own research about AI,

33:36

right? At that point, this stuff

33:38

starts to advance much, much,

33:40

much, much faster. If

33:42

we can see on the horizon the

33:44

day that AI might teach itself, then

33:47

the question of who's in charge of it right

33:49

now feels pretty important. And

33:51

remember, OpenAI itself had foreseen this

33:54

problem. That's the very reason it had

33:56

created the nonprofit board as a

33:58

safety measure. And the problem for

34:00

Sam Altman in 2023, and that

34:02

while chat GBT had been taking over the world,

34:05

the composition of his nonprofit board

34:07

had changed. Some of his

34:09

natural allies, business minded folks like Reid

34:11

Hoffman, had left the board, which

34:14

had tipped the balance of power over

34:16

to the academics towards the people associated

34:18

with the effective altruism movement. And

34:21

that's what set in motion the coup, the

34:23

very recent attempt by the board to take

34:25

out Sam. When news

34:28

of Sam's firing first broke, the reasonable

34:30

guess was that he tried to push

34:32

AI forward too fast in a way

34:34

that had alarmed the board's safety minded

34:36

people. In the

34:38

aftermath of all this, it's pretty clear that that's

34:40

not what happened. According to

34:42

the Wall Street Journal, here's how things broke down.

34:47

The departure of some of Sam's allies had left an

34:49

imbalance of power. And afterwards, the

34:51

two sides began to feud. One

34:54

of the effective altruists, an academic

34:56

named Helen Toner, co authored

34:59

a paper about AI safety, where

35:01

she criticized OpenAI, the company whose

35:03

board she was sitting on. A

35:05

normal enough thing to do in the

35:07

spirit of academia, but an arguably passive

35:10

aggressive violation of the spirit of corporate

35:12

America. Sam Altman

35:14

confronted her about it. Then,

35:17

sometime after that, some of Altman's

35:19

allies got on slack and started

35:21

complaining about how these effective altruist

35:23

safety people were making the company

35:25

look bad in public. The

35:27

company should be more independent of them, they said,

35:31

on slack. The problem

35:33

is that on that slack channel was

35:35

Ilya Sutskivar, a member of the board

35:38

and someone who is both a sometime Altman

35:40

ally, but also someone

35:42

who is deeply concerned with

35:44

AI safety. How many companies

35:46

have been destroyed by the

35:49

actually already nuclear technology that

35:51

is slack? Anyway,

35:53

two days later, it's Sutskivar who

35:55

delivers the killing blow. Sam

35:58

is in Vegas that Friday out of. Formula

36:00

One race, trying to raise more billions for

36:02

open AI, he's invited

36:04

at noon to a Google Meet, where

36:06

Sudskivir and the other three board members

36:08

tell Oppmann he's been fired. Afterwards,

36:11

like any laid off tech worker,

36:13

he finds his laptop has been

36:15

remotely deactivated. Over

36:20

the weekend, as the company's employees and

36:22

executives get angrier and angrier about the

36:25

coup, they confront Helen Toner, the academic

36:27

who wrote the spicy paper. They

36:29

tell her that the board's actions might destroy

36:32

the company. According to the

36:34

Wall Street Journal, Helen Toner

36:36

responds, quote, "'That would

36:38

actually be consistent with the mission.'" In

36:41

other words, she's saying, "'The board should kill

36:43

the company "'if the board has decided it's

36:45

the right thing to do.'" Casey

36:48

told me that in the days after, a

36:50

public consensus is quickly congealed against these effective

36:52

altruists, who had knowingly damaged the company, but

36:54

then had been unable to provide evidence that

36:57

they'd done it for any good reason. Part

37:00

of the fact that the EAs have a really bad

37:02

reputation right now is that if

37:05

you have not thought that much about AI, and

37:08

it's very hard for you to imagine that a

37:10

killer AI is anything other than a fantasy from

37:12

the Terminator movies, and you find

37:14

out that out there in San

37:17

Francisco, which is already a kooky town, you

37:19

think this, there's

37:21

