The A.I. Trade Secrets War + Economists Say ‘We Must Act Now’ + HatGPT

Introducing The Problem Solvers, a new series from Anthropic about the startup founders who are going further with Claude, like Max Unstrand, co-founder and CEO at Lagora, a collaborative AI workspace for lawyers.

Our system gets better with every new model release, and we certainly saw this when we started embedding more and more of the Anthropic models into Lagora.

This, uh, exchange of ideas, what's working, what's not working, where do we need to push further, that has been incredible.

Watch the series at claude.com/problemsolvers.

Casey NewtonHost0:32

Well, Casey, I had a very confusing experience this week-

Kevin RooseHost0:35

Yeah?

Casey NewtonHost0:35

... because I ordered a new ice cream scoop.

Kevin RooseHost0:38

Hmm.

Casey NewtonHost0:39

Uh, this ice cream scoop is the, uh, the, the one I have used for many years, but one of the drawbacks about it is if you put it in the dishwasher it gets totally ruined.

Kevin RooseHost0:47

Okay.

Casey NewtonHost0:48

Um, so put it in the dishwasher-

Kevin RooseHost0:49

Which I have to say, major design flaw for an ice cream scoop-

Casey NewtonHost0:51

[laughs]

Kevin RooseHost0:51

... a device that is going to want to be washed at regular intervals.

Casey NewtonHost0:55

Yes, you have to hand wash this thing.

Casey NewtonHost0:56

So I get a new one, it comes in a box, and there's a little sticker on the ice cream scoop that says, "This product contains AI."

Kevin RooseHost1:05

[laughs] Oh, no.

Casey NewtonHost1:06

And I'm thinking-

Kevin RooseHost1:07

Not artificial ice cream.

Casey NewtonHost1:09

[laughs] I'm thinking, "What?" I have been using this same ice cream scoop-

Kevin RooseHost1:14

Yeah

Casey NewtonHost1:14

... for, like, for like 10 years, and all of a sudden it is upgraded to AI? Like, what is the AI? I did not order this.

Casey NewtonHost1:20

Like, is it connecting to, you know, to, to ChatGPT to, like, tell it what, what flavor-

Kevin RooseHost1:25

Yeah

Casey NewtonHost1:25

... ice cream I'm doing? What is going on here?

Kevin RooseHost1:27

I really am bracing myself 'cause the, the number of things in my life that used to be very simple and now require something called a firmware update is out of control.

Casey NewtonHost1:34

[laughs] Yeah.

Kevin RooseHost1:34

And so if I, if my ice cream scoop needed a firmware update, I'd be very upset.

Casey NewtonHost1:37

It also, like, doesn't ship with a charger or anything.

Kevin RooseHost1:40

Right.

Casey NewtonHost1:40

It doesn't have any ports on it, so I'm like, "How the hell does this ice cream scoop have AI?"

Kevin RooseHost1:44

Yeah.

Casey NewtonHost1:45

And I stare at this just, like, bewildered for like five minutes, and then I realize it's talking about aluminum.

Kevin RooseHost1:52

[laughs]

Casey NewtonHost1:52

It's using the chemical symbol for aluminum.

Casey NewtonHost1:56

This product contains Al.

Kevin RooseHost2:00

[laughs]

Casey NewtonHost2:00

And then I felt stupid.

Casey NewtonHost2:01

[upbeat music]

Kevin RooseHost2:07

I'm Kevin Roose, a tech columnist at The New York Times.

Casey NewtonHost2:09

I'm Casey Newton from Platformer.

Kevin RooseHost2:10

And this is Hard Fork.

Casey NewtonHost2:11

This week, OpenAI has a great new model and a big new legal problem.

Casey NewtonHost2:16

Then, Stanford economist Erik Brynjolfsson is here to discuss a new warning from hundreds of economists and researchers about the growing risk of AI job loss.

Casey NewtonHost2:24

And finally, some HatGPT starring Lorde.

Casey NewtonHost2:28

You can remember the green light.

Kevin RooseHost2:29

Is that a Lorde joke?

Casey NewtonHost2:30

Kevin, would you listen to one K-pop song?

Kevin RooseHost2:32

[laughs] [upbeat music] Well, Casey, lot is happening in AI this week, specifically, uh, i- feel like it's been a very big OpenAI news week.

Casey NewtonHost2:48

Yeah, and I feel like it's actually been a minute since we checked in on them, and so I was glad that there was, like, such a series of events that I felt like, okay, yeah, we can actually return to this company now.

Kevin RooseHost2:58

So before we do this, of course, we should make our AI disclosures.

Kevin RooseHost3:00

I work for The New York Times, which is suing OpenAI, along with Microsoft and Perplexity.

Casey NewtonHost3:05

And my fiance works at Anthropic.

Kevin RooseHost3:07

So let's just get a few things ticked off here.

Kevin RooseHost3:10

The first, uh, thing was a personnel shift.

Kevin RooseHost3:12

Uh, Fiji Simo, the former CEO of applications, I believe, was her title at OpenAI.

Casey NewtonHost3:17

That was her initial title.

Casey NewtonHost3:18

It was later changed to CEO of AGI Deployment.

Kevin RooseHost3:22

Okay, so she's leaving the company.

Kevin RooseHost3:24

She is, was on an extended medical leave due to some health problems she had had.

Kevin RooseHost3:28

Um, she is now officially sort of stepping back from that role, and her roles will be divided up among other executives.

Kevin RooseHost3:34

Casey, what did you make of this?

Casey NewtonHost3:35

I mean, very sad story.

Casey NewtonHost3:36

Obviously anytime somebody has to leave a job for, like, reasons other than they wanted to, uh, it is sad.

Casey NewtonHost3:42

Uh, Simo was there for less than a year, and I think was able to accomplish some of what she wanted and did play a role in pushing the company to streamline some of its offerings and cut back on some of the projects that, in my view, were distractions, such as the Sora, you know, AI, uh, video generator.

Casey NewtonHost4:01

Um, but I think she had much bigger ambitions there that she didn't get to realize.

Casey NewtonHost4:05

And, you know, in, in terms of OpenAI going forward, it is yet more executive turmoil at a company that has had a lot of that, uh, over the past year.

Kevin RooseHost4:14

And speaking of turmoil, we should also talk about this lawsuit that Apple filed last week against OpenAI alleging that they stole some trade secrets.

Kevin RooseHost4:22

Casey, tell us the story of this lawsuit.

Casey NewtonHost4:24

This came out on Friday.

Casey NewtonHost4:27

Uh, Apple filed a lawsuit against OpenAI and IO, which was the, uh, sort of Jony Ive co-founded company that they, uh, bought for more than $6 billion.

Casey NewtonHost4:38

The suit also names two, uh, current OpenAI employees, Qiang Lu and Tang Tan, who incidentally are former Apple employees who now work at OpenAI.

Casey NewtonHost4:48

Um, they write in this filing, quote, "At every level, from members of its technical staff to its chief hardware officer," that's Tang Tan, "and in coordination with business partners, OpenAI has been stealing Apple's trade secrets and confidential information.

Casey NewtonHost5:02

As a natural result, OpenAI's nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets." So that was the kind of, uh, you know, uh, quote that you saw in a lot of stories about this particular case, but the details are really pretty wild.

Kevin RooseHost5:22

Yeah, let's talk about some of those details.

Kevin RooseHost5:23

It was, it was a long filing.

Kevin RooseHost5:24

There was a lot of juicy stuff in there.

Kevin RooseHost5:26

Um, the thing that I think got the most attention was this allegation by Apple that, uh, Tang Tan, this OpenAI chief hardware officer, had directed job candidates who were applying for jobs at OpenAI to bring physical hardware components, blueprints, and prototypes into job interviews for what the filing calls show and tell sessions.

Casey NewtonHost5:47

And, like, you think it's hard enough to get a job in this economy- And then your recruiter says, "Hey, um, when you come in, make sure to bring, um, two forms of ID, um, and then your resume, and also an unreleased Apple product-

Kevin RooseHost6:02

[laughs]

Casey NewtonHost6:02

... that you sort of secret out of the lab." Like, what? That's, that's such a, a big ask.

Kevin RooseHost6:08

Yeah.

Casey NewtonHost6:09

You know?

Kevin RooseHost6:09

How are you supposed to deal to get a job in this climate?

Casey NewtonHost6:11

[laughs] I mean, my goodness.

Kevin RooseHost6:14

So then there's also these allegations about Chang Liu, the former senior system electrical engineer at Apple.

Kevin RooseHost6:20

Uh, Apple says in its filing that he kept a work laptop after leaving the company.

Kevin RooseHost6:25

Um, he then apparently exploited an unknown security vulnerability to access confidential information, and bragged about this to a colleague.

Kevin RooseHost6:34

Uh, I'll just read his apparent quote-

Casey NewtonHost6:36

Please

Kevin RooseHost6:36

... from the filing.

Kevin RooseHost6:38

"Lol, I found out I can access the network storage.

Kevin RooseHost6:42

So funny."

Casey NewtonHost6:43

You know, and I just...

Casey NewtonHost6:43

And I'm glad you, you read that in full, 'cause here's what I want people to know.

Casey NewtonHost6:46

If you say "lol" in a message, you should not add, "So funny."

