The Indicator from Planet Money

Episode•January 24, 2025•9 min

Why China's DeepSeek AI is such a big deal

Indicators of the Week unpacks numbers from the news. This week, we delve into what President Trump's first AI announcement reveals about the economics of the industry, a Chinese company's answer to OpenAI, and the reason why Prince Harry may have settled with Rupert Murdoch's media empire. Related listening: Is AI overrated or underrated? (https://www.npr.org/2024/09/11/1198748511/is-ai-overrated-or-underrated) (Apple (https://podcasts.apple.com/us/podcast/planet-money/id290783428?i=1000669194968) / Spotify (https://open.spotify.com/episode/4rETAdDnEOH6PVsdTvRCt3?si=71205a9cabf34218)) AI creates, transforms and destroys... jobs (https://www.npr.org/2023/12/08/1197958787/ai-jobs-friday) For sponsor-free episodes of The Indicator from Planet Money, subscribe to Planet Money+ via Apple Podcasts or at plus.npr.org (http://plus.npr.org/). Fact-checking by Sierra Juarez (https://www.npr.org/people/g-s1-26724/sierra-juarez#:~:text=for%20Planet%20Money-,Sierra%20Juarez%20is%20a%20researcher%20and%20fact%20checker%20at%20the,and%20fact%20checking%20in%20Mexico.). Music by Drop Electric (https://dropelectric.bandcamp.com/). Find us: TikTok (https://www.tiktok.com/@planetmoney), Instagram (https://www.instagram.com/planetmoney/), Facebook (https://www.facebook.com/planetmoney), Newsletter (https://www.npr.org/newsletter/money). To manage podcast ad preferences, review the links below: See pcm.adswizz.com (https://pcm.adswizz.com) for information about our collection and use of personal data for sponsorship and to manage your podcast sponsorship preferences. Learn more about sponsor message choices: podcastchoices.com/adchoices (https://podcastchoices.com/adchoices) NPR Privacy Policy (https://www.npr.org/about-npr/179878450/privacy-policy)

Apple Podcasts Overcast

Transcript

0:01

Npr.

0:11

This is the Indicator from Planet Money. I'm Darian Woods.

0:14

I'm Waylon Wong. And joining us today is Planet Money's very own Jeff Guo. Hello, Jeff.

0:22

Always nice to be here. Waylon.

0:25

It is time for Indicators of the Week. I'll look at the numbers from around the news this week. We have figures from President Trump's first major AI announcement.

0:35

We've also got Chinese AI makers making

0:37

waves and a royal settlement with some eye popping details.

0:42

That's all after the break.

0:46

It's Indicators of the Week. Darian, you're up first today.

0:49

My indicator is $500 billion. And that was the focus of a White House press conference on Tuesday.

0:56

Future, a new American company that will

0:58

invest $500 billion at least in AI

1:02

infrastructure in the United States.

1:03

And President Donald Trump stood at the podium flanked by tech titans Larry Ellison and OpenAI, Sam Altman, as well as tech investor Masayoshi Sun. He was announcing that this new company would immediately invest $100 billion in building new servers to run AI models. It would then aim to invest a total of 500 billion DOL dollars within four years.

1:26

That's a lot of money even for a tech company, you know.

1:29

Yes. And what I really want to talk about today is how this $500 billion price tag highlights something unusual about the economics of AI.

1:37

I'm listening.

1:38

So to set the scene, a big reason we have tech monopolies like Google or Microsoft is because it costs a lot of upfront money to build a good search engine or spreadsheet software. But, but then each new search result or copy of Microsoft Excel costs the company close to zero. As economists say, the marginal cost is really low. And so that means it's hard for competitors to go up against a Google or a Microsoft. The tech giants have done that upfront investment, building their products and can do whatever it takes to retain customers. But for AI, the cost of that question you're asking ChatGPT or Claude is is not close to zero.

2:20

The other week my husband was like, we should ask ChatGPT about this. And I was like, but what about our utility bill? And I was obviously being flipped. But I've read that, you know, like each AI query takes enough energy to power a light bulb for a few minutes, like a lot of electricity.

2:35

Yeah, it's 10 times more energy than a traditional Google search. And that's for the common models. The more advanced AI models are way more intensive. And, and that's where the old economic framework of super low marginal costs might be quite different for AI. If the costs of an AI query is still quite expensive, a company like OpenAI might not be able to amass as much domination as the old tech companies did.

3:02

Right, because it's not enough just to have the best AI models, you also have to have the giant data center to back it up.

3:09

That's right.

3:09

Okay, well, my indicator this week, coincidentally, also has to do with AI, but it has to do with these AI models. So on Monday, a company released what might be the most powerful AI model in the world right now.

3:25

Okay.

3:25

Yeah.

3:26

Whoa.

3:26

Yeah, but it wasn't OpenAI. It wasn't Google or Meta. It was this little Chinese startup called DeepSync. And apparently their new model, it goes toe to toe with the most powerful model from OpenAI itself. Deepseek claims that it even beats OpenAI's model on search certain math tests.

3:45

So people kicked the tires on this.

3:47

Well, that leads me to my indicator of the week, which is the number zero. Because this super powerful new AI model, it is free, it is open source. Deepseek released the whole thing to everybody on Monday.

4:02

Oh, so anybody can download it and run this model?

4:05

Yeah, they've even released, like, smaller versions of the model you can run on your laptop. So that is what I've been downloading literally right now. So.

