Stephanomics

Jason Furman discusses AI replacing economists, policy impacts of AI data centers, and his EconVitals platform

34 min Listen to the original Episode page
Artificial IntelligenceEconomic PolicyUS ElectionsLabor MarketMacroeconomics

Summary

Stephanie Flanders interviews Jason Furman about the economic and political implications of artificial intelligence. Furman argues that voter backlash against AI data centers stems from concerns over utility costs and job displacement. He outlines policy recommendations including healthcare affordability and permitting reform to lower energy costs. Furman also discusses his EconVitals platform, an AI-built macroeconomic tool he claims replaces the work of 20 economists. He expresses concern that AI may reduce on-the-job training opportunities for junior economists.

Key takeaways

  1. 1
    The thing is, those policies are only for about 10% of households.

    Jason Furman · 4:13

  2. 2
    Furman argues divided government is fiscally expansionary because Democrats will object to deficit-cutting measures (like Medicaid cuts) but not tax cuts, leading to higher deficits.

    Jason Furman · 10:32

  3. 3
    Unfortunately, the data centers is a little bit of a distraction and a sideshow. They'll get built somewhere in the world, some which way, as long as there's an economic return to them, which is itself a question. that we could discuss. But if there's an economic concern, they're going to get built. For me, the the two things are number one, a regulatory framework for safety, and two, an economic framework for economic security.

    Jason Furman · 14:38

Episode sections

  1. 0:01 – 1:31

    Intro and Sponsor Read

    The episode begins with a sponsor read for Samsung Galaxy S26 Ultra and Bell 5G network, followed by an introduction to the Bloomberg Audio Studios podcast Stephanomics.

  2. 1:31 – 3:24

    AI's Impact on the Economy and Midterms

    Stephanie Flanders discusses the US midterm elections, noting voter backlash against AI data centers and concerns about affordability. She introduces Jason Furman, who built an AI platform to replace the work of a team of economists.

  3. 3:24 – 7:00

    Healthcare and Tariffs Policy

    Furman identifies healthcare affordability as a key election issue, suggesting Democrats should restore ACA tax credits and cut tariffs on Canada. He argues that permitting reform is necessary to increase energy supply and lower costs.

  4. 7:00 – 13:59

    Tariffs and Fiscal Deficit

    Furman predicts most tariffs will remain due to bipartisan support for China tariffs and revenue needs. He argues that divided government will likely increase the deficit, as Democrats will not raise taxes to fund spending cuts they oppose.

  5. 13:59 – 17:17

    AI Regulation and Economic Security

    Furman argues that data center construction is a sideshow; the real issues are safety regulation and economic security. He suggests policymakers should focus on making work more rewarding and improving vocational training rather than regulating AI models directly.

  6. 17:17 – 20:55

    EconVitals Platform and AI Capabilities

    Furman describes EconVitals.org, a free macroeconomic platform he built using 500,000 lines of 'vibe-coded' AI. He claims it performs the work of 20 economists and has a low error rate, comparable to major financial publications.

  7. 20:55 – 31:32

    Impact on Junior Economists and Training

    Furman expresses concern that AI reduces opportunities for junior economists to learn through hands-on analysis. He suggests that on-the-job training may decline, potentially requiring a shift toward vocational education and non-compete agreements to retain trained workers.

  8. 31:32 – 34:05

    Advice for Students

    Furman advises students to become proficient in using AI agents and to understand the limitations of AI to fill in gaps that machines cannot handle.

Claims worth verifying

Jason Furman built a system of over 500,000 lines of code using 'vibe-coding' that recreates the work of a 35-person team at the Council of Economic Advisors.

Jason Furman ·1:16

How to verify: Check the code repository for EconVitals.org if public, or verify Furman's claim of building the system via his own statements and the platform's functionality. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Voters swung dramatically against AI data centers in the US midterm elections, particularly where they could vote against further construction.

Unidentified speaker (UNKNOWN) ·1:45

How to verify: Review US midterm election results and ballot measures related to data centers and AI infrastructure in 2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Healthcare affordability was a dominant issue in the midterms, with Medicaid cuts and ACA tax credits being key topics.

Jason Furman ·4:13

How to verify: Analyze midterm election exit polls and campaign advertising spending data from sources like Ad Impact or Kantar Media. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Policies to undo Medicaid cuts and restore ACA tax credits affect only about 10% of households.

Jason Furman ·4:33

How to verify: Compare the number of households receiving ACA subsidies or Medicaid expansion benefits to the total number of US households using Census Bureau data. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

A permitting reform bill is moving through the Senate to increase energy supply and affordability.

Jason Furman ·5:12

How to verify: Track the legislative status of the permitting reform bill in the US Senate via Congress.gov. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Divided government is generally fiscally expansionary, and Furman expects this time to be even more so because Democrats will not raise taxes to fund spending.

Jason Furman ·10:32

How to verify: Compare historical federal deficit data during periods of divided vs. unified government using Congressional Budget Office (CBO) reports. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Most tariffs will last beyond the next two years due to bipartisan support for China tariffs, industry lobbying, and revenue needs.

Jason Furman ·7:50

How to verify: Monitor US trade policy announcements and Congressional voting records on tariff legislation in 2026-2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

EconVitals.org is a free platform that provides macroeconomic tracking, data analysis, and policy simulation, with 25 weekly emails, most written by LLMs.