Kevin RooseHost6:50

[laughs]

Casey NewtonHost6:50

Right? Because b- those two phrases are doing the same thing.

Kevin RooseHost6:53

Yes.

Kevin RooseHost6:54

Omit needless words.

Casey NewtonHost6:55

Yeah.

Kevin RooseHost6:56

So there's a lot more allegations in the filing.

Kevin RooseHost6:59

Uh, one of the things it says is that OpenAI allegedly approached one of Apple's manufacturing partners and tricked them into showing them some sort of secret metal treatment finishing process that Apple uses.

Casey NewtonHost7:10

Yeah.

Casey NewtonHost7:11

If you've ever picked up an iPhone, you know that it has a very beautiful metal finish, and the allegation here is that OpenAI was, like, wanting to know how that happened.

Casey NewtonHost7:18

And so they just went to the supplier, and they were like, "Just show us what you do for Apple.

Casey NewtonHost7:22

It's cool.

Casey NewtonHost7:23

They told us it's cool." [laughs]

Kevin RooseHost7:24

Yeah, Tim Cook said-

Casey NewtonHost7:26

Yeah

Kevin RooseHost7:26

... said, "We have permission to, uh, to do this."

Casey NewtonHost7:27

Yeah, Tim Cook said, "I would love for Sam Altman to know how to do this."

Kevin RooseHost7:30

[laughs]

Casey NewtonHost7:30

"Just go ahead and show them," and then they show...

Casey NewtonHost7:34

The- these are the allegations.

Kevin RooseHost7:35

Yes.

Casey NewtonHost7:36

Yeah.

Kevin RooseHost7:36

Um, but Casey, what did you make of this filing as you looked through all of the juicy details?

Casey NewtonHost7:41

So I'm of two minds here, because it is true that Apple in general treats everything that has ever happened in the city of Cupertino as an Apple trade secret.

Kevin RooseHost7:52

Yes.

Casey NewtonHost7:53

Right? Uh, they defend their, uh, intellectual property extremely vigorously.

Casey NewtonHost7:58

And so I would not be surprised if as this case unfolds, um, you know, a judge or a jury might push back on some of the claims here and says, "You know what? It's actually not illegal for Tang Tan to mention the code name of an unreleased Apple product to a prospective candidate," right? Okay.

Casey NewtonHost8:18

That said, with some of the particular allegations against the lol-so-funny person, um, the allegation that they were, like, coaching job candidates to bring unreleased products into job interviews, that does just seem transparently wrong to me.

Casey NewtonHost8:37

And so my real question as this case unfolds now, Kevin, is basically how high up the chain did this go? You know, some bloggers have noted that the case goes out of its way not to mention Jony Ive, who's, like, a foundational figure at this company.

Casey NewtonHost8:55

I- if there was actually some sort of, you know, scheme to do this, it seems extremely unlikely that Jony Ive would be, uh, unaware of it, and yet he's not in the filing.

Casey NewtonHost9:04

Neither is Sam Altman, right? So these are some of the questions that I have.

Casey NewtonHost9:09

Um, but I do think, you know, i- if Apple is successful here, it could have some, you know, uh, very tough consequences for OpenAI.

Casey NewtonHost9:15

What would you make of it?

Kevin RooseHost9:16

Yeah, I mean, I think these guys ju- knowing nothing else about this case other than the filing, I'm sure there will be, you know, counter filings and, and explanations from OpenAI about why this all wasn't as bad as it looks on paper, but, like, to me, these guys seem guilty as hell.

Kevin RooseHost9:30

That said, it is not clear that OpenAI is actually building anything that makes use of Apple's trade secrets.

Kevin RooseHost9:38

Um, we got a report this week from Mark Gurman over at Bloomberg, who is sort of the closest and most plugged in Apple reporter, who reported on OpenAI's first hardware product, the thing that Jony Ive has been working on with them, being some sort of a mobile screen-free home smart speaker.

Kevin RooseHost9:56

That, uh, doesn't sound to me like anything that Apple has built or maybe is planning on building.

Casey NewtonHost10:01

Well, I mean, they have that HomePod, right? You know, like, what Gurman described, uh, as I read it, was basically a next generation, like, Amazon Echo, um, which many companies now have their own versions of, including Apple with the HomePod.

Casey NewtonHost10:16

And so a screen-free speaker that connects to AI, like, Apple does have one of those, and I would not be surprised-

Kevin RooseHost10:22

[laughs]

Casey NewtonHost10:22

...

Casey NewtonHost10:22

if they said that whatever OpenAI is building, uh, somehow builds off of their efforts here.

Kevin RooseHost10:27

Yeah, I mean, I am not, uh, particularly bullish on the OpenAI hardware push in general, uh, but I think this made me even more skeptical that they are going to come out with some category-breaking, uh, you know, definitive technology here.

Kevin RooseHost10:42

Um, I could be wrong, of course.

Kevin RooseHost10:43

But the, you know, asking employees, job candidates, to bring in prototypes from their Apple secret work does not seem to me like a company that, that has a real vision for what it is doing on its own.

Casey NewtonHost10:54

You know what I thought was funny about this was that a little over a year ago, at the first Hard Fork live, uh, I asked Sam Altman about their hardware project, and I said, "Is it just Alexa?" And he said, "It's not just Alexa.

Casey NewtonHost11:07

Like, you know, give us some ti- Like, we're thinking we're gonna do something better than Alexa." And I read the Gurman report this week, and I thought, "Sounds a lot like Alexa."

Kevin RooseHost11:14

[laughs]

Casey NewtonHost11:14

So, you know, we'll see.

Casey NewtonHost11:15

Like, I mean, the AI in it is gonna be better, you know, but is it, is it really more than that? Anyways.

Kevin RooseHost11:20

Yeah, I mean, I think this is very much in keeping with Apple's sort of historical posture for secrecy.

Kevin RooseHost11:26

You know, you remember when the, when the iPhone, I think it was the iPhone 4, was sort of leaked- There were law enforcement investigations.

Kevin RooseHost11:33

Like this, this company takes its trade secrets very seriously.

Kevin RooseHost11:37

At the same time, Apple is also struggling in AI.

Kevin RooseHost11:40

I think it's fair to say that, and so this to me feels very much like the kind of lawsuit that someone who is not in the lead or close to the lead would file.

Casey NewtonHost11:49

Yeah, I think that that is, uh, reasonable.

Casey NewtonHost11:51

You know, in addition to the, you know, alleged, uh, [laughs] inspection of unreleased devices, over the past couple of years, uh, OpenAI has hired more than 400 people away from Apple.

Casey NewtonHost12:04

So even at a big company like Apple, that is the kind of recruiting effort that you notice, and I'm sure that has been extremely frustrating to the folks over at Apple as they try to build their own next generation devices.

Casey NewtonHost12:15

So I think that there, uh, was clearly a wound there and, and I think it is fair to read this lawsuit in part as, uh, Apple lashing back.

Kevin RooseHost12:25

Yeah.

Kevin RooseHost12:25

I mean, this is a bigger conversation than we, than we have time for today, but I'm always just fascinated by how leaky and sort of porous the boundaries between these AI companies are.

Kevin RooseHost12:36

I mean, one of the sort of hidden factors in this AI race is I think the fact that in California you can't really enforce a non-compete.

Kevin RooseHost12:43

So people are always moving jobs from OpenAI to Anthropic, from Google DeepMind to OpenAI.

Kevin RooseHost12:49

Like people are just bouncing around all of the time, and they're taking their secrets with them, right? Like you can't take intellectual property.

Kevin RooseHost12:56

You can't like literally take the code for your model, but all this stuff like exists in their head.

Kevin RooseHost13:01

And so part of the reason that you would hire someone from Apple if you're OpenAI is like they have this stuff in their head, and it's just a matter of like, oh, they can't literally bring the designs or the prototype or the IP over.

Kevin RooseHost13:14

That's illegal.

Kevin RooseHost13:15

But just knowing how to build something is not a protected class of IP.

Casey NewtonHost13:19

Right, and I'm sure that OpenAI will make that argument as this case moves forward.

Casey NewtonHost13:24

Kevin, what has OpenAI said in the meantime, uh, about these allegations?

Kevin RooseHost13:28

So a spokesperson for OpenAI posted on X, quote, "We have no interest in other companies' trade secrets.

Kevin RooseHost13:35

We remain focused on building innovative technology that empowers people everywhere."

Casey NewtonHost13:39

Hmm.

Kevin RooseHost13:39

How did that strike you?

Casey NewtonHost13:40

That struck me as what we call a non-denial denial.

Kevin RooseHost13:43

[laughs]

Casey NewtonHost13:44

It struck me as we're just learning about this, and we need to put out a statement, but we can't say anything 'cause we don't know anything.

Casey NewtonHost13:51

So what is a series of words that journalists would reprint that would sort of buy us a few more days?

Kevin RooseHost13:57

Yes.

Casey NewtonHost13:57

That's kind of how that struck me.

Kevin RooseHost13:58

There's an art form to this stuff.

Casey NewtonHost13:59

Mm-hmm.

Kevin RooseHost14:00

Okay.

Kevin RooseHost14:00

Another thing that is happening at OpenAI this week is that people are starting to play around with Sol, this new model, GPT 5.6, um, that was not available for a period of time under request from the White House and the federal government, uh, but that people are now using and experimenting with.