4:14

So hang on, what's your status bar? How long do we have to sit here? Jeff?

4:17

Wait, so. No, no, no, it's downloaded. Hold on, let me share my screen for a second. Yeah, all I got to do, I hit run. Would you like to meet your new AI overlord?

4:27

Sure.

4:27

Anybody want to ask you a question?

4:29

What is the sound of one hand clapping?

4:31

What is the sound of one hand clapping?

4:34

Let's do it.

4:36

It's processing. Okay. So you see, it takes you through its, like, chain of thought. So it's like, okay, I'm trying to figure out this question about one hand clapping. Is it this like a. Is it like a philosophical question? Is it a riddle? Yada, yada, yada. So it just goes on and on and on.

4:50

Oh, it's so chatty.

4:51

Yeah. And it like, self critiques itself to try to find holes in its own answer. And it kind of just like goes in circles and circles.

4:58

It said in summary. And now it's giving us a numbered list with more answers.

5:03

Yeah. So this is a huge deal in the world of AI. It's a huge deal for the entire world for two reasons. Number one, there are, like, major geopolitical implications.

5:15

Right.

5:15

Because for the past few years, the US has been trying to Slow down China's AI progress. It's been blocking the sale of advanced microchips to China.

5:24

Seems like that hasn't worked.

5:25

Yeah, it seems like even without those advanced chips, Chinese companies are still able to make cutting edge AI.

5:31

Which maybe relates to my point earlier about the economics of AI being perhaps different than traditional software. Even more so if the actual research and development is more easy to copy, but then the actual running is where that's going to be costlier and require companies competing to do.

5:53

Yeah, exactly. That's kind of the other reason this is such a big deal. Because, you know, companies like OpenAI, they've been able to charge a premium because they have, you know, supposedly the most advanced models in the world, but now they're competing with free. And on top of that, you know, they have all these data center costs that they have to, you know, pay. This is your new AI overlord. This is the cutting edge right here in front of us.

6:15

So what's the tldr on our clapping question?

6:18

So while physically there's no sound from one hand clapping, metaphorically, it speaks to individuality, focus, and the limitations of singular actions without sound support.

6:28

Thank you. Thank you. Deep sea.

6:30

That's very thoughtful. Okay.

6:32

It really is deep.

6:34

And that brings us to your indicator. Whan across the very deep Atlantic to the United Kingdom.

6:43

That's right. My indicator is about Prince Harry, AKA the Duke of Sussex. He settled a lawsuit against News Group Newspapers. This is the Rupert Murdoch owned media company that Prince Harry had accused of hacking his phone and doing other unlawful things. So as part of this court settlement, the company issued a quote, full and unequivocal apology for intruding into Prince Harry's private life for 15 years. That is my indicator. 15 years.

7:12

And so this is the same phone hacking scandal that we've been hearing about in the news for quite a long time.

7:18

Yes. This scandal is almost 20 years old, depending on where you want to start the timeline. And the settlement with Prince Harry comes at the last minute. The two sides will now avoid a trial. And there actually have been over 3, 1300 settlements in this scandal. Murdoch's companies have paid out over a billion dollars in settlements over the years.

7:38

Wow. So did Prince Harry get what he wanted?

7:41

Well, he did get an undisclosed amount of money. The New York Times reported that it was at least $12 million. He also got a public apology, plus an apology to his late mother, Princess Diana, for the intrusion into her private life when Prince Harry was young. But the thing is that he had also talked about how airing all of this out trial was necessary for full accountability, and now there's no trial. And what I found really interesting is that economics might have played a role here.

8:12

Okay, so some game theory.

8:14

Yeah, it's always time for some game theory. So under British law, if they had gone to trial and Prince Harry had won, but the damages he got awarded were less than the settlement offer, he would have had to pay the other side's legal costs.

8:29

And that could be a lot of money, even for a prince or a duke or whatever his title is now.

8:35

Yeah, I mean, last year, the actor Hugh Grant settled his lawsuit against the same Murdoch owned media company. And he said at the time, the prospect of having to pay the other side's legal costs was just too risky. Like he couldn't, he couldn't do it. So I think he said something like, I am shying from that fence, which I think is like an equestrian thing, but I'm not sure.

8:53

And this is not the way it

8:54

works in the U.S. uh, no. So I think the incentives are quite different when it comes to settling this kind of thing.

9:01

He's just a boy standing in front of a court asking to settle a super expensive lawsuit.

9:07

Darian. And the Oscar goes to Darian Woods. Who would you like to thank?

9:13

Darian.

9:14

Well, I'll do it. This episode was produced and fact checked by Julia Richie with engineering by Neal Rauch. It was edited by Patty Hirsch. Katie Canyon is the show's editor, and the indicator is a production of npr.

Why China's DeepSeek AI is such a big deal

0:00

0:00

Related Episodes

Why Paramount went looney tunes for Warner Bros.

Why Paramount went looney tunes for Warner Bros.

Mar 3, 20269 min

Paramount SkydanceWarner Bros. DiscoveryDavid Ellison

Are we in an economic 'doom loop'?

Are we in an economic 'doom loop'?

Feb 9, 20269 min

Eswar PrasadDoom LoopGlobalization

America's next top Fed Chair

America's next top Fed Chair

Feb 2, 20269 min

Kevin WarshFederal ReserveJerome Powell

Why isn’t corporate America standing up to Trump?

Why isn’t corporate America standing up to Trump?

Jan 28, 20268 min

Donald TrumpJamie DimonJPMorgan Chase