a bunch of people working for some rich

37:23

person philanthropy, and all they do is they

37:26

sit around all day and

37:28

they think about the worst case scenarios that could ever come

37:30

out of computers, you would think, it seems like

37:33

kind of weird and culty to me. It's

37:35

like, these are like the goths of Silicon

37:37

Valley. There's something almost religious

37:39

about their belief that this AI god

37:41

is about to come out of the

37:43

machine. So these people kind of get

37:46

dismissed. And so when the open AI, Sam

37:48

Altman firing goes down, there's a lot

37:51

of discussion of like, well, here go

37:53

the creepy AI kids again, the gods

37:55

of Silicon Valley and their religious belief

37:58

and killer AI, they've all conspired. fired

38:00

to destroy what was a really

38:02

great business. And that becomes, I

38:04

would say, maybe the first big

38:06

narrative to emerge in the aftermath

38:08

of Sam's firing. We all

38:10

know what happens next. On November

38:13

21st, five days after the shocking

38:15

firing of Sam Altman, he

38:17

gets his job back. He is once

38:19

again CEO of OpenAI. And

38:21

while he won't get to keep his seat on the board, he

38:23

seems to have defeated the goths of

38:26

Silicon Valley. There is

38:28

a big party at OpenAI's headquarters.

38:30

Someone pulls the fire alarm because

38:32

there was a fog machine going.

38:35

But by all accounts, everyone had a great time. They

38:37

stay up very late. And what

38:39

about the board? These people that tried and failed

38:41

to do a coup. So

38:44

three of the four members of the

38:46

board are leaving it. That's Sasha McCauley,

38:48

Helen Toner, and Ilya Sutskivar. A

38:50

fourth member, one of the people who had voted

38:53

to fire Sam, Adam D'Angelo, who's the CEO of

38:55

Quora. He is staying on the board. And

38:57

then they have brought in Larry Summers, who

38:59

is a well-known former U.S. Treasury Secretary, and

39:02

Brett Taylor, who is the former chair

39:04

of the Twitter board, the former co-CEO

39:07

of Salesforce. So the three

39:09

of them are going to appoint a new

39:11

board of up to nine members. And they're

39:13

also going to conduct an investigation into what

39:15

happened. And my hope is that in that

39:17

investigation, we will get some more details finally

39:20

on why the board actually fired Sam Altman.

39:28

After the break, we get

39:30

to the question attended here. Who

39:33

should actually be in charge of AI? This

39:57

episode is brought to you by Shopify.

40:00

Selling a little This episode or a

40:02

lot? by Klaviyo. Do

40:04

your thing however you cha-ching with Shopify,

40:06

the global commerce platform that helps you

40:09

sell at every stage of your business.

40:12

Shopify helps you turn browsers into buyers

40:14

with the internet's best converting checkout. 36%

40:17

better on average compared to other leading

40:19

commerce platforms. Get a

40:22

$1 per month trial period at

40:24

shopify.com/ offer23. The

40:30

platform that powers smarter digital relationships. With

40:32

Klaviyo, you can activate all your customer

40:34

data in real time. Connect seamlessly with

40:36

your customers across all channels. Guide

40:39

your marketing strategy with AI-powered insights,

40:41

recommendations, and automated assistance. Deliver experiences

40:43

that feel individually designed at scale

40:45

and grow your business faster. Power

40:48

smarter digital relationships with Klaviyo. Learn

40:51

more at klaviyo.com. That's k-l-a-v-i-y-o.com.

40:53

That's k-l-a-v-i-y-o.com. That's k-l-a-v-i-y-o.com. That's

40:55

k-l-a-v-i-y-o.com. That's k-l-a-v-i-y-o.com. That's k-l-a-v-i-y-o.com.

41:01

What is the future of

41:04

digital communication? Something

41:08

like once a week in America,

41:10

some institution implodes. And

41:12

it pretty much always goes the same way. A

41:14

confusing private conflict breaks out onto the

41:16

internet. The combatants plead their versions of

41:18

the story to the public. And

41:21

we, reporters, gawkers, people online,

41:24

render a quick, noisy verdict.

41:27

The desire to participate in all this is human

41:29

nature. I am doing it right now.