Kevin RooseHost14:19

Um, I have not spent much time with Sol so far, but people I know who I- I trust on this stuff have said it's really good.

Casey NewtonHost14:27

Yeah.

Kevin RooseHost14:27

It's a very capable model.

Casey NewtonHost14:28

Yeah, it absolutely is.

Casey NewtonHost14:30

It is a notable step up from 5.5. I've been doing some vibe coding projects recently.

Casey NewtonHost14:36

I've had, uh, the GPT 5.6 Sol kind of critique them, suggest changes, and I've been really pleased with the feedback.

Casey NewtonHost14:46

So I think if you are a ChatGPT user and you're on one of the paid plans, you're gonna be really excited that you have GPT Sol.

Kevin RooseHost14:52

Is this a thing you do to like, uh, you know, get back at your fiance when you're having arguments at home?

Casey NewtonHost14:57

[laughs]

Kevin RooseHost14:58

You just like casually stroll in and you're like, "Hey, um, been doing some really cool vibe coding with GPT 5.6"?

Casey NewtonHost15:05

[laughs] Yes.

Casey NewtonHost15:06

I mean, I think he understands that, you know, it's in the nature of my job to always be testing all of the models.

Kevin RooseHost15:12

Yeah, you're not a monogamist when it comes to-

Casey NewtonHost15:16

[laughs]

Kevin RooseHost15:16

...

Kevin RooseHost15:16

AI models.

Casey NewtonHost15:16

No.

Kevin RooseHost15:16

[laughs]

Casey NewtonHost15:16

I'm extremely poly and, um, I think he appreciates that.

Kevin RooseHost15:20

Yeah.

Kevin RooseHost15:20

So a couple things that have come out of this new model release.

Kevin RooseHost15:23

One is that it has really changed the, uh, plans of Anthropic.

Kevin RooseHost15:28

Uh, as we talked about on the show, they had announced that, you know, starting on some arbitrary day in early July, they were going to revoke access to Fable for people using their Claude subscription plans.

Kevin RooseHost15:42

Um, this is their most powerful model.

Kevin RooseHost15:44

Basically, you'd still be able to use it through the API, but you're gonna have to pay for it on a per token basis.

Kevin RooseHost15:50

You're not just gonna like get some Fable access, uh, included in your bundle of a subscription.

Kevin RooseHost15:56

And they then extended that deadline by a few days, um, to give people a few more days to play with Fable.

Kevin RooseHost16:05

Then they have extended it again.

Kevin RooseHost16:07

Um, and my read on this is like this is in direct response to the fact that OpenAI has a, by all accounts, very good model out right now, and they are afraid that if they cut off access to Fable, people are gonna switch to using ChatGPT or Codex.

Casey NewtonHost16:21

Absolutely.

Casey NewtonHost16:21

So I mean, you think about what their original plan was, which is, you know, we'll put out Fable, and th- this happened, you know, several weeks before 5.6 eventually came out.

Casey NewtonHost16:30

And I assume that their thinking was people will have a few weeks to see how good Fable is.

Casey NewtonHost16:35

And so then whenever some other company launches a new model, it won't matter because people are already gonna be addicted to Fable.

Casey NewtonHost16:42

Instead, they had a big fight with the US government, and by the time they finally got Fable out the door, there were essentially only a few days between that and 5.6 arriving.

Casey NewtonHost16:50

And so my belief now- ... is that they are trying to, uh, essentially get people hooked on Fable, um, and, and eventually push them toward paying per token.

Casey NewtonHost16:58

Um, but you know, 5.6 is out now, and people are gonna be able to maybe try that for a lot cheaper, so.

Kevin RooseHost17:03

Right.

Kevin RooseHost17:03

And the companies, we should say, have been getting a little catty.

Kevin RooseHost17:07

Um-

Casey NewtonHost17:07

A little?

Kevin RooseHost17:08

[laughs]

Casey NewtonHost17:09

I've seen episodes of Drag Race that were more polite than what's been happening on Twitter over the past few days, Kevin.

Kevin RooseHost17:14

Yeah, so let's just talk about that a little bit because I think it's interesting, but also I think potentially very important.

Kevin RooseHost17:20

Um, so Anthropic posted a new ad, uh, for Claude that is sort of a, a longer ad.

Kevin RooseHost17:27

It's kind of a, a, a sort of montage.

Kevin RooseHost17:29

It's talking about hard questions.

Kevin RooseHost17:32

It starts with this very bizarre sequence of images, of one is like a, a building on fire, um, and one is with a, a cemetery, like a row of headstones.

Kevin RooseHost17:43

It looks to be, like, a military cemetery.

Casey NewtonHost17:45

It's Arlington National Cemetery, where my grandfather is buried, Kevin.

Kevin RooseHost17:49

[laughs]

Casey NewtonHost17:49

What is it doing in this ad?

Kevin RooseHost17:51

And so the, the sort of voiceover on the ad is talking about how, how can it, you know, h- who's gonna, who's gonna hit the brakes if we need to slow down this technology? Can AI be trusted? Like, these are the kinds of hard questions.

Kevin RooseHost18:02

You know, the, the ad eventually gets to a more hopeful and positive message, but it starts with this kind of, like, very doomy picture of the world on fire literally as a result of this technology, and Sam Altman did not like this ad.

Kevin RooseHost18:17

Many other people did not like this ad.

Kevin RooseHost18:19

Um-

Casey NewtonHost18:19

I mean, I'll say, I think this ad is a disaster.

Kevin RooseHost18:21

Yeah.

Kevin RooseHost18:21

[laughs]

Casey NewtonHost18:21

Like, I really do.

Casey NewtonHost18:22

Like, it's, it's an extremely confusing set of messages, and I think it's v- like, it is unlikely to me that the average person would take away a, a positive message about Anthropic or AI from this ad.

Kevin RooseHost18:37

Oh, I don't know.

Kevin RooseHost18:38

I, I wanna sort of, like, meekly defend the ad by saying that I think it's good if AI companies are honest about the fears that they have.

Kevin RooseHost18:47

I think it would be very bad if they just all turned into, like, sunny, optimistic marketing companies.

Casey NewtonHost18:52

Well, sure, but is the right format for that, like, brand marketing? You know, like, I understand you think this might go really, really badly and you're trying to prepare the world for that, but I think, like, a serious video with your leaders talking about that or, like, a thoughtful blog post are a better avenue for that than, like, video posted on X of burning building.

Casey NewtonHost19:12

You know what I mean?

Kevin RooseHost19:13

Yeah, no, I, I take that point.

Casey NewtonHost19:14

Yeah.

Kevin RooseHost19:14

I just, like, I don't wanna have the message that comes out of this be like, "AI companies should just lie to us, and if they're scared about AI, they should just not tell us that, and they should just put out ads about how great this stuff is."

Casey NewtonHost19:25

Well, h- here to me is I'm not gonna call it a lie, but here's what I find confusing.

Casey NewtonHost19:29

So at the end of this ad, wh- which just es- essentially I believe has, like, uh, sound clips of average people asking their questions about AI or d- discussing their wishes for AI, at the very end of it, uh, Anthropic shows its tagline, which is "Keep thinking." And when Anthropic tells me, "Keep thinking," I feel like, um, I'm reading one of those blog posts where they tell senior citizens to keep doing crossword puzzles to keep-

Kevin RooseHost19:54

[laughs]

Casey NewtonHost19:54

... their mind active.

Casey NewtonHost19:55

You know what I mean? It's like-

Kevin RooseHost19:57

Mm-hmm

Casey NewtonHost19:57

...

Casey NewtonHost19:57

if you're building super intelligence that is going to do all of the thinking, what exactly do you want me to do the thinking for?

Kevin RooseHost20:03

Right.

Casey NewtonHost20:03

Do you know what I mean?

Kevin RooseHost20:04

Right.

Casey NewtonHost20:04

So I just think the entire brand message is getting very, very confused, and it is making people feel [laughs] bad, and I think Anthropic needs to think about that.

Kevin RooseHost20:13

Yeah, I think that's right.

Kevin RooseHost20:15

Um, but let's get back to the OpenAI of it-

Casey NewtonHost20:17

Yeah

Kevin RooseHost20:17

... because that was actually where this started to heat up.

Kevin RooseHost20:19

So Sam Altman posted on Monday a, he reposts the ad, and he says, "I thought this was satire.

Kevin RooseHost20:26

Kept looking for the handle to be spelled C-1audAI," like basically a sort of scam account or something.

Kevin RooseHost20:33

And then he sort of mocks the hard questions tagline of the ad.

Kevin RooseHost20:36

He says, "Hard questions are great, but only if we deem you worthy enough not to silently downgrade you or even get access at all," of course, referring to Anthropic's on-again, off-again, uh, Fable rollout and some decisions they had made about sort of, like, silently downgrading people to a lesser model if it detected they were working on certain types of things.

Casey NewtonHost20:55

Yeah.

Casey NewtonHost20:55

So here is my impression of what is happening, is the folks at OpenAI, uh, understand that they are now seen as an underdog, and they're thrilled about it [laughs].

Casey NewtonHost21:05

Like, think about the past year or two that they have had where they were the sort of carrying the banner of American AI, and every misstep that they made was under a microscope.

Casey NewtonHost21:18

And now fast-forward to today, Anthropic is seen as being in the lead.