Unidentified speaker (UNKNOWN) ·17:51

How to verify: Subscribe to the EconVitals newsletter and analyze the authorship attribution and content quality compared to human-written economic reports. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Furman holds four Claude Max accounts, noting that Harvard pays for them and that usage has become more efficient.

Unidentified speaker (UNKNOWN) ·24:34

How to verify: Verify Furman's statement about his subscription accounts and Harvard's payment policy through public disclosures or Furman's own statements. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Opus 5.5 is cheaper and faster than earlier Opus models, and Fable 5.2 is more advanced than needed for most business use cases.

Jason Furman ·26:11

How to verify: Check Anthropic's pricing page for Opus 5.5 and Fable 5.2 to verify cost and performance claims. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Do your own research

  • Visit econvitals.org to review the platform's capabilities and code documentation if available.
  • Check Furman's GitHub for the EconVitals project repository.
  • Review US midterm election exit polls and local ballot measure results regarding data center zoning or taxes.
  • Compare the number of households receiving ACA subsidies against total US households using Census Bureau data.
  • Track the legislative status of the permitting reform bill in the US Senate.
  • Monitor US trade policy announcements and Congressional voting records on tariff legislation.
  • Compare historical deficit trends during periods of divided vs. unified government using CBO data.
  • Subscribe to the EconVitals newsletter to analyze authorship attribution and content quality compared to human-written economic reports.
Full transcriptexpand
0:01
Unidentified speaker (UNKNOWN): Oh,
0:01
Unidentified speaker (speaker_0): group shot.
0:02
Unidentified speaker (speaker_1): I'll take it.
0:03
Unidentified speaker (speaker_0): No, we're good. There's one in every group:
0:06
Unidentified speaker (speaker_1): the
0:06
Unidentified speaker (speaker_0): notographer.
0:07
Unidentified speaker (speaker_1): Best
0:07
Unidentified speaker (speaker_0): intentions,
0:08
Unidentified speaker (speaker_1): worst shots. But there's hope. The Samsung Galaxy S26 Ultra on Bell. Super steady videos and perfectly framed shots with my FamCam. And nitography for stellar picks even in low light.
0:19
Unidentified speaker (speaker_2): I want that. Yep,
0:20
Unidentified speaker (speaker_1): postworthy with just a to click. All on Bell, Canada's fastest 5G plus network. Visit a bell store or bell.ca slash Samsung for details. Bell, connection is everything.
0:32
Unidentified speaker (speaker_0): Bloomberg Audio Studios,
0:35
Unidentified speaker (speaker_2): podcasts, radio, news.
0:38
Jason Furman: When I was chair of the Council of Economic Advisors, I had about 35 people working for me. And I just had the most amazing data inflow, constant monitoring, updating, memos, analysis. coming my way. And also a team of people that could basically answer any question about the economy I had
0:57
Jason Furman: with a combination of theory, data, evidence, other things. And I basically set about recreating that. It's over 500,000 lines of code. That would have been a lot of coders to build that system. And it was all built by me. And I didn't read a single one of those lines of code. It was all just
1:16
Jason Furman: vibe-coded.
1:20
Unidentified speaker (UNKNOWN): I'm
1:20
Unidentified speaker (speaker_0): Stephanie Flanders,
1:21
Unidentified speaker (speaker_2): Head of Government
1:22
Unidentified speaker (UNKNOWN): and Economics at
1:23
Unidentified speaker (speaker_2): Bloomberg, and
1:24
Unidentified speaker (UNKNOWN): this is once again
1:25
Unidentified speaker (speaker_2): Stefanomics, your weekly guide to the forces reshaping the global economy. We're recording
1:31
Unidentified speaker (UNKNOWN): this on Wednesday, the 11th of November. It's just a week after the midterm elections in the
1:35
Unidentified speaker (speaker_2): US, in which
1:36
Unidentified speaker (UNKNOWN): voters swung dramatically against.
1:39
Unidentified speaker (speaker_2): Republicans, but at least
1:41
Unidentified speaker (UNKNOWN): in some places the swing against AI was even
1:44
Unidentified speaker (speaker_2): greater. With one
1:45
Unidentified speaker (UNKNOWN): exception,
1:46
Unidentified speaker (speaker_2): anywhere
1:46
Unidentified speaker (UNKNOWN): that voters could vote against further
1:48
Unidentified speaker (speaker_2): AI data centers being built, they did. People often associated AI with a higher cost of living, especially higher utility costs, and they fear, We many of them, that AI is about to destroy millions of jobs, certainly make it much harder for their children to get their first foot on the job ladder. That suggests quite a thorny challenge for
2:08
Unidentified speaker (speaker_2): the incoming Democratic majority, which I think we've already seen highlighted in the Cesaw in tech stocks in the days following their election win. Democrats need to respond to voters' concerns about AI, along with that much bigger, broader complaint about higher costs, but not seriously undermine. a massive
2:28
Unidentified speaker (speaker_2): bet on artificial intelligence, which is sustaining a large part of America's economic growth. Well, Jason Furman, I know, will be someone that many of those Democrats will be turning to for advice on this. He is professor at Harvard University and was chairman of President Barack Obama's Council of Economic Advisors. He's also been
2:48
Unidentified speaker (speaker_2): on this show a few times. But it has to be said he's not. someone who's going to provide much reassurance about the job implications of widespread adoption of AI. Because his own experiments with it, illustrated by a free platform, econvitals.org, that he launched a few weeks ago, suggests it's going to be pretty transformational for future economic policymakers,
3:08
Unidentified speaker (speaker_2): maybe even replacing a lot of the PhD economists who might previously have worked in, for example, President Obama's Council. of Economic Advisors.
3:16
Unidentified speaker (UNKNOWN): But
3:16
Unidentified speaker (speaker_2): I'll let him tell you all about that in a minute. Jason, thanks for coming back.
3:20
Jason Furman: Great to be back with
3:21
Unidentified speaker (UNKNOWN): you.
3:24
Unidentified speaker (UNKNOWN): Can
3:24
Unidentified speaker (speaker_2): we spend a few minutes on the midterms? AI, as I mentioned, but also the sort of broader cost of living concerns were obviously a massive factor. And now Democrats face this question of how to address those realistically, given that the levers they have in their hands are quite limited, at least when it comes to things like tariffs and the