41:32

You are doing it with me. Neither

41:34

of us chose this system, but we're stuck with

41:36

it. Institutions right

41:38

now are fragile. The internet is powerful. And

41:41

we're all addicted to being entertained. In

41:44

my wiser moments, though, what I try

41:46

to remember is that whoever is actually at fault in

41:48

any of these fights of the week, the

41:51

truth is institutions are supposed

41:53

to have conflict. And

41:55

we all, put together, will disagree. A

41:58

healthy institution is one capable of... of

42:00

mediating those disagreements. When

42:03

we, the public, watch a private fight break

42:05

out online, it's hard to ever

42:07

really know for sure who was actually right

42:09

or wrong. What we can

42:11

know is that we are watching as the

42:13

institution itself breaks. OpenAI

42:16

was set up from the beginning to be

42:18

an unusual kind of company with an unusual

42:21

governance structure. As unusual

42:23

as it was, I'm not convinced

42:25

from the available evidence that the structure

42:27

was the problem. The

42:30

fashion of revolutionaries who took over OpenAI,

42:32

who governed it for a little over

42:34

a weekend, it just seems like

42:36

they didn't know how to be in charge. They

42:39

couldn't articulate what they thought was wrong. They

42:41

couldn't articulate why their revolution would fix it. They

42:44

never even bothered to try to win over the people in the

42:46

building with their mission. They

42:48

thought they saw someone acting like a

42:50

king, and so they acted imperially themselves.

42:54

In the aftermath, what I found myself wondering this

42:56

week was this. This

42:58

new version of OpenAI, could

43:01

it tolerate conflict? Could it

43:03

have, productively, the fights you'd hope would

43:05

take place somewhere as important as this,

43:08

in the rooms we'll never see inside of? Casey

43:11

Newton, who is better at spying into those

43:13

rooms than you or me, he

43:16

says he feels optimistic. I

43:18

think the most important thing about the Newport

43:20

is that Adam DiAngelo is on it. This

43:22

is someone who voted to fire Sam Altman

43:24

and who is still there, and who has

43:26

a say on who else comes onto that

43:28

board, who will have a say on who

43:30

gets picked to investigate all of the circumstances.

43:33

To me, that is like, if

43:35

you're somebody who is worried that, oh no, OpenAI is just going

43:37

to go gas to the pedal. If

43:41

you're worried that OpenAI is going to go foot

43:44

to the gas, why can't I figure out this gas to

43:46

the foot pedal? If you're worried

43:48

that OpenAI is going to go gas to

43:50

the foot pedal, don't worry. Because Adam DiAngelo

43:52

is there. What's how I'm feeling about

43:54

it, anyway? Is that how you're feeling about it?

43:56

Are you feeling like, well, I mean, look, let

43:58

me take a step back. I'm having too much of your

44:00

podcast, PJ. But let me tell you something. I love when you take

44:03

a step back. Okay, great. Take a step back.

44:05

Okay, great. One of the big narratives

44:07

that came out of this whole drama

44:09

was there was the forces of corporate

44:11

money-making and there were the forces of AI

44:14

safety. And the forces of AI safety kicked

44:16

out Sam Altman and then the forces of

44:18

corporate money-making stepped in to ensure that Sam

44:20

Altman would be put back in his role

44:23

to continue the corporate money-making. And

44:26

it is true that the forces of

44:28

capitalism intervened to restore Sam Altman. That

44:30

part is true. But from

44:33

my own reporting, I truly believe

44:35

that the core conflict was not

44:37

really about AI safety in

44:39

the sense that Sam Altman was behind

44:41

the scenes saying, like, we have to

44:43

go accelerate all these projects while the

44:45

board isn't looking. And that's why he

44:47

got fired. I do not think that

44:50

was what happened. I think the board

44:52

was actually fairly comfortable where things were

44:54

from a safety perspective. I think they

44:56

were just worried about the lying, that

44:58

they say that he was doing. But they have

45:01

not pointed to a single instance of- Perhaps because

45:03

he's such a good liar that you

45:05

can never catch him, but you can sometimes smell

45:07

the sulfurous smell of a lie that

45:09

went undetected and passed by him. They

45:11

do talk about him like a mischievous

45:14

leprechaun or like Rumpelstiltskin

45:16

or something. I

45:18

like having interviewed Sam. I think, no, that's not my

45:20

impression of him. Maybe it's like a Kaiser Soze thing

45:22

where it's like his greatest trick was convincing me that

45:25

he didn't exist. But

45:27

anyways, you were saying that, and this

45:29

fits with my general worldview, which is

45:31

that when institutions explode, it's

45:34

always described as people

45:36

represent one value versus representing another. And

45:38

sometimes that's true. And often it's actually

45:41

about either things that are more subtle

45:43

or just sort of power. And

45:46

you're saying that from your reporting, your sense

45:48

is not that the board was saying, hey,

45:50

you're screening into the apocalypse, we have to

45:52

stop you. The board had some hard

45:55

to define problems with his leadership

45:57

style, and they pulled the big... red

46:00

lever that they're really only supposed to pull if he's

46:02

inventing a Death Star. But what you're

46:04

saying is if you were worried about the AI Death

46:06

Star, you don't necessarily have to

46:08

feel like the AI Death Star

46:10

is coming. That's right.