Casey NewtonHost21:22

And so it's now all of Anthropic's mistakes that are sort of getting magnified, and that lets OpenAI come in and throw some blows and look like a scrappy challenger again.

Casey NewtonHost21:31

And it's just much more fun to play the scrappy challenger than the sort of, you know, solemn, uh, standard-bearer of American AI.

Kevin RooseHost21:38

Yeah, I, I buy that.

Kevin RooseHost21:39

I don't think they're thrilled about being in second place, at least as the public perceives it.

Kevin RooseHost21:43

I think there's a lot of paranoia and fear over there, that they are, you know, losing talented researchers to Anthropic, that they are sort of losing what they call the mandate of heaven, um, to Anthropic by not having the absolute best models on the market.

Casey NewtonHost21:58

Wait, what is the mandate of heaven in this context?

Kevin RooseHost22:00

Oh my God, you haven't seen this meme?

Casey NewtonHost22:01

No.

Kevin RooseHost22:02

So this is a thing that, like, insane people on social media post about, where they're like, "Oh, you know, uh, the new Gemini is really good.

Kevin RooseHost22:11

Maybe, maybe Google has the mandate of heaven."

Casey NewtonHost22:13

[laughs]

Kevin RooseHost22:13

It's just sort of like [laughs]...

Casey NewtonHost22:14

Like sort of like God has chosen you to invent AGI?

Kevin RooseHost22:17

Yes.

Casey NewtonHost22:17

Okay.

Casey NewtonHost22:18

Wonderful.

Kevin RooseHost22:18

Uh, it's, it's very dumb.

Kevin RooseHost22:19

Anyone who says the mandate of heaven in, in sincerity should be checked into a mental hospital.

Casey NewtonHost22:24

[laughs]

Kevin RooseHost22:24

Um, but anyway, I think there's a lot of paranoia.

Kevin RooseHost22:27

They really, like, these companies, like I- I don't know if the public understands this, but these companies hate each other.

Casey NewtonHost22:34

Yeah.

Kevin RooseHost22:34

This is more than a Coke and Pepsi thing.

Casey NewtonHost22:36

Yeah.

Kevin RooseHost22:36

Like, what's funny is I was, I was researching this for my book, and I learned that technically these companies may still have a non-disparagement agreement with each other.

Kevin RooseHost22:45

[laughs]

Casey NewtonHost22:45

[laughs]

Kevin RooseHost22:47

So when, uh, Dario and, and the Anthropic folks-

Casey NewtonHost22:49

Yeah

Kevin RooseHost22:49

... were leaving OpenAI in late 2020, they sort of signed this non-disparagement agreement that says, like, "We won't badmouth you, and you won't badmouth us, and, like, we'll all sort of keep things, um, you know, placid on the surface." Um, and that has never been rescinded to my knowledge.

Casey NewtonHost23:05

Yeah.

Kevin RooseHost23:05

And so technically they still have a document on the books that says we're not supposed to trash each other in public.

Casey NewtonHost23:10

From the outside, reading sort of how they talk about each other, I would assume that they had a disparagement agreement.

Kevin RooseHost23:15

Yes [laughs].

Casey NewtonHost23:15

Which-

Kevin RooseHost23:15

An o- only disparagement agreement

Casey NewtonHost23:17

Yeah.

Casey NewtonHost23:17

The minute we leave here, it's on, bro.

Kevin RooseHost23:19

[laughs]

Casey NewtonHost23:19

It's on like Donkey Kong.

Kevin RooseHost23:21

[laughs] Yes, so these companies absolutely hate each other.

Kevin RooseHost23:24

Um, they are bickering over the, the sort of terms of these models being accessible to subscribers.

Kevin RooseHost23:32

I think they are going to start sort of doing some kind of price war with each other.

Kevin RooseHost23:36

That's my expectation, is like, you know, Anthropic will kind of reduce the price of tokens, and then OpenAI will reduce it below that.

Kevin RooseHost23:42

And it'll sort of be, I think, good for consumers in a very sort of narrow, you know, consumer interest perspective because competition among frontier AI labs is good, I would argue.

Casey NewtonHost23:54

Uh, yeah, it's, it's definitely good for consumers.

Casey NewtonHost23:56

I think that there are elements of the, you know, AI race itself that are, uh, scary, but perhaps that's a different conversation.

Kevin RooseHost24:01

No, I think that's, that's where I wanna go with this-

Casey NewtonHost24:03

Yeah

Kevin RooseHost24:03

... because I think, like, this feud between OpenAI and Anthropic, which does not seem to be subsiding and in fact seems to be getting worse, I think it is very dangerous.

Kevin RooseHost24:12

I do not like this way that these labs are conducting, uh, their relations with each other.

Kevin RooseHost24:18

I do not like how much the people at OpenAI hate Anthropic and are motivated by wanting to crush them.

Kevin RooseHost24:24

I do not like how much the people at Anthropic hate OpenAI and wanna crush them.

Kevin RooseHost24:28

I think that we are heading into a period where there may be real and dangerous risks associated with these very powerful systems, and these companies should be open to collaboration-

Casey NewtonHost24:39

Yes

Kevin RooseHost24:39

... and solving this problem together.

Kevin RooseHost24:41

And coordination between labs I think is going to become increasingly important, and you can't do that if you're constantly fighting each other.

Casey NewtonHost24:47

I completely agree.

Casey NewtonHost24:48

You know, uh, Demis Hassabis, who runs DeepMind, put out a piece this week where he talked about, um, an idea for a kind of government regulator that would, um, you know, cr- create a process around when a frontier lab could release a new model, maybe solve some of the issues that we've seen in recent months.

Casey NewtonHost25:03

And one of the things that he talks about is it would be great if the labs at some point could successfully collaborate, uh, a- around a slowdown in the development of AI.

Casey NewtonHost25:14

And now maybe that's something that the government will just, you know, decree.

Casey NewtonHost25:17

But in the absence of that, uh, you want to see exactly what you're talking about, Kevin, which is, like, the leaders of these companies being able to pick up the phone and call each other and think about, like, what is good for the rest of the world and not just what is gonna get a bunch of retweets.

Kevin RooseHost25:29

Yeah.

Kevin RooseHost25:29

I just desperately need Dario and Sam and Demis to, like, all get together, take a little MDMA-

Casey NewtonHost25:38

[laughs]

Kevin RooseHost25:39

... work it out-

Casey NewtonHost25:40

Mm-hmm

Kevin RooseHost25:40

... find some compassion-

Casey NewtonHost25:42

Yeah

Kevin RooseHost25:42

... find some common ground.

Casey NewtonHost25:44

Yeah.

Kevin RooseHost25:44

You know, put the beef aside.

Casey NewtonHost25:46

Yep.

Kevin RooseHost25:47

We're all in this together.

Casey NewtonHost25:48

Somet- some- what I've found is sometimes business leaders will forget that we are all individual facets of the same universal consciousness-

Kevin RooseHost25:54

[laughs]

Casey NewtonHost25:54

... and that's when they get themselves into trouble.

Kevin RooseHost25:57

[laughs] [upbeat music]

Casey NewtonHost26:01

When we come back, the loudest warning yet on AI and jobs.

Casey NewtonHost26:05

[upbeat music]

Anthropic's Claude is the AI for problem solvers who build things faster, fix things better, and create things that haven't existed before, like Kai Xu, CTO and co-founder at Genspark, who dreamed up an all-in-one AI workspace.

We do AI spreadsheets, AI slides, AI document, all kind of work for normal white-collar workers.

We are a very small startup, and working with Claude as a partner is really, really important.

Watch the series at claude.com/problemsolvers.

That's claude.com/problemsolvers.

I'm Dani Blum, a journalist at The New York Times.

I spend a lot of my time reporting on GLP-1 medications.

People call them weight loss drugs.

They are doing so much more in the body, and there's a lot of hope pinned on them.

I am spending part of truly every single workday making sure that I am up to date on the latest science.

I really wanna make sure that I am talking to the leading researchers, keeping an eye on trial results, and that I'm talking to people who are on these medications to learn more from them, to translate the really messy and murky and shifting science into a way that people can understand.

It's really important to have evidence-based answers to some of the questions that people have, especially about something as vital as a medication they're taking.

That's what you get when you subscribe to The New York Times.

If you already subscribe, thanks.

If you'd like to, go to mytimes.com/subscribe.

Casey NewtonHost28:03

All right, Kevin, let's turn our attention to the question of AI and jobs.

Casey NewtonHost28:07

This is something that I have been fascinated by myself over the past few months.

Casey NewtonHost28:11

I've been speaking to several economists and tech leaders about the short, medium, and long-term risk of an AI getting so good that all of a sudden employers are going to feel like they don't need people anymore, like they can just spend tokens.

Casey NewtonHost28:26

And there's frankly still a lot of skepticism on this point, and so that is why I was fascinated to see a new letter come out this week that had been signed by a group of nearly 200 economists and researchers initially, uh, and they were warning about the impact that AI could have on the economy.

Kevin RooseHost28:43

Yeah, I saw this too.

Kevin RooseHost28:44

This was a statement.

Kevin RooseHost28:45

It was very short, just three short points.

Kevin RooseHost28:47

Uh, it was titled We Must Act Now.

Kevin RooseHost28:50

And basically, I mean, we can probably just read the whole thing.