3:44
Unidentified speaker (speaker_2): Iran war. What do you think they should be looking to show that they are concerned about this in a real way rather than just having symbolic votes on things?
3:53
Jason Furman: Yeah, one of the biggest issues, and it goes under the heading of affordability in the election, was actually healthcare, which is interesting. It used to be the dominant issue. in American elections. It faded for a bit, but if you look at the amount of money spent on advertising what voters said in the exit polls, it was incredibly
4:13
Jason Furman: salient. The first thing Democrats, I think, will do and should do, and I fully support it, is undoing some of the cuts that President Trump made to Medicaid and bringing back expanded tax credits. for the Affordable Care Act. The thing is, those policies are only for about 10% of households.
4:33
Jason Furman: So most people watching those ads, listening to them, worried about it, it won't reach them. To do that, you need the sort of harder, longer, tougher work of reforming healthcare delivery systems. And I've heard a much less serious conversation about that part. Healthcare of course, is just one thing. Tariffs are
4:53
Jason Furman: another. I would get rid of all of them, but the place they'll probably start is the tariffs on Canada. And those may be the single stupidest set of tariffs that were done. So I think it's a perfectly great place to start. And then right now already Congress can do something about it. There's a big permitting reform bill making its way through the Senate would make
5:12
Jason Furman: energy supply increase more affordable. I think that would be a great place to start as well.
5:19
Unidentified speaker (speaker_2): Will you see the effects of that soon enough? I mean I just have a feeling that sort of permitting isn't going to make a typical kind of democratic socialist as heart beat faster as they've kind of made their way into Congress. And I guess they would also say that some of those big supply side reforms under President Biden didn't do him much good politically.
5:39
Jason Furman: Yeah, well, President Biden's supply side reforms in part didn't do him as much good politically because he didn't do the permitting reform with them. And so you passed this transformational legislation on things like broadband and electric cars, and then nothing ended up getting built because it wasn't able to be built. So
5:59
Jason Furman: I think there is some complementarity. But look, I think people need to. be honest about the trade-offs and difficult things here. Ultimately, the best solutions are on the supply side, but those tend to be relatively small. There's things you can do on the demand side that help some people, but can't help everyone. Because when you do things on the demand side, they
6:19
Jason Furman: also tend to drive up before tax prices. And so if you want to help some groups, great, give them subsidies, but we can't really afford to give subsidies for everyone. And then the last thing that works right away is price controls, but it works right away for some people who are lucky enough to get the price controlled good. And it makes it worse and
6:39
Jason Furman: hurts the supply for everyone else. So there's some combination of demand subsidies, supply enhancements, and probably very little in the way of price controls, at least in my drone druthers, that one needs to do. But none of these are going to be commensurate to the frustrations and problems that Americans
6:59
Jason Furman: have.
7:00
Unidentified speaker (UNKNOWN): Just
7:00
Unidentified speaker (speaker_2): quickly to go back on tariffs, we had a conversation about this last week, and then I noticed that Hakeem Jeffries, one presumes the new speaker of the House, has said he'll eliminate Donald Trump's harmful tariffs. And then there's obviously this lack of clarity. about whether it's just the ones that he considers harmful or all of them being harmful. They're raising
7:20
Unidentified speaker (speaker_2): quite a lot of money. So some would say they're going to stick around for a while, maybe not the Canada ones, as you mentioned. But the other ones, we haven't heard a real alternative, at least in terms of those revenues.
7:30
Jason Furman: Yeah, I mean there well, there's three things. One is a bunch of them are on China, and there's bipartisan consensus about being tough on China. China. The second is certain industries have come to rely on them and will lobby for keeping them. Now, I don't think that's a good argument for them, but just as a political prediction. And then the last issue, as you said, is
7:50
Jason Furman: the revenue one, which is why I do think the majority of the tariffs are going to last at least for the next two years, and my guess is beyond the next two years. And what you'll see is maybe certain salient ones on certain types of consumer goods, certain salient, extraordinarily high ones on certain
8:10
Jason Furman: countries like Canada. But yes, I think unfortunately the bulk of the tariffs are going to be with us.
8:17
Unidentified speaker (speaker_2): They are enormously unpopular there. I
8:19
Unidentified speaker (UNKNOWN): mean,
8:19
Unidentified speaker (speaker_2): they've become a lot more unpopular in the last two years, for good reason in some ways, but the ordinary voter, I think. probably exaggerates the impact on inflation and may think, oh, you have to get rid of them in order to reduce prices. How are Democrats going to get around that?
8:35
Jason Furman: Look, I mean, ultimately part of why they're unpopular, I wish it was because people understood Ricardian equivalents and all the things I teach about
8:42
Unidentified speaker (UNKNOWN): trade. I wish
8:43
Jason Furman: it was all of that. I think a lot of it is that they're basically a tax. And we actually do need more taxes. That's the sad thing from my perspective. perspective. And taxes are gonna hurt somebody. I think there's less dumb ways to raise revenue, there's broader ways, there's more progressive ways, there's less distortionary ways, etc.
9:03
Jason Furman: But the fundamental objection is that, and that's part of why they're gonna last. Because if you put them up with legislation head to head with another way to raise revenue, it's not like the other. way to raise revenue would be a whole lot more popular than the tariffs themselves.
9:20
Unidentified speaker (speaker_2): When you think about raising money, on the first day after the election there was this kind of surprisingly kind of benign market reaction. And there were lots of what I thought were quite kind of contorted, slightly odd explanations for that. But one of them was and long long-term term interest rates, bond yields also remained calm. And that was based on this kind of idea that was around that if
9:40
