How to think about AI progress
The Zvi has a good survey post on what is going on with the actual evidence. I have a more general point to make, which I am drawing from my background in Austrian capital theory.
There are easy projects, and there are hard projects. You might also say short-term vs. long-term investments.
The easier, shorter-term projects get done first. For instance, the best LLMs now have near-perfect answers for a wide range of queries. Those answers will not be getting much better, though they may be integrated into different services in higher productivity ways.
Those improvements will yield an ongoing stream of benefits, but you will not see much incremental progress in the underlying models themselves. Ten years from now, the word “strawberry” still will have three r’s, and the LLMs still will tell us that. There are other questions, such as “what is the meaning of life?” where the AI answers also will not get much better. I do not mean that statement as AI pessimism, rather the answers can only get so good because the question is not ideally specified in the first place.
Then there are the very difficult concrete problems, such as in the biosciences or with math olympiad problems, and so on. Progress in these areas seems quite steady and I would call it impressive. But it will take quite a few years before that progress is turned into improvements in daily life. Again, that does not have to be AI pessimism. Just look at how we run our clinical trials, or how long the FDA approval process takes for new drugs, or how many people are reluctant to accept beneficial vaccines. I predict that AI will not speed up those processes nearly as much as it ideally might.
So the AI world before us is rather rapidly being bifurcated into two sectors:
a) progress already is extreme, and is hard to improve upon, and
b) progress is ongoing, but will take a long time to be visible to actual users and consumers
And so people will complain that AI progress is failing us, but mostly they will be wrong. They will be the victim of cognitive error and biases. The reality is that progress is continuing apace, but it swallows up and renders ordinary some of its more visible successes. What is left behind for future progress can be pretty slow.
The politics of depression in young adults
From a recent paper by Catherine Gimbrone, et.al.:
From 2005 to 2018, 19.8% of students identified as liberal and 18.1% identified as conservative, with little change over time. Depressive affect (DA) scores increased for all adolescents after 2010, but increases were most pronounced for female liberal adolescents (b for interaction = 0.17, 95% CI: 0.01, 0.32), and scores were highest overall for female liberal adolescents with low parental education (Mean DA 2010: 2.02, SD 0.81/2018: 2.75, SD 0.92). Findings were consistent across multiple internalizing symptoms outcomes. Trends in adolescent internalizing symptoms diverged by political beliefs, sex, and parental education over time, with female liberal adolescents experiencing the largest increases in depressive symptoms, especially in the context of demographic risk factors including parental education.
Here is the link. This is further evidence for what is by now a well-known proposition.
Wednesday assorted links
1. Flora Yuknovich, painter (NYT).
2. Further comments on Milei and Argentina.
3. My TA Zixuan Ma is starting a blog on China and also recommends these six books.
4. How is New College of Florida doing?
5. Some new Substacks from economics graduate students.
6. Machine learning for economists. And double descent and econometrics.
7. Flood the zone with AI-generated podcasts? “I think that people who are still referring to all AI-generated content as AI slop are probably lazy luddites.” I still think people do not really want this, but I suppose we will see.
8. Roland Fryer on education reform (WSJ).
AI-led job interviews
We study the impact of replacing human recruiters with AI voice agents to conduct job interviews. Partnering with a recruitment firm, we conducted a natural field experiment in which 70,000 applicants were randomly assigned to be interviewed by human recruiters, AI voice agents, or given a choice between the two. In all three conditions, human recruiters evaluated interviews and made hiring decisions based on applicants’ performance in the interview and a standardized test. Contrary to the forecasts of professional recruiters, we find that AI-led interviews increase job offers by 12%, job starts by 18%, and 30-day retention by 17% among all applicants. Applicants accept job offers with a similar likelihood and rate interview, as well as recruiter quality, similarly in a customer experience survey. When offered the choice, 78% of applicants choose the AI recruiter, and we find evidence that applicants with lower test scores are more likely to choose AI. Analyzing interview transcripts reveals that AI-led interviews elicit more hiring-relevant information from applicants compared to human-led interviews. Recruiters score the interview performance of AI-interviewed applicants higher, but place greater weight on standardized tests in their hiring decisions. Overall, we provide evidence that AI can match human recruiters in conducting job interviews while preserving applicants’ satisfaction and firm operations.
That is from a new paper by Brian Jabarian and Luca Henkel.