Kevin RooseHost28:53

It's-

Casey NewtonHost28:53

Read the whole thing

Kevin RooseHost28:53

... it's not that long And it says, "Point one, AI may become radically more powerful over the next 10 years.

Kevin RooseHost29:02

Point two, this could drive an unprecedented transformation of our economy larger than the Industrial Revolution, but unfolding over a vastly shorter timeframe.

Kevin RooseHost29:11

It could bring risks, including large-scale job displacement, as well as opportunities, such as major gains in living standards.

Kevin RooseHost29:18

Point three, economists, policymakers, and technology leaders must act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society."

Casey NewtonHost29:32

[sighs]

Kevin RooseHost29:33

So the statement itself, I think, is pretty anodyne.

Kevin RooseHost29:36

It's basically like, "We should take AI really seriously 'cause it could do all these things.

Kevin RooseHost29:41

It could do some good things, it could do some bad things, and we should pay attention to it." I think what was more interesting was the list of signatories.

Kevin RooseHost29:48

Former Hard Fork guests Anton Korinek from the University of Virginia and Anthropic was on there, as well as, I think, some more skeptical economists who have previously been kind of dismissive of the idea that AI could cause large-scale job transformation.

Casey NewtonHost30:03

At least in the short term.

Kevin RooseHost30:04

Yes, at least in the short term.

Kevin RooseHost30:06

It includes, uh, several Nobel laureates, including Daron Acemoglu, sort of a well-known skeptic of the short-term AI job loss hypothesis, David Autor at MIT, lots of other just sort of, like, well-known economists and thinkers who are coming together to make this statement, and I thought this was really interesting.

Casey NewtonHost30:26

Yeah, and so this week we wanted to talk to a person who helped to put this statement together.

Casey NewtonHost30:31

Uh, that person is Erik Brynjolfsson.

Casey NewtonHost30:34

He is a Stanford economist who helped organize this statement, and he is somebody who takes these threats really, really seriously.

Kevin RooseHost30:41

Yeah.

Kevin RooseHost30:41

I've known Erik for years.

Kevin RooseHost30:43

He's a great economist, and he's been writing about the danger that AI poses to jobs and the economy, um, for a very long time, like, longer than most economists, longer than ChatGPT has existed, uh, he's been writing about this stuff.

Kevin RooseHost30:59

He's a senior fellow at the Stanford Institute for Human-Centered AI and the director of the Stanford Digital Economy Lab.

Kevin RooseHost31:06

Uh, he's been studying AI's impacts on human workers and firms for a very long time.

Casey NewtonHost31:10

So to meet the demand for the questions that we have for him, we now supply Erik Brynjolfsson.

Casey NewtonHost31:15

[upbeat music]

Kevin RooseHost31:23

Erik Brynjolfsson, welcome to Hard Fork.

Erik BrynjolfssonGuest31:25

It's a pleasure to be talking to you again, Kevin.

Kevin RooseHost31:27

So you have been writing and talking and thinking about the transformative potential of AI on the economy for a very long time, longer than, uh, almost any economist I know, and I'm curious what inspired you to organize this letter, uh, this We Must Act Now statement from a bunch of your peers and colleagues.

Erik BrynjolfssonGuest31:48

Well, this is something that's been brewing for a long time.

Erik BrynjolfssonGuest31:51

You're right, I've been thinking about it a long time.

Erik BrynjolfssonGuest31:53

I taught my first course about artificial intelligence back in 1987, before most of your listeners were probably born.

Erik BrynjolfssonGuest32:01

I, I've just been very convinced about what was happening.

Erik BrynjolfssonGuest32:03

But, you know, the nature of exponential curves is they grow, you know, slowly at first and then suddenly, and, uh, and now, especially this year, we're in the, that suddenly part of the curve.

Erik BrynjolfssonGuest32:12

The technology's going so fast, but our economic understanding is nowhere close to keeping up and, and so I'm doing what we can to, to try to change that.

Casey NewtonHost32:21

So the letter is titled We Must Act Now.

Casey NewtonHost32:24

Uh, I would say it doesn't really specify an action.

Casey NewtonHost32:28

So tell us about what your goals are with this letter and, and what you think the policymakers who read it and all the, the sort of esteemed signatories, what should they do?

Erik BrynjolfssonGuest32:38

Well, there's two main purposes.

Erik BrynjolfssonGuest32:39

First is, is that, uh, there's been a sea change, and we're hoping to catalyze that sea change in attitudes about how seriously we take AI.

Erik BrynjolfssonGuest32:48

Uh, as recently as a year or two ago, and certainly before that, most of the economists I talked to really were not very AI pilled.

Erik BrynjolfssonGuest32:56

Uh, they thought, "Hey, it's, it's not a big deal." I remember one, uh, secretary of treasury saying to me, "Oh, do you think that maybe, you know, there'll be a couple of tenths of a percent of change in productivity over the coming years?" So it was sort of like imagining a, a, a sort of a, a very small technological impact.

Erik BrynjolfssonGuest33:12

And I, I personally think it's going to be huge [chuckles] and I, I wanted more people to take it seriously and realize that, uh, I think this is the most important thing economists should be working on.

Erik BrynjolfssonGuest33:22

Whatever other projects they're working on, they should look at it through a lens of AI.

Erik BrynjolfssonGuest33:25

Part of what we're calling for is, is a better understanding, um, more economists thinking hard about this question, uh, more research going into it.

Erik BrynjolfssonGuest33:33

That said, there was the beginning of a direction that, you know, Daron Acemoglu and I and others have been talking about for a while, which is this idea of complementing humans rather than substituting for them.

Erik BrynjolfssonGuest33:44

Now, fleshing out exactly how to do that is a, is a bigger question, but at least if we could get, you know, a couple hundred economists and, and a lot of Nobel Prize winners and, and other AI researchers around the world to agree on, on that direction, uh, that's a good start.

Casey NewtonHost33:57

L- Let me, uh, hit you with some of the objections that I have heard when I've talked to other tech leaders or, um, uh, economists about the idea that AI might take a lot of jobs.

Casey NewtonHost34:09

Um, I myself am totally prepared to believe that there could be a real disruption here, but I do hear some skepticism.

Casey NewtonHost34:14

So one thing I hear is that job losses almost never have any one cause, and it's basically impossible to pinpoint when you're looking at the data.

Erik BrynjolfssonGuest34:22

Well, it's certainly true there's lots of things going on in the economy.

Erik BrynjolfssonGuest34:24

AI's not the only thing.

Erik BrynjolfssonGuest34:25

It's one of the most important and increasingly the most important thing.

Erik BrynjolfssonGuest34:29

You know, we wrote this paper, Canaries in the Coal Mine, that, that you, uh, may be familiar with.

Erik BrynjolfssonGuest34:33

We're gonna have a new version of it out in a, in a few days, uh, probably in time for your, your airing this podcast, um, where we look at some of the alternatives as well.

Erik BrynjolfssonGuest34:40

We look at what's going on with interest rates.

Erik BrynjolfssonGuest34:42

We look at remote work.

Erik BrynjolfssonGuest34:43

We look at the tech over-hiring.

Erik BrynjolfssonGuest34:45

Uh, we look at changes in education.

Erik BrynjolfssonGuest34:47

Um, all of those are important, uh, and we find that all of them make a difference, but even when you put them in individually or all at the same time into these regressions to try to, to, to parse out what's happening, AI continues to be a, a, an important factor.

Erik BrynjolfssonGuest35:01

Um, and it's not just destroying jobs.

Erik BrynjolfssonGuest35:02

AI is also creating jobs, as, as a lot of people will tell you.

Erik BrynjolfssonGuest35:05

Um, and it's the job of us [chuckles] people like me, like economists, to try to sort that out and see which parts are important as early as possible and as fine-grained as possible.

Erik BrynjolfssonGuest35:14

Just because it's not easy doesn't mean we shouldn't try to sort it out.

Kevin RooseHost35:17

Yeah, I, I mean, I confess that I have been wrong about the economic expectations I had for what would happen when AI systems got more powerful.

Kevin RooseHost35:28

I think if you had time traveled back to 2020 and showed me, like, Claude Fable or GPT 5.6, and then you asked me, like, "This technology will exist in 2026, what is the headline unemployment rate?" Uh, I probably would have said, like, 10 or 20%.

Kevin RooseHost35:46

And what we see instead is that employment rates today are basically the same as they were, uh, a few years ago.

Kevin RooseHost35:52

Headline unemployment is hovering around 4%.

Kevin RooseHost35:55

So are you as surprised as I am that this technology becoming so powerful has not created more unemployment?

Erik BrynjolfssonGuest36:02

No [laughs].

Erik BrynjolfssonGuest36:03

N- N- Not at all.

Erik BrynjolfssonGuest36:04

So I, I, I've, I've written many papers, um, on this idea of the productivity J-curve and on the way that AI complements changes.

Erik BrynjolfssonGuest36:12

Um, you may have heard, I'll, I'll tell, I'll retell the story about how, uh, when electricity was first introduced, it took about 20 to 30 years before it fundamentally changed the way work was in American factories.

Erik BrynjolfssonGuest36:24

At first, there was little or no productivity change, there was little or no employment change because people were just kind of paving the cow paths.

Erik BrynjolfssonGuest36:30

They were redoing the same things they used to do, but now with electric motors, and nothing fundamentally changed.