Unidentified speaker (speaker_2): you had the Congress was in a different con under different control from the White House, nothing really could be done to make the deficit worse. And therefore it would be kind of relatively neutral for the cost of borrowing for the government for bond yields. And yet, when you actually look at the real situation and the kind of thing you're describing, I can
10:00
Unidentified speaker (speaker_2): imagine the Democrat majority doing deals with the president when it comes to even restoring the Affordable Care Act credits for healthcare. I can't imagine him raising taxes to pay for those things.
10:12
Jason Furman: Yes, I think that diagnosis is exactly right. Divided government has generally been somewhat fiscally expansionary or deficit increasing. I think this time I would expect it to be even more so because when it came to Trump's first term or President Bush, the main things Democrats objected
10:32
Jason Furman: to were ways in which they had cut taxes, which is to say, the main democratic objections, if they had been able to act on them, would be deficit reducing. Here, the main democratic objections are to ways in which they cut the deficit through Medicaid, through tariffs, and the like. There wasn't a whole lot of
10:52
Jason Furman: concern over the Trump tax cuts expressed in the campaign. And even if there was any concern, as you said, President Trump could veto it. And by the way, some of those new tax cuts, like tax cuts on ticks, tax cuts on overtime, I don't know a single economist that liked those, but I don't know a single politician that objected in any serious way
11:12
Jason Furman: to them either.
11:17
Unidentified speaker (speaker_0): Oh, group shot.
11:19
Unidentified speaker (speaker_1): I'll take
11:20
Unidentified speaker (speaker_0): it. No, we're good.
11:21
Unidentified speaker (speaker_1): There's one in every group. The notographer. Best intentions, worst shots. But there's hope. The Samsung Galaxy S26 Ultra on Bell. Super steady videos and perfectly framed shots with my fan cam. And nitography for stellar picks even in low light.
11:36
Unidentified speaker (speaker_2): I want that.
11:36
Unidentified speaker (UNKNOWN): Yep,
11:37
Unidentified speaker (speaker_1): postworthy with just a click. All on Bell, Canada's fastest 5G plus network. Visit a Bell store or bell.ca slash Samsung for details. Bell. Connection is everything.
11:47
Jason Furman: Hi, I'm Barry Ritholtz, inviting you to join me for the Masters in Business Podcast. Every week, we bring you conversations with the people who shape markets, investing, and business. I speak with CEOs, Nobel laureates, market innovators, and legendary. investors. Whether you own stocks, bonds, real estate commodities, even
12:06
Jason Furman: crypto, these are discussions you absolutely need to hear. Subscribe to the Masters in Business podcast on
12:14
Unidentified speaker (speaker_0): Apple, Spotify, or anywhere you listen.
12:21
Unidentified speaker (UNKNOWN): There
12:22
Unidentified speaker (speaker_2): surely is a risk to mortgage rates on the back of this. And I mean, when we get back to this question of how much can the Democratic majority really control, I mean, if one of the levers they can pull to help at least some people is things that are going to also increase deficit spending, then that could potentially be adding to higher mortgage rates for everybody else.
12:40
Jason Furman: Yeah, absolutely. I mean, I can think of a single broad agreement between the two parties on anything that would reduce the deficit, you know, raise taxes or cut spending. There's no item on that. list. But there's a number of items where you have not consensus, but some broad agreement about spending more, cutting back the
13:00
Jason Furman: tariffs. And yeah, so buckle up. I think we're making our fiscal problems a bit worse over the next two years. Now, the big question mark there is do we get more productivity growth? And does that sort of luck luckily help our way out of the deficit in debt? Another thing that could help our way
13:20
Jason Furman: out of the deficit in debt, I'm not at all in favor of this, would be if inflation stays quite high. In some ways, the debt has been rising less than we thought as a share of the economy of the last couple of years, because the economy itself has been inflating itself up. But that's not a way I would recommend dealing with. it.
13:39
Unidentified speaker (speaker_2): You mentioned productivity and I raised this kind of massive question on AI at the beginning, which we probably won't be able to completely resolve. But Democrats, you know, the other thing that they're going to be dealing with is this enormous concern around AI. And I think there's a general feeling that AI is going to feel much more present in people's lives when we get to the the general election in in in twenty 2028. twenty eight. But in the
13:59
Unidentified speaker (speaker_2): meantime, how risky is it to slow down the construction of these data centers? Managing the fact that such a huge amount of investment is now pinned on this AI revolution. How do you think a sensible democratic leader leadership should be responding to these things?
14:19
Jason Furman: Yeah, so people's concerns with AI grew rapidly. People are almost entirely on one side of the issue. and Americans much more negative about AI than people in any other country. It wasn't the main thing people were talking about. But insofar as they were talking about it, they were all saying the same thing.
14:38
Jason Furman: Unfortunately, the data centers is a little bit of a distraction and a sideshow. They'll get built somewhere in the world, some which way, as long as there's an economic return to them, which is itself a question.
14:51
Jason Furman: that we could discuss. But if there's an economic concern, they're going to get built. For me, the the two things are number one, a regulatory framework for safety, and two, an economic framework for economic security. And let me explain those are a little bit different. You know, when it comes to safety, I don't think the
15:11
Jason Furman: right response is saying they can help you build bioweapons with an AI and we're going to give everyone gas masks so that they don't die. We have to stop them from building the bioweapons in the first place. And I think it's sort of madness that we have no federal legislation around transparency, around
15:30
Jason Furman: evaluation, around the ways in which these are deployed. And there's ideas that are out there. When it comes to the economic effects, I actually don't think we can regulate the models. We have no idea how to regulate models so that they add to jobs and help productivity and don't take away jobs and
15:50
Jason Furman: subtract from productivity. I think it's a little bit like the way AI affects college campuses. It helps our students cheat. We can't tell the anthropic to stop helping our students cheat. We have to move our class.