46:12

That's right. There's no reason to

46:14

believe that now that the old board is out of

46:16

the way, open AI can just go

46:19

absolutely nuts. I don't think that's what is going to

46:21

happen. And also, by the way, there's going to be

46:23

way more scrutiny on open AI as it releases next

46:26

generation models and new features. And

46:28

so I think there's a way in which this was

46:31

very bad for the AI safety community because they were

46:33

made to look like a bunch of Goths who were

46:35

bad at governance. But

46:38

I think it was good in the sense

46:40

that now everyone is talking about AI safety.

46:43

Regulators are very interested in AI

46:45

safety and regulations are being written

46:48

in Europe about AI safety. So

46:50

I actually don't think we have to panic just yet. Got

46:53

it. And then I guess like I've

46:56

began this episode by saying like one way

46:58

that you can think about this as it

47:00

being like a bunch of silly corporate drama.

47:02

And like that is true. And

47:05

at the same time, can I just say I've

47:07

been reading these stories. It's like, oh,

47:09

well, looks like the Silicon Valley tech

47:11

bros have gotten themselves embroiled in a

47:14

little drama. And

47:16

like the only people who can feel that way are

47:18

the people who truly do not care about the future.

47:21

Sorry, you want to convince yourself

47:23

that like there's nothing at stake here

47:25

that like I truly wish my brain

47:27

were as smooth as yours because it

47:29

actually does matter like how people will

47:31

make money in the future. It

47:34

matters if a machine will be able to do

47:36

everyone's job. So count me on the side of

47:38

those who are interested and who do not think

47:40

that this is just like a fun little Netflix

47:42

series for us all. What are you going to

47:45

bet? I'm with you

47:47

and I appreciate you ranting and raving because

47:49

I feel the exact same way. And I'm

47:51

also just like there's this really annoying to

47:53

me thing in technology and it's not just

47:55

civilians. It's like also sometimes journalists who cover

47:58

it where they're like I know. going

48:00

on is the thing that happened last time. So it's

48:02

like, people who are like,

48:04

AI is just NFTs. I'm like, no, those

48:07

are just pieces of technology that

48:09

are described with letters. They're

48:11

very different. Like, the

48:13

future and the present are informed by

48:16

the past, but it's not just a

48:18

movie that you can say you saw

48:20

the end of. Some journalism is just

48:22

people who don't care posturing for other

48:24

people who don't care. And I think

48:26

that is like, we've seen so

48:29

much of that during the open AI story. But

48:31

we're right. And we're smart. Good

48:33

for we're killing it over here. So

48:36

if we agree, and we do that,

48:38

like, whether or not there were shenanigans

48:40

this week, the shenanigans were inspired

48:43

by a real question. And that real

48:45

question matters. AI is likely

48:48

transformative technology. And the

48:50

idea of how it should be governed is

48:52

really tricky. We're focusing on

48:54

open AI because they are the leader in the space. But

48:57

if you zoom out from open AI, there's

48:59

a ton of other companies developing artificial

49:01

intelligence. There's a ton of other countries

49:03

where this is happening, you know, it's

49:05

being developed all over the world. And

49:08

I don't know the right answer. If

49:11

this technology has a potential to be as powerful

49:13

as the people developing it fear, I don't

49:16

know what you do around that. And I'm curious what you

49:18

think like if you were king of

49:20

the world, but you were leaving next year,

49:22

and you had to set up a regime

49:24

for artificial intelligence that everyone would actually follow.