Erik BrynjolfssonGuest36:36

It took a lot of reinvention of those factories, and ultimately the whole economy, to unleash a big change, and that reinvention meant eliminating millions of jobs.

Erik BrynjolfssonGuest36:46

It meant creating millions of jobs.

Erik BrynjolfssonGuest36:49

All of that, in the case of electricity, like I said, took 20 to 30 years.

Erik BrynjolfssonGuest36:53

I don't expect it to take 20 to 30 years this time around, but it's not the kind of thing that happens in, in 20 to 30 months either.

Erik BrynjolfssonGuest36:59

It's something that plays out over, over a period of years.

Erik BrynjolfssonGuest37:02

Um, I think we're in the early stages of that transformation.

Erik BrynjolfssonGuest37:06

Um, I have no doubt there's going to be a lot of disruption, a lot of jobs destroyed.

Erik BrynjolfssonGuest37:11

There's also gonna be a lot of jobs created.

Erik BrynjolfssonGuest37:13

Uh, we're just in the earliest stages of that right now, but it, it's going to get bigger.

Kevin RooseHost37:17

You mentioned the Stanford Canaries dashboard, which I've been following with great interest.

Kevin RooseHost37:22

It showed that early career jobs shrank 2.7% year over year, while mid-career jobs grew 1.6% year over year.

Kevin RooseHost37:32

Is there growing consensus, uh, among economists that this is AI-driven impact, and how should we respond to th- those numbers?

Erik BrynjolfssonGuest37:41

I can't speak for the economics profession as a whole.

Erik BrynjolfssonGuest37:44

I mean, I, I, I believe the numbers.

Erik BrynjolfssonGuest37:45

I've, I've spent a lot of time looking at them.

Erik BrynjolfssonGuest37:47

I think over time it'll become more and more obvious.

Erik BrynjolfssonGuest37:49

Maybe one of the most convincing things in addition, I mentioned we, we, we tested all these other hypotheses and, and they didn't knock out the AI result.

Erik BrynjolfssonGuest37:56

But the other thing, just to...

Erik BrynjolfssonGuest37:57

You know, we first published a version of this back in August of 2025, and we saw, you know, basically the same pattern.

Erik BrynjolfssonGuest38:05

Um, and people said, "Oh, it's a blip.

Erik BrynjolfssonGuest38:06

It's due to, you know, interest rates or whatever." It's just grown over time.

Erik BrynjolfssonGuest38:10

You can go to the Canaries, uh, you can go to the dashboard, um, AI Economic Indicators, uh, at Stanford, and, um, what you'll see is that these trends have just been continuing to grow over time.

Erik BrynjolfssonGuest38:22

It's not something that's just due to COVID or something else.

Kevin RooseHost38:26

I wanna ask one more question, Erik, about this letter, this, this statement.

Kevin RooseHost38:30

Um, obviously the coalition that signed the letter, like hugely impressive group of economists and thinkers, tech executives, um, a, a real who's who of the field.

Kevin RooseHost38:40

Um, but there were some people who didn't like this letter and, and didn't sign it.

Kevin RooseHost38:44

Uh, either they thought it was too vague or it didn't make enough concrete policy recommendations.

Kevin RooseHost38:48

Um, but I wanted to single out one reaction from Noah Smith, who's a, an, an economics writer and blogger.

Kevin RooseHost38:54

Um, he wrote that he didn't sign this statement because, among other reasons, he was just very skeptical of this line about trying to steer AI in a direction that complements humans and benefits society.

Kevin RooseHost39:06

Um, he basically wrote, like, "This is not possible.

Kevin RooseHost39:08

We, we don't know that we can steer this technology toward being complementary." He writes, "How the heck are business people and inventors supposed to steer AI toward being complementary to human workers? They obviously couldn't predict the labor market effects of the last round of AI, at least in the short term, so why should anyone believe the technologists have the ability to purposefully invent different forms of AI with different labor market effects?" What's your response to that?

Erik BrynjolfssonGuest39:33

Well, I love Noah.

Erik BrynjolfssonGuest39:33

He's super smart.

Erik BrynjolfssonGuest39:34

He's a friend of mine.

Erik BrynjolfssonGuest39:35

We hang out, uh, in the Bay Area from time to time.

Erik BrynjolfssonGuest39:38

I'm so glad he, he engaged with the letter, and I'm also glad that not everybody just reflexively agreed with it.

Erik BrynjolfssonGuest39:43

I think we had just the right amount of friction there, a few people disagreeing, just to make it interesting.

Erik BrynjolfssonGuest39:48

It would've been boring if everybody just said, "Yeah, yeah, this is obvious." Uh, that said, Noah's wrong about this.

Kevin RooseHost39:53

[laughs]

Erik BrynjolfssonGuest39:53

Um, you know, uh, w- we have lots of examples in history of technology being directed in different ways.

Erik BrynjolfssonGuest39:59

One of the ideas behind carbon taxes is that, you know, you wanna do that, and there's research that shows that, you know, uh, uh, countries that impose more taxes, engine technology evolved in different ways.

Erik BrynjolfssonGuest40:10

Um, for computers specifically, there's long been two strains.

Erik BrynjolfssonGuest40:14

Uh, you know, Alan Turing talked about the Turing test, imitating, replacing humans.

Erik BrynjolfssonGuest40:19

There's another stream, Doug Engelbart and others, um, who talked about, uh, AI amplifying people.

Erik BrynjolfssonGuest40:26

And one of the things I do, I spend a lot of time talking to CEOs.

Erik BrynjolfssonGuest40:29

Um, let me tell you a, a, a very concrete story.

Erik BrynjolfssonGuest40:31

I, I met with a CFO, and she told me that she wanted to use AI and get higher ROI for it.

Erik BrynjolfssonGuest40:37

And I said, "That's great.

Erik BrynjolfssonGuest40:38

That's what you should be doing." And then she said, "And therefore, what we're doing is requiring measuring in every department, um, how much headcount reduction they have thanks to AI." And I said, "Whoa, hold it just a minute there.

Erik BrynjolfssonGuest40:49

That's one way to measure AI is in replacing workers, and you know, and sometimes that's a good thing, but there's so many other things you can be doing.

Erik BrynjolfssonGuest40:55

You're a CFO.

Erik BrynjolfssonGuest40:56

You're being paid millions of dollars." I didn't actually say that to her, but that was what I was thinking.

Erik BrynjolfssonGuest41:00

Um, "You should be more creative.

Erik BrynjolfssonGuest41:02

Like, let's look at the upside.

Erik BrynjolfssonGuest41:03

What kinds of new products and services? What kind of better customer service? What kind of empl- you know, reduced employee turnover?" There are all these other metrics that a good manager can look at that will increase the upside.

Erik BrynjolfssonGuest41:14

And if they think that way, if they change their mentality- Believe me, they're gonna steer the technology.

Erik BrynjolfssonGuest41:19

You're gonna have outcomes that involve people more, that complement people more, rather than just replace them.

Kevin RooseHost41:25

Yeah, I th- I, I'm just a little more cynical about that than, than maybe you are.

Kevin RooseHost41:29

I, I wanna believe that you're right, and that these enlightened corporate, uh, leaders will say, "Well, you know, we, we really don't want the technology to replace humans, 'cause humans are so valuable, and, and there's all these th- ways that they can be, you know, be complemented by the, by the AI technology."

Casey NewtonHost41:42

We're a family here at the Olive Garden.

Kevin RooseHost41:44

Right.

Kevin RooseHost41:44

[laughs]

Casey NewtonHost41:44

Yeah.

Kevin RooseHost41:45

[laughs] Right.

Kevin RooseHost41:46

But, like-

Erik BrynjolfssonGuest41:46

Well, let, let me just say, uh, th- this, you know...

Erik BrynjolfssonGuest41:48

Th- there's three reasons for doing this, okay? One is that I do think it's the right thing for society, but let's just set that one aside for now.

Erik BrynjolfssonGuest41:54

The second reason is that I think it's in their enlightened self-interest, that they should...

Erik BrynjolfssonGuest41:59

That the more defen- I'm, I, and I could teach at business schools corporate strategy.

Erik BrynjolfssonGuest42:02

The, the, the, the, the better strategy, the more defensible strategy is creating new products and services and barriers to entry, not just cutting costs.

Erik BrynjolfssonGuest42:11

And then the third thing I'll say is it- it's true, there are a lot of tax and other incentives right now that skew companies disproportionately towards investing in capital versus labor.

Erik BrynjolfssonGuest42:22

For instance, marginal tax rates on capital are lower than, than, than marginal tax rates on labor.

Erik BrynjolfssonGuest42:26

And so, so if, if you have a strategy that replaces a lot of workers with machines, you end up paying lower taxes.

Erik BrynjolfssonGuest42:33

That's a flaw, I think, in the tax system.

Erik BrynjolfssonGuest42:34

That's something we can fix and create better incentives for work, for, for managers to treat them more evenly.

Erik BrynjolfssonGuest42:40

And then last but not least, I, I do think there are a lot of people who will, will embrace this because it is the right thing for the world, and, and I think their workforce is gonna like working at that company if it's a company that does right by them.

Kevin RooseHost42:53

Hmm.

Casey NewtonHost42:54

You told Ben Casselman at the New York Times that economists need more reliable data.