16:03
Jason Furman: our exams, and make them closed, closed book and the like. So when it comes to the economy, it's much more reshaping our economic system. I don't have all the answers to that, but for me, work is at the center of it. Making work more rewarding with things like tax credits, making sure
16:23
Jason Furman: we're beefing up our fiscal system so that we're raising the revenue we're going to. gonna need for our solutions. I don't think we need anything fancy, just old ideas about how you tax capital income and tax high incomes and incomes more broadly should be a part of that. Things like improving our community colleges and job search and placement and things like
16:43
Jason Furman: that have long been on the agenda. I think the evidence for them is quite good. Um, they'll only be better. At universities like mine. we might need to think more about making our education a little bit more vac vocational because there's going to be less incentive to provide people on-the-job training in this world. So I don't know the
17:03
Jason Furman: full economic list. I think almost none of it, though, is things that anthropic or OpenAI could do. Almost all of it is a matter for policymakers and institutions. throughout our economy and society.
17:17
Unidentified speaker (UNKNOWN): Well,
17:18
Unidentified speaker (speaker_2): I wanted to ask you about that. There's been a sort of glib, uh, been at various kind of policy conferences where economists sort of say, oh, that's there's still plenty of demand for macroeconomic analysis and thinking around the impact of AI, and they sort of joke about it. But at the same time, you're one of many economists, but probably the one the one that's most publicly shown their kind of
17:37
Unidentified speaker (speaker_2): living the dream. that AI has become part of your lived reality. And you sort of demonstrated that by releasing a few weeks ago Econvitals. So uh for those listening who don't know, just tell us a little bit about that.
17:51
Unidentified speaker (UNKNOWN): Yeah,
17:51
Jason Furman: it's econvitals.org. It's free and it does macroeconomic tracking and data. Um, it does analysis. You can simulate different um policy changes. You can run the Taylor rule. You can look at trends and inequality. And there's also a set of emails you can sign up for, about 25 a week.
18:11
Jason Furman: And a few of those are written by me, and the rest of them are written by LLMs. Originally, I actually built this for myself. When I was chair of the Council of Economic Advisors, I had about 35 people working for me. And I just had the most amazing data inflow. constant monitoring, updating, memos,
18:30
Jason Furman: analysis coming my way. And also a team of people that could basically answer any question about the economy I had with a combination of theory, data, evidence, other things. And I basically set about recreating that for myself. And then I thought, you know, it's not going to be that hard to make it free and
18:51
Jason Furman: share it with the rest of the world. Turned out to be a little bit harder than I thought, but but I don't really regret it. I don't think it's a major competitor to Bloomberg. If you're like, if your job is like buying and selling stocks at a high frequency, it is probably worth you continuing your Bloomberg subscription. But if you're just a member of the public, you're
19:10
Jason Furman: interested in public policy, you you'd need to sort of pay some attention to the U.S. economy.
19:15
Jason Furman: I think this system does the work of the 20 people would have done. And that's just the economist. The last thing I should say is over 500,000 lines of code. That would have been a lot of coders to build that system. And it was all built by me. And I didn't read a single one of those lines of code. It was all just vibe coded.
19:35
Unidentified speaker (speaker_2): Yeah, I saw that you'd asked Claude how much it would have. taken to build that same website before AI. And it said it would have taken 50 people three years and cost 20 million dollars.
19:47
Jason Furman: Yeah, it would have been a lot of people and a lot of money to build this. So for me, it's amazing. It's augmenting me, it's extending me. I have all sorts of ideas that I didn't have time to pursue before, and I'm pursuing now. And so so in some ways, I think I am personally thinking. better and extending myself better than I have before.
20:07
Jason Furman: But then I also feel somewhat guilty, somewhat bad that I'm working less with research assistants. I still spend a ton of time mentoring students, but I'm not mentoring them nearly as much through this process. And then I wonder at this old place I used to run, the Council of Economic Advisor. Advisors,
20:27
Jason Furman: why two years from now you'd need to hire a junior staffer. And that's painful for me because I myself started as a junior staffer there. And a lot of what I know today I learned by doing the types of numbers and analysis that Claude can just do so much better and faster than a person can these
20:46
Jason Furman: days.
20:55
Unidentified speaker (UNKNOWN): We
20:56
Unidentified speaker (speaker_2): should end on the career thing because it's one of those things that lots of people have are asking the same question in lots of different industries. But these emails are going out. Most of them, the ones that don't have your name on them, are entirely automated. They're not being checked. And you do that as a sort of warning to people. But you know, how often are they wrong? I mean, how much can people
21:14
Unidentified speaker (UNKNOWN): be able to do that?
21:15
Unidentified speaker (speaker_2): rely on them relative to what would have been the kind of regular human error rate?
21:19
Jason Furman: Yeah. So first of all, I I it's really important to me to be transparent. So if my name is on it, I wrote it. And if my name is not on it, the AI wrote it. I'm not pretending otherwise. And there's some people I know that just want to read things with my name on it, don't want to read the other stuff. That's completely fine with me. I think transparency is an important thing here. In terms of the error rate,
21:39
Jason Furman: there have been one or two errors in the system. I have a lot of agents that do fact checking, that do error correcting. So my guess is in terms of raw numbers, it's sort of on par with a very good publication, maybe a little bit worse than Bloomberg, but Bloomberg is the gold standard in this regard. But I also think it's going to
21:59
Jason Furman: get better very, very fast because,
22:02
Unidentified speaker (UNKNOWN): you know, for
22:02