49:26

What do you do? Well,

49:29

one, I do think this is a place

49:31

where we want the government to play a

49:33

role, right? Like if a technology is created

49:36

that does have the effect of causing massive

49:39

job losses and introduces novel

49:41

new risks into like, you

49:44

know, bioweapons and cybersecurity

49:46

and all sorts of other

49:48

things. I think you do want the

49:50

government paying attention to that. I think

49:52

that there's a good case to be made that

49:54

the government should like be funding its own large

49:56

language, but it should be doing its own fundamental

49:59

research. into how these models

50:01

work and maybe how to build some of

50:03

its own safely because I'm not sure that

50:05

the the for-profit model is the

50:07

one that is going to deliver us to the

50:09

best result here. In terms of what

50:11

would government oversight look like,

50:14

some folks I talked to you talk about it

50:16

just in terms of capabilities. Like we should identify

50:18

capabilities that's like once a model is able

50:21

to do this then we

50:23

would introduce some breaks on how it

50:25

is distributed, how it is released

50:27

into the world. Maybe there are some safety tests

50:29

we make you go through and in a world

50:32

where the government can and does regulate this, which

50:35

government? Is it the US? Is it the

50:37

UN? Like how do you do it? It

50:39

generally winds up being a mix of Western

50:41

democracies that lay the blueprint. You know

50:44

the US doesn't typically regulate technology very

50:46

much but Europe does and so

50:49

Europe essentially writes the rules for the internet

50:51

that the rest of us live on and

50:53

it basically works out okay because their values

50:55

are basically aligned with American values and so

50:57

like yes we have to click on a

50:59

little cookie pop-up every website that we visit

51:01

because Europe is making us and we hate

51:03

it but it's also fine you know. Yeah

51:05

and so like AI is probably going to

51:07

be the same thing where Europe is going

51:10

to say well AI should basically be like

51:12

this and the US will have hearings

51:14

where they sort of gesture in similar directions and

51:16

then never pass the law and like that will

51:18

be the medium-term future of AI. Where

51:20

I think it will change is if there is

51:22

some sort of incident where like thousands of people

51:24

die and AI plays a direct result like that

51:27

is when the US will finally get around to

51:29

doing something. Maybe. Maybe.

51:32

It's weird it's

51:34

weird to feel both scared and excited like I'm

51:36

not used to having two feelings at the same

51:38

time. There's this feeling

51:40

that I just call AI vertigo which I mean

51:42

and this is the sort of staring into the

51:45

abyss feeling where you can imagine

51:47

all of the good that could come with

51:49

you know having a universal translator and essentially

51:51

omniscient assistant that is just living in every

51:54

device that you have like that's incredibly powerful

51:56

and good but like yes

51:58

it will also generate both like

52:00

a huge number of new harms and like

52:02

at a huge volume. And

52:05

so your imagination can just run wild and I

52:07

think it's important to let your imagination run wild

52:09

a little bit and it is also possible to

52:11

go too far in that direction and sometimes you

52:13

just need to like you know chill out and

52:15

go play Marvel stuff a little bit. Casey,

52:18

that's exactly what I'm gonna do. Okay, that's like good.

52:24

Thank you. Thank you. Casey

52:29

Newton. You should subscribe to his

52:31

excellent newsletter platformer and to his

52:33

podcast Hard Fork which he co-hosts

52:35

Kevin Roos. They've had some wonderful

52:37

episodes on the subject. You should go check them out.

52:45

Also just in general this blow up at

52:47

OpenAI has been an occasion for some wonderful

52:49

tech reporting. People have been all over this

52:52

story explaining a very complicated situation

52:54

very quickly. I'm going to put links to

52:56

some of the pieces that I enjoyed and

52:58

drew from for this story. You can find

53:00

them as always at the newsletter for this

53:02

show. There's a link to that

53:04

newsletter in the show. Search

54:19

Engine is a presentation of Odyssey and

54:22

Jigsaw Productions. It was created by

54:24

me, PJ Vogt, and Shruti Pena Menaini, and

54:26

is produced by Garrett Graham and Noah John.

54:29

Theme, original composition, and mixing by

54:31

Armin Bazarian. Our executive

54:34

producers are Jenna Weiss-Berman and Leah Reese

54:36

Dennis. Thanks to the team

54:38

at Jigsaw, Alex Gibney, Rich Pirello, and John

54:40

Schmidt. And to the

54:42

team at Odyssey, JD Crowley, Rob

54:45

Mirandi, Craig Cox, Eric Donnelly, Matt

54:47

Casey, Laura Curran, Josephina Francis, Kurt

54:49

Courtney, and Hilary Scheff. Our

54:51

agent is Orin Rosenbaum at UTA. Our

54:54

social media is by the team at Public Opinion

54:56

NYC. Follow and

54:58

listen to Search Engine with PJ Vogt now for

55:01

free on the Odyssey app or wherever you get

55:03

your podcasts. Also, if

55:05

you would like to become a paid subscriber, you

55:07

can head to pjvogt.com. There's a link in the

55:09

show notes. Or another way

55:12

to help the show is to go to Apple

55:14

Podcasts and rate and review us. Highly

55:16

would be nice. All right, that's it for

55:18

this week. Thank you for listening. We'll see you next week.