Erik BrynjolfssonGuest42:59

Yes.

Casey NewtonHost42:59

What are some of the problems with the data that we have right now, and what would you like to see more of?

Erik BrynjolfssonGuest43:05

You know, this is something I'm, I'm super focused on right now.

Erik BrynjolfssonGuest43:08

Um, sadly, as we're going into one of the most turbulent periods in all of economic history, we are destroying our instrument panel.

Erik BrynjolfssonGuest43:15

Um, the government is cutting back on data collection, uh, eviscerating some of the statistical agencies.

Erik BrynjolfssonGuest43:20

I feel for those people as they're trying to do it.

Erik BrynjolfssonGuest43:22

This is the time, if anything, we should be doubling and, you know, investing more in visibility.

Erik BrynjolfssonGuest43:27

It's a really high ROI activity.

Erik BrynjolfssonGuest43:29

It just costs a tiny bit of the budget and will help us make better decisions, not only as policymakers, but also as managers.

Erik BrynjolfssonGuest43:36

That said, you know, there's only so much I can do about the federal budget.

Erik BrynjolfssonGuest43:38

What I can do is at the Digital Economy Lab, create the AI economic indicators, partner with private organizations.

Erik BrynjolfssonGuest43:45

We're incredibly lucky that there's a bunch of real-time data that's being just collected by companies anyways, you know, Ramp, Revelio, ADP.

Erik BrynjolfssonGuest43:55

And so we can tap into these data sources and get some kinds of visibility.

Erik BrynjolfssonGuest44:00

They're not always collected for the purpose of being representative, so you have to do some work to kind of normalize them and, and, and measure them.

Erik BrynjolfssonGuest44:06

But by, uh, creating an infrastructure to get more visibility, we can do more.

Erik BrynjolfssonGuest44:11

And we're creating a new, a new taxonomy of the kinds of tasks that are being done in the economy.

Erik BrynjolfssonGuest44:15

There's something called O*NET that lists all the tasks that workers do in the American economy, and we're creating an updated version of that.

Erik BrynjolfssonGuest44:22

I think in, you know, over the coming months and years we'll be able to get much more real-time visibility in a much more fine-grained way of what's happening in the economy, and that will allow us to sense and respond to what's doing.

Erik BrynjolfssonGuest44:33

I mean, you can only do so much about predicting what's gonna happen next, but it's sometimes easier to at least, uh, measure it quickly and then allow people to respond to that.

Kevin RooseHost44:42

Eric, I wanna get your take on a policy idea that's been making a lot of waves in recent weeks, which is this idea of a, some sort of a sovereign wealth fund that would be, uh, sort of paid for by the AI companies.

Kevin RooseHost44:56

Like, every American would get a stake-

Erik BrynjolfssonGuest44:58

Mm-hmm

Kevin RooseHost44:58

... effectively in-

Erik BrynjolfssonGuest44:59

Sure

Kevin RooseHost44:59

...

Kevin RooseHost44:59

the frontier AI companies, and you would kind of use that to capture some of the value and redistribute it to American citizens.

Kevin RooseHost45:07

From an economist perspective, do you like that idea, or is there some other way that we should be, uh, getting Americans some share of the proceeds from this booming industry?

Erik BrynjolfssonGuest45:16

It's not my first choice, um, but I think we need sort of an all-of-the-above solution.

Erik BrynjolfssonGuest45:20

So I'd like to see us thinking more about ways that we implement those kinds of, of ways of, of sharing the gains.

Erik BrynjolfssonGuest45:27

My concerns with it, with just a, a, a payout and a sovereign wealth fund is, like, who decides who gets the money and who doesn't? And, and if you don't agree with it, what, what leverage do you have? Um, if you don't pay a farmer, they're not gonna give you their food, or you don't pay a factory worker, they're not gonna give you the, the, the, the, the effort that they put into it.

Erik BrynjolfssonGuest45:45

Um, but they, if you're just getting a dole, you don't really have any leverage to, to step back.

Erik BrynjolfssonGuest45:49

I mean, you hope you can work through your representatives, but, um, yeah, I, I, I think it's, it's m- a less stable equilibrium to just have people getting a, a check that way.

Erik BrynjolfssonGuest45:59

That said, as I said at the beginning, we should, you know, explore a variety of approaches.

Casey NewtonHost46:03

Yeah.

Casey NewtonHost46:03

My understanding about how, like, the new dole would work is that if you had some sort of, uh, issue with it, you would tweet at David Sachs, and if your tweet got enough retweets-

Erik BrynjolfssonGuest46:12

[laughs]

Casey NewtonHost46:12

... then he would respond to you.

Casey NewtonHost46:14

So I agree-

Erik BrynjolfssonGuest46:14

Well, that, that could work.

Erik BrynjolfssonGuest46:15

[laughs]

Casey NewtonHost46:15

Yeah, we could probably come up with a better system.

Casey NewtonHost46:17

Let me throw a couple other ideas up on the board that are, are getting, uh, kicked around in, in this space, Eric.

Casey NewtonHost46:22

What do you think about, like, wage insurance if, uh, you know, you have to take another job because your job, um, goes away and, and you're, you have to... you're forced to work a, a lower paying one?

Erik BrynjolfssonGuest46:32

Yeah, wage insurance is, is another thing economists have been talking about this for, for decades, that there's more potential.

Erik BrynjolfssonGuest46:37

It's kind of a missing market there, just like you have housing insurance and fire insurance and health insurance.

Erik BrynjolfssonGuest46:43

Um, it, it, it's tricky 'cause of, uh, moral hazard, as you call it, that like, okay, you know, you still wanna have incentives for people to work.

Erik BrynjolfssonGuest46:50

Um, but we could have better tools to understand which jobs are being affected and how much of that is due to exogenous forces and s- how many aren't.

Erik BrynjolfssonGuest46:59

I'm, I'm all for exploring that a little bit further.

Erik BrynjolfssonGuest47:01

Again, I, I don't think it's gonna solve the problem, but it could be one piece of the puzzle.

Casey NewtonHost47:06

What about government incentives to hire young workers?

Erik BrynjolfssonGuest47:09

workers?I think some of that actually makes sense as well.

Erik BrynjolfssonGuest47:12

And, and the reason for that is that you don't get, you know, mid-career workers unless they first start as, as young workers.

Erik BrynjolfssonGuest47:19

So there's kind of a, a, a, a market failure there that needs to be stepped into.

Erik BrynjolfssonGuest47:24

Historically, uh, education training have been, to some extent, public goods, where there's value that goes beyond the people who are paying for them, and that's why in almost every country, uh, government and nonprofits have had a role in trying to support them.

Erik BrynjolfssonGuest47:40

I mean, the United States has been for many, for many centuries the number one in the world at making investments in education training.

Erik BrynjolfssonGuest47:47

For, you know, the 1800s, we were way ahead of Europe or other country- uh, other, uh, places in the world, um, at, um, at investing in primary education and secondary education and college education.

Erik BrynjolfssonGuest47:58

It's not true anymore, but, but when we did, that was a big reason why the United States not only had the highest productivity, but also one of the, uh, biggest middle classes in the world.

Kevin RooseHost48:08

Hmm.

Kevin RooseHost48:09

Eric, I know we're almost out of time, but I, I wanted to ask one final question.

Kevin RooseHost48:12

Um, Eric, your field, economics, has been going through some pretty radical changes, and the one that I see the most is just that a lot of your colleagues in academia are going to take jobs at AI labs.

Kevin RooseHost48:25

[laughs]

Erik BrynjolfssonGuest48:25

Yeah, sure.

Kevin RooseHost48:25

It seems like every week I'm- I read on the internet that some esteemed economist is taking a leave or leaving their tenured position at some university to go work for a frontier AI company.

Kevin RooseHost48:36

Um, are you worried that that is going to hollow out the ranks of academic economics departments? And what is your poaching price for any AI-

Erik BrynjolfssonGuest48:44

[laughs]

Kevin RooseHost48:44

... frontier lab that is interested in hiring you?

Erik BrynjolfssonGuest48:47

I've been resisting so far, but I do see, like, uh, my colleagues, yeah, in the offices next door, my friends, my students, um, a lot of them have gone to, uh, to take jobs at frontier labs.

Erik BrynjolfssonGuest48:56

And I gotta say, by and large, you know, these frontier jobs, they're, they're doing great work, um, and they get access to amazing data, and so that's exciting.

Erik BrynjolfssonGuest49:04

At the same time, I think there's a role for people that, you know, uh, don't have a boss who, who, you know, is looking to make profits for the, uh, frontier labs.

Erik BrynjolfssonGuest49:13

And, uh, you know, even if the reality is that they're, uh, unbiased, there's gonna be a perception that, um, they're speaking for that company, and we need people who, who aren't, uh, uh, uh, as affected by that.

Erik BrynjolfssonGuest49:26

I'm not saying academia is completely, uh, uh, unaffected either.

Erik BrynjolfssonGuest49:30

Um, so, you know, I, I do hope people will, will stay a good... lot of smart people will stay in academia [laughs] and continue to do what I'm doing.

Erik BrynjolfssonGuest49:38

I've sat down very hard.

Erik BrynjolfssonGuest49:39

I've had offers from frontier labs and, and I've thought about it.