The evolution of the economics job market
In the halcyon days of 2015-19, openings on the economics job market hovered at around 1900 per year. In 2020, Covid was a major shock, but the market bounced back quickly in 2021 and 2022. Since then, though, the market has clearly been in a funk. 2023, my job market year, saw a sudden dip in postings. 2024 was even worse, with openings falling 16% lower than the 2015-19 average.
At the time, the sudden fall in 2023 seemed mysterious—it was an otherwise healthy year for the broader labor market. In hindsight, it seems like the 2021-22 recovery masked some underlying weakness. The 2020 job market had 500 fewer openings than the 2014-19 average; 2021 and 2022 together produced only around 100 more jobs than the 2014-19 average. In other words, the recovery never made up for the pandemic; by this crude logic, around 400 economist jobs were “destroyed”.
…And of course, all of this decline occurred before the litany of disasters that have recently hit the Econ job market. In May, Jerome Powell announced that the Federal Reserve—perhaps the largest employer of economists in America—would cut its workforce by 10%. The federal government has frozen hiring, as has the World Bank. Hit by the dual threat of fines and looming cuts to federal funding, Harvard, MIT, the University of Washington, Notre Dame, Northwestern University, among others, have announced hiring freezes and budget cuts.
Here is more from Oliver Kim, who also offers a much broader discussion of the meaning of all this.
UAP okie-dokie
The real action starts at 0:54:
Here is the same video with under oath reactions from witness experts. Here are remarks from Dylan Borland.
Tuesday assorted links
1. The Islamic argument for competence.
2. Worry claims about Egypt and Israel.
3. PEPFAR will distribute Gilead’s new anti-HIV drug.
4. A quick U.S. geography lesson.
5. Results from Italian experiments with tenure.
6. New results on the fiscal impact of immigration.
7. Observations on redistricting (New Yorker).
9. New results on TRAPPIST-1e. Only about forty light years away.
Housing 101
John Arnold points us to this table on new apartments and pointedly notes that the population of LA (18.5 m) is more than 7 times that of Austin (2.5m).
MR readers will not be surprised to learn that apartment prices are falling in Austin.
Meanwhile the WSJ reports another shocker, New York’s Airbnb Crackdown, in Force for Two Years, Hasn’t Improved Housing Supply. But guess what has happened? Ok, you don’t have to guess. Hotel prices have increased:
Hotels tend to benefit from tighter Airbnb restrictions, especially in New York City. Significantly reducing the number of apartments that can be rented for less than 30 days undeniably boosts demand for hotel rooms in a city visited by tens of millions of tourists a year.
Without the law, “we would be in a catastrophic situation,” said hotelier Richard Born, who owns 24 hotels across the city.
The “Marvel Universe” of faith
In a recent video posted to the AI Bible’s Youtube channel, buildings crumble and terrified-looking people claw their way through the rubble. Horns blare, and an angel appears floating above the chaos. Then come monsters, including a seven-headed dragon that looks like something out of a Dungeons and Dragons rulebook.
The visuals in this eight-minute video, which depicts a section of the Book of Revelation, are entirely generated by artificial intelligence tools. At times it feels like a high-budget Hollywood movie, at times more like a scene from a video game, and at times like fantasy art. Despite the somewhat muddled visual styles, viewers seem to like what they see – it has racked up over 750,000 views in the two months since it was posted.
The viewers are mostly under 30, and skew male.
Here is the full story, via Ari Armstrong.
One look at negative emotional contagion
This paper studies how peers’ genetic predisposition to depression affects own mental health during adolescence and early adulthood using data from the National Longitudinal Study of Adolescent to Adult Health (Add Health). I exploit variation within schools and across grades in same-gender grademates’ average polygenic score—a linear index of genetic variants—for major depressive disorder (the MDD score). An increase in peers’ genetic risk for depression has immediate negative impacts on own mental health. A one standard deviation increase in same-gender grademates’ average MDD score significantly increases the probability of being depressed by 1.9 and 3.8 percentage points for adolescent girls (a 7.2% increase) and boys (a 25% increase), respectively. The effects persist into adulthood for females, but not males. I explore several potential mechanisms underlying the effects and find that an increase in peers’ genetic risk for depression in adolescence worsens friendship, increases substance use, and leads to lower socioeconomic status. These effects are stronger for females than males. Overall, the results suggest that there are important social-genetic effects in the context of mental health.