Jason Furman: example, it made a error that involved the way it rounded numbers and it rounded it to two digits and then to one digit. And that would have been different if it went straight from three digits to one digit. That meant something printed as 5.0 instead of 4.9. I caught that, fortunately, and it's never going to make that error again. And so that is the difference. And I think that's the difference in all
22:22
Jason Furman: of these systems, whether it's automated driving or anything else. is it it tends to not be like the person that you explain the problem to. They say yeah, yeah, yeah. And then a month later they sort of slip back into making the problem.
22:39
Unidentified speaker (speaker_2): What are you finding? And this may change every time Claude produces a new model, but what can it do that Fed staff or a previous kind of CEA junior economist couldn't do? but I guess also what can it still not do that you would need to have that bright young thing for?
22:56
Jason Furman: Right. So there's the dimming old thing, which is to say me. Um, I think the things I write are more interesting than the things it writes. And I would rather listen to your podcast than an avatar of you doing a podcast, even
23:13
Unidentified speaker (UNKNOWN): though the
23:13
Jason Furman: avatars are getting sort of more and more realistic looking. They're not as interesting and engaging and and human as you are, Steph. But it's the juniors where I'm more worried. I mean, I just did something that simulated alternative policies to everything from the fiscal response to the financial
23:33
Jason Furman: crisis, to the fiscal response to COVID to the full-scale full scale Russian invasion of Ukraine. And it was a massive amount of modeling that went into it. And it just sort of did all that. And that would have been a couple weeks of work for junior people. And by doing it, they would have gotten
23:53
Jason Furman: to know the Ferbus Federal Reserve's macroeconomic Economic model better. I know that model some. I've worked with it in the past. I've worked with people that have worked with. it. But now from a distance, I can just do just very, very ambitious things of a type that you would have only bit off a piece
24:13
Jason Furman: of before. The problem is I can do that. The person who's doing it isn't there doing it anymore. And I think that has some real costs and downsides to it. Aaron
24:23
Unidentified speaker (UNKNOWN): Powell
24:23
Unidentified speaker (speaker_2): If you spread that around the economy, I mean the thing that people are hitting now is just a cost constraint. Because as Anthropic and others, they're starting to cost quite a lot of money. So
24:32
Jason Furman: I have four Claude Max accounts. And
24:34
Unidentified speaker (UNKNOWN): uh
24:35
Jason Furman: but Harvard is paying for them. So um so that's I guess they're called ultra accounts now. So that's $800.
24:40
Unidentified speaker (UNKNOWN): I
24:41
Unidentified speaker (speaker_2): suspect you're a power user at this point.
24:42
Jason Furman: I'm a power user, sort of switch back and forth as they run out, et cetera. Um but you know, just in the last six months, I it's actually become cheaper. Things happen faster. They use less tokens. Claude does certain things that let you save the amount that you use on it.
25:02
Jason Furman: They kept threatening that they're going to take Fable out or raise the price or lower the limits, and they keep not doing that. And so you're right, spend is going up. But spending on anything is price times. quantity. And the thing that's going up is the quantity. We want to use it much more, not the price. But if anything, I might be
25:22
Jason Furman: on the verge of canceling one of my accounts. I might even be able to cancel a second one. And these were things where three months ago I might use up all four accounts in a given week. And they're just actually getting more efficient. And if anything, falling in price. And for me, that actually is great for users. It raises a question for
25:42
Jason Furman: me about the economics of these Frontier Labs and the amount of competition they face and the degree to which they will or won't be able to monetize some of their innovations.
25:53
Unidentified speaker (UNKNOWN): And
25:53
Unidentified speaker (speaker_2): it actually seems it's slightly the opposite for them because it seems like each model has actually cost them more, not less, as they go as as it becomes more sophisticated. I can take your point about. the quantity of use, but in terms of raising the quality of each model, that seems to be more expensive, not less, each time they do it.
26:11
Jason Furman: Yeah, that's exactly right. And then just to be a little bit granular, Opus 5.5, the latest of the one rung down model from Claude, that is cheaper and faster than the earlier Opus models. So they put a lot of money into it. They made it better and they actually made it cheaper. Now they have Fable
26:30
Jason Furman: 5.2 that came out this week. I haven't spent a lot of time with it, but so far it seems like way more than is needed for the stuff that I do. And a lot of businesses are the same way. They're just not using Fable that much. And because, in some sense, these companies already have cars that go at 300 miles per hour.
26:50
Jason Furman: They put a lot of effort into making them go at 500 miles an hour. Um, if you're a race car driver, you should be buying the 500 mile per hour car. But for the rest of us driving around the city, 300 miles per hour was already a lot faster than what we needed. And that's, I think, the quandary that they're gonna face going forward.
27:08
Unidentified speaker (speaker_2): In this conversation, subconsciously or not, you're sort of it reassuring your fellow economists who maybe don't have quite the illustrious career that you have, but would consider themselves to be serious, established economists. Because you're suggesting that the only problem here is the junior people aren't going to be able to kind of get their foot in the door. But what you've described, I mean, we have
27:27
Unidentified speaker (speaker_2): 60 economists at Bloomberg. I think they won't find it super reassuring because it sounds like you might only need a handful to do all the modelling modeling that each team, current team of like four or five people does. So this it this doesn't seem like just an issue for the first rung on the ladder.
27:46
Unidentified speaker (UNKNOWN): I
27:46
Jason Furman: mean, look, we had um unbelievable PhD economists who were faculty members on leave at the Council of Economic Advisors. And my view is that some of what they did was interpersonal, going to meetings, engaging in the policy process. and you'd still need people for that. But an awful lot of what they did
28:06
Jason Furman: was analysis at a very high and sophisticated level, but it can operate at a very high and sophisticated level. And any reassurance I'm giving you is reassurance about the capabilities in 2026. I mean, we're not talking about 2027 twenty twenty-seven or 2028, twenty twenty-eight. and it can write already.