Erik BrynjolfssonGuest49:42

And, um, for me, the, the marginal utility of more money isn't that high, so I, I just sort of like what I'm doing right now.

Erik BrynjolfssonGuest49:48

Stanford's been great to me, and, uh, I like just being able to have a perch where I can, I can do the research that I'm doing.

Erik BrynjolfssonGuest49:55

Um, so-

Kevin RooseHost49:57

What I, what I'm hearing you say, Eric, is that it would take some kind of, like, meta super intelligence lab-

Erik BrynjolfssonGuest50:02

[laughs]

Kevin RooseHost50:02

... style offer, like maybe a billion dollars to get you out of your chair and, and go work-

Erik BrynjolfssonGuest50:08

Honestly-

Kevin RooseHost50:08

... for a frontier lab

Erik BrynjolfssonGuest50:09

...

Erik BrynjolfssonGuest50:09

I don't know.

Erik BrynjolfssonGuest50:09

I, I... maybe there's something wrong with me, but I feel like, uh, like what am I gonna do? Get a bigger house, more cars? I don't know.

Kevin RooseHost50:15

[laughs]

Erik BrynjolfssonGuest50:15

I, you know, so, so I kind of...

Erik BrynjolfssonGuest50:17

I feel like I have the greatest job.

Erik BrynjolfssonGuest50:19

You know, don't tell my dean or he'll cut my salary.

Kevin RooseHost50:21

[laughs]

Erik BrynjolfssonGuest50:21

But, um, but I feel like I'm...

Erik BrynjolfssonGuest50:23

I've like basically very happy with what I'm doing.

Erik BrynjolfssonGuest50:26

Uh, whatever amount of money I had, I would be doing basically what I'm doing right now.

Erik BrynjolfssonGuest50:30

So I'm lucky in that not everybody's lucky enough to have a job like that.

Erik BrynjolfssonGuest50:33

Maybe you guys do.

Erik BrynjolfssonGuest50:34

You guys look like you're having fun.

Erik BrynjolfssonGuest50:36

Um, and hopefully we'll be entering a world in the future where more and more people can just be doing what they really enjoy doing, like I do, and, uh, and money won't be the main reason that they pick one job versus the other.

Kevin RooseHost50:47

Oh, wouldn't that be nice? [laughs] I like it.

Kevin RooseHost50:49

Well, uh, well, uh, we really do appreciate the work that you do.

Kevin RooseHost50:52

It's great to have, you know, strong independent research and data about this that we can, you know, rely on, uh, as, as we do our work.

Kevin RooseHost50:59

So, uh, thank you very much, Eric.

Kevin RooseHost51:01

Thanks, Eric.

Erik BrynjolfssonGuest51:02

Thank you.

Erik BrynjolfssonGuest51:02

Such fun talking to you guys.

Kevin RooseHost51:05

When we come back, there's cheating at Brown University and other stories from this week's installment of Hat GPT.

Kevin RooseHost51:11

[upbeat music]

The Problem Solvers is a series from Anthropic about the startup founders building new things faster with Claude, like Scott Wu, co-founder and CEO of Cognition.

We build Devin, the first AI software engineer, and a suite of tools around that.

By getting to work with Anthropic, we're able to go and deliver customers the exact capabilities that they're looking for.

This is a world where building software is just as easy as telling your computer what to do.

To hear more from startup founders like Scott, go to claude.com/problemsolvers.

This is Maurice Chammah, host of The Last 12 Weeks, a new podcast from Serial Productions, The Marshall Project, and The New York Times.

A couple of years back, I got an email from a defense lawyer who wanted me to write about his client.

The client, David Wood, was on death row in Texas, had been there for more than 30 years.

The lawyer was writing because David Wood had lost all of his appeals.

He was set to be executed.

The lawyer's plan to stop the execution was to try and prove something that nobody had successfully done in three decades, that one of Texas's most notorious serial killers was actually innocent.

It wasn't that the lawyers didn't have a case to make.

I know two people fabricated testimony to get a guy executed.

It's just that they had so little time to make it.

The Last 12 Weeks.

Listen wherever you get your podcasts.

Kevin RooseHost52:59

Well, Casey, so much news this week that we can't talk about it all in depth, but we can do a round of Hat GPT.

Kevin RooseHost53:10

[upbeat music]

Casey NewtonHost53:10

[upbeat music]That's right, Hat GPT, of course, our segment where we toss the week's news into a hat, pull out items one by one, and discuss them until one of us gets bored and says, "Stop generating."

Kevin RooseHost53:26

Okay.

Kevin RooseHost53:27

Casey, we got a bunch of stuff in the hat this week.

Kevin RooseHost53:29

You go first.

Casey NewtonHost53:31

All right.

Casey NewtonHost53:31

Let's see, Kevin.

Casey NewtonHost53:33

This first story comes to us from The Times and our friend Mike Isaac.

Casey NewtonHost53:36

OpenAI is showing Kalshi's World Cup odds in ChatGPT, Kevin.

Casey NewtonHost53:42

As part of a deal between the two companies, ChatGPT started displaying the Kalshi odds of victory in response to user queries about various World Cup matches.

Casey NewtonHost53:50

For example, in the case of France and Spain, there was a graphic showing that the prediction market had assigned the French a 60% chance of winning, which was a very important information to have, uh, before, uh, France would go on to lose 2-0.

Kevin RooseHost54:04

[laughs]

Casey NewtonHost54:04

So I think this is a perfect story because now there is something in ChatGPT that can help you lose money in a way that does not involve your monthly subscription payment to OpenAI.

Kevin RooseHost54:15

It's beautiful.

Kevin RooseHost54:16

I am, I am circling a theory that actually has to do with this particular story, which is that the two most important forces operating in the world today are, one, AI capabilities, and two, gambling addiction.

Casey NewtonHost54:30

[laughs] Yes.

Kevin RooseHost54:32

This combines them both in a, into a sort of, like, societal Voltron that is probably going to become unstoppable.

Casey NewtonHost54:40

Yeah.

Casey NewtonHost54:40

Yeah, and this one is a really funny one because of course, like, prediction markets are so closely associated with the rationalists, who were a driving force in, like, bringing about this current AI moment, you know? So it sort of looks like, um, you know, maybe a feature in ChatGPT that is, like, steeped in the cultural traditions of the place that birthed it.

Casey NewtonHost54:58

But on the other hand, it is just gambling, and I, as a, you know, like, a paying subscriber to ChatGPT, uh, it, it does offend me that I have to see gambling advertisements, uh, in my, uh, responses now.

Kevin RooseHost55:10

Yeah.

Kevin RooseHost55:10

I ju- I just think this is going to be sort of everywhere, unless there's some sort of regulatory response.

Kevin RooseHost55:15

Like, I expect to, like, open a box of cereal, and it should just give me some Kalshi odds for something-

Casey NewtonHost55:21

[laughs] Yeah

Kevin RooseHost55:21

... like in a year or two.

Kevin RooseHost55:22

I just think there's going to be no surface for our attention that does not in some way have a gambling ad attached to it.

Casey NewtonHost55:28

All right.

Casey NewtonHost55:29

[laughs] Stop generating.

Kevin RooseHost55:30

Okay.

Kevin RooseHost55:31

Next out of the hat, New York enacts nation's first statewide moratorium on data centers.

Casey NewtonHost55:37

Ooh.

Kevin RooseHost55:37

Governor Kathy Hochul, former Hard Fork guest, signed an executive order recently pausing the construction of the largest data centers for one year while the state assesses their effects on the environment and energy use.

Kevin RooseHost55:49

This is a temporary pause on hyperscale data centers which use 50 or more megawatts of power to operate.

Kevin RooseHost55:56

Um, and President Trump responded to this on Truth Social, calling it a terrible decision and saying, "The radical left Democrats must not be allowed to cause us to lose data centers, AI, and all this incredible new technology to China and other countries." Casey, what do you make of the statewide moratorium in New York?

Casey NewtonHost56:16

So I, I, I, I have to say I'm quite sympathetic here.

Casey NewtonHost56:20

Um, the state legislature had passed a moratorium that would've been even more restrictive than this, and so my, um, belief here is that the governor wanted to find a kind of middle path and throw a sop to people who are worried about data centers in their communities without sort of going as far as some of the, the legislature did.

Kevin RooseHost56:38

It was sort of a, a less-atorium.

Casey NewtonHost56:40

[laughs] Yes.

Kevin RooseHost56:42

[laughs]

Casey NewtonHost56:42

It really was.

Kevin RooseHost56:45

[laughs]

Casey NewtonHost56:45

Very good.

Kevin RooseHost56:46

Thank you.

Casey NewtonHost56:46

Um, and so look, I think it is reasonable to say that, uh, there are a lot of questions about the impact of these data centers, and we want to investigate it.

Casey NewtonHost56:56

Now, you know, my hope is that during the next year, they do what they're saying here, and they get a handle on what is the actual impact, um, to, like, the average ratepayer, right? And can we address some of these environmental concerns that are out there meaningfully? Um, you know, we, we've talked on the show in the past about, well, maybe some of the, you know, water use claims are, uh, are, are overstated here, a- and I do believe that, but I'm reading other stuff about data centers contributing to, like, an urban heat island effect, for example.

Casey NewtonHost57:25

Um, or like, a, like, w- like, um, uh, sort of, uh, pol- polluting the water supply.