That is from a recent paper by Yeongmi Jeong, via the excellent Kevin Lewis.
Facts about Norway
But some worry that a “Norwegian disease” is developing through the use of an ever-increasing withdrawal from the fund each year. That amount — which reached NKr542bn ($54bn) this year — amounts to about a quarter of the government budget.
This year, it helped Norway boost contributions to Ukraine without having to cut spending elsewhere or raise taxes.
Spending on sickness and disability is the highest in the OECD group of rich countries, and four times the average. High-school dropout rates are well above the European average. Meanwhile, productivity growth has slowed, worrying policymakers.
Here is more from Richard Milne in the FT.
Monday assorted links
1. How Arnold Kling reads with AI.
2. Criticism of the Singaporean educational system.
3. Patrick Collison on the uses of crypto. And Matt Huang on permissionless.
4. GPT-5 on Huemer and immortality. And more.
5. How about Chinese stablecoins? (NYT)
6. Milei’s party loses in the elections.
7. How much do AIs push back against psychosis?
8. The Oakland B’s will experiment with being managed by AI (NYT).
9. Good Khodorkovsky thread on Ukraine, Russia, and the war.
Patrick Collison on the Irish Enlightenment
Most of all, the Irish Enlightenment seems to me an instance of small group theory. I’m fond of the thought that between great man and structuralist theories of history there lies an intermediate position: the small group, a colocated cauldron for iconoclastic thinking, can as a collective pioneer a novel direction. The romantics in Jena, the founders of Silicon Valley, the musicians behind punk. Unsurprisingly, the early Irish thinkers are closely connected. Swift and Berkeley attended the same school and were good friends. Hutcheson and Berkeley debated publicly, while Burke’s work is clearly downstream of Hutcheson’s.
And this:
How should we view the movement as a whole? Well, the timing is important: Cantillon published his Essai in 1755, Swift Drapier’s Letters in 1724, and Berkeley The Querist in 1735. It seems to me that, before 1750, the Irish thinkers have a strong claim to leading the world in the field of economics and to having collectively sketched out much of the core of the field in broadly correct terms. In Petty you have economic statistics; in Cantillon you have risk, market pricing, and much else; in Berkeley, you have a theory of national banking plus development economics; in Swift you have proto-monetarism. The claim is not that they figured everything out or were right on all points, but which other school or group could you rank ahead of them? Smith published Wealth of Nations in 1776 and The French physiocrats, who were very important, came later: Quesnay’s first piece wasn’t published until 1756.
Do read the whole short essay.
France fact of the day
If nothing is done, interest payments will become the biggest expense in the French budget in four years, Mr. Bayrou has warned.
Here is the full NYT article, noting that government spending is 57 percent of the economy and the French have the longest financed retirements ever seen in the history of the world.
Eli Dourado on trains and abundance
One thing I got a bit of crap for in the hallways of the Abundance conference is my not infrequent mockery of trains on Twitter. I’m sorry, trains are not an abundance technology. I think many people in the abundance scene like trains because:
1. America’s inability to build HSR is the leading example of low state capacity, and we all more or less agree that state capacity is a tenet of the abundance agenda.
2. Trains have high transport efficiency, and people coming to abundance out of the climate movement can’t shake their old habits of caring about energy efficiency ahead of other considerations.
Obviously if we spend billions of dollars on high-speed rail, there should at least be some high-speed rail service. But a deeper element of state capacity is not picking dumb things for the state to build in the first place. And trains are a dumb thing to build in the 21st century.
A true transportation abundance agenda has to revolve around airplanes and autonomous vehicles. The goal should be able to go from any point in the country to any other point in the country in, like, two hours, door to door.
We should have supersonic airplanes made out of cheap titanium and powered by electro-LCH4. An autonomous vehicle should be available to pick you up within 30 seconds and whisk you to a nearby airfield. Security should be painless and instant (another state capacity task). If your trip doesn’t require an airplane, the autonomous vehicle should get you straight there at 100+ mph since it’s good at avoiding accidents. In cities, autonomous buses with dynamic route planning based on riders’ actual needs beat subways’ 1-dimensional tracks.
We should not be trying to build marginally better versions of 20th century (or 19th century!) technology. We should be more ambitious than that. Trains are unbefitting of a country as wealthy as I aspire for us to be.
Please join the anti-train faction of the abundance movement.
Here is the link to the tweet.