28:27
Jason Furman: a a quite good economics paper without a whole lot of guidance from economists. Will it be able to get all the way there and fully replace academic economists or the economists at Bloomberg? I think people will want people, a certain amount of trust, someone that puts their name behind something. How
28:46
Jason Furman: much are you going to be willing to pay for that? How much are you going to need that in the future? I confess I'm growing increasingly unsure about that. And some of my being unsure is less about the economic data and economic history and more about the vibes and maybe making the mistake that this time actually could be different.
29:07
Unidentified speaker (UNKNOWN): And
29:07
Unidentified speaker (speaker_2): you mentioned something interesting when you're talking about the Harvard context that you're teaching students walking out into this job market every day.
29:15
Unidentified speaker (speaker_2): And I know from my own experience, my both of my children at college, one of them is just being sort of deluged with sort of negativity around how awful and impossible it's going to be to get a job in a year or two's time. But you said that maybe you should think of it as a more vocational education. What do you mean by that?
29:31
Unidentified speaker (UNKNOWN): Yeah,
29:31
Jason Furman: I mean, the Nobel Prize-winning economist Gary Becker observed that most businesses that train people are not. keeping that employee. So our graduates who go to an investment bank, they'll usually go to that bank for two years and then they'll go someplace else, on to business school, to a private equity firm, whatever
29:51
Jason Furman: it is. And that investment bank is putting a lot of resources into training them. Why do they do that? Because the person's actually willing to work for lower pay, making up numbers here. They might. produce four hundred thousand dollars a year for the investment bank. The investment bank only pays them $250,000. And
30:11
Jason Furman: so they're being charged effectively tuition for training. And you see that throughout our economy. It's not just super high paid jobs like that. In your 20s, you are effectively trading off lower pay for more training so that you can have a better trajectory. in the future.
30:28
Unidentified speaker (UNKNOWN): Well,
30:28
Jason Furman: that only works if you're producing something for the business that's meaningful to them. If you're not producing enough for the business, they're not going to want to just hire you for the training. Moreover, a business model where people actually pay tuition to investment banks or to Bloomberg Economics to get
30:48
Jason Furman: their training, that would be absurd. That's not gonna happen. And so the first thing that I'm most nervous about is not the hiring, but is some of the on-the-job training that goes together with the hiring. And what's one solution to that? It's doing more training at places that can charge people money for the training, which is to say
31:08
Jason Furman: schools. tools. And I think there's other solutions too. You know, I have never really liked non-compete agreements. in labor markets. I'm starting to rethink that a little bit. Maybe you do need some way to keep your employee with you so that in the first year or two, when they're not that productive, they're not doing that much for you, you'll still take the time, effort, and cost
31:28
Jason Furman: because you know they'll be with you later on.
31:31
Unidentified speaker (UNKNOWN): Okay.
31:32
Unidentified speaker (speaker_2): And then the obvious question I'm sure many of your current students wouldn't mind having your old CEA job at some point being the the chief economic advisor to the president. Right now, if they're, you know, 20, 21 years old, they're really good at economics. How should they think about the path to that kind of job?
31:50
Jason Furman: Yeah. I mean, the two things I say are number one, be good at using AI. Our median student at Harvard still uses it like a chat bot and a search engine, and are not building systems of aid. agents to extend them and be that much more ambitious about what the way they approach things. That's thing number one,
32:10
Jason Furman: is be very good at AI. But the second thing is understand what AI does really well, what it doesn't do well, so you can fill in and do those other things. Now, that second piece of advice is a lot easier said than done because the capabilities of AI are constantly. changing and constantly evolving.
32:30
Jason Furman: But figuring out where your edge, your alpha as a human is, and a lot of times that's going to be in interpersonal, in management and creativity, and things that are a little bit harder to teach and learn, but are going to be hard for the machines to remote
32:47
Unidentified speaker (UNKNOWN): I know
32:47
Unidentified speaker (speaker_2): you now have to go off and provide them with some of that alpha. So I won't keep you any longer, but it's fascinating
32:52
Unidentified speaker (UNKNOWN): as
32:52
Unidentified speaker (speaker_2): always, Jason. Thanks so much.
32:54
Jason Furman: Great talking.
32:59
Unidentified speaker (UNKNOWN): Thanks
32:59
Unidentified speaker (speaker_2): for listening to Stefanomics from Bloomberg. It was hosted by me, Stephanie Flanders, and I was joined by Harvard's Jason Furman. Stefanomics was produced by Sama Sadi and Moses Andam, with editorial guidance from Megan O'Neill and Beth Ponsert. Amy Keane is the executive producer for. Talk Podcasts at Bloomberg, and Cheryl Brumley is our head of podcasts.
33:19
Unidentified speaker (speaker_2): Sound design was by Blake Maples and Erin Casper and Kelly Garry. And to help others find this Defonomics show, please rate and review it highly wherever you listen.
33:34
Unidentified speaker (speaker_1): Gain insight on the innovators, disruptors, and tech-driven trends shaping today. today's complex economy. I'm Carol Masser.
33:41
Unidentified speaker (UNKNOWN): And
33:41
Unidentified speaker (speaker_0): I'm Tim Stenevek. Wrap up your workday with the Bloomberg Business Week Daily Podcast.
33:46
Unidentified speaker (speaker_1): We bring you deeper dives into the story shaping your world, from the evolution of AI to the shifting priorities of global business. Plus,
33:53
Unidentified speaker (speaker_0): Silicon Valley Power Players and the latest tech trends.
33:56
Unidentified speaker (speaker_1): Catch up on the conversations you miss during the day.
33:59
Unidentified speaker (speaker_0): Subscribe to the Bloomberg Business Week Daily Podcast on Apple, Spotify, or anywhere you listen.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from Stephanomics

Stephanomics

Reintroducing Stephanomics

Episode discussion: Stephanomics returns to cover global economic shifts beyond US policy

Speakers’ claims: The host introduces the podcast's renewed focus on the global economy's seismic shift driven by inflation, AI, energy shocks, and political threats. The show previously tracked Trump's economic policies under the name 'Trumponics' but is now widening its lens to cover broader global economic forces. Stephanie Flanders, Bloomberg's head of government and economics, outlines the new format. The show will use a global newsroom and interview economists and policymakers to explore big questions reshaping the global economy.

Global EconomyMonetary PolicyGeopolitics
2 min0 assets5 claims to verify
Stephanomics

The Midterms Changed Washington. Will They Change the Economy?

Episode discussion: Midterm results signal voter dissatisfaction with affordability, constraining Trump's economic agenda and shaping future GOP/DEM dynamics.

Speakers’ claims: Host Stephanie Flanders discusses the Democratic midterm victory with guests Oren Cass and Josh Wingrave, framing the win as a protest against the cost of living rather than a policy mandate. Cass argues that traditional metrics mask a deep 'affordability disconnect,' while Wingrave notes Trump faces constraints from expiring tax cuts and political backlash. The speakers analyze the persistence of tariffs, the structural limits on presidential power over the Federal Reserve, and the evolving ideological factions within both major parties.

Midterm ElectionsAffordability CrisisTariffs
36 min0 assets17 claims to verify
All insights