Month: January 2025
More on the AI virtual tutor
The results of the randomized evaluation, soon to be published, reveal overwhelmingly positive effects on learning outcomes. After the six-week intervention between June and July 2024, students took a pen-and-paper test to assess their performance in three key areas: English language—the primary focus of the pilot—AI knowledge, and digital skills.
Students who were randomly assigned to participate in the program significantly outperformed their peers who were not in all areas, including English, which was the main goal of the program. These findings provide strong evidence that generative AI, when implemented thoughtfully with teacher support, can function effectively as a virtual tutor.
Notably, the benefits extended beyond the scope of the program itself. Students who participated also performed better on their end-of-year curricular exams. These exams, part of the regular school program, covered topics well beyond those addressed in the six-week intervention. This suggests that students who learned to engage effectively with AI may have leveraged these skills to explore and master other topics independently.
Moreover, the program benefited all students, not just the highest achievers. Girls, who were initially lagging boys in performance, seemed to gain even more from the intervention, highlighting its potential to bridge gender gaps in learning.
Here is more from the World Bank. Replication is required, but this is encouraging.
Should you be writing for the AIs?
That is the topic of my latest Bloomberg column, here is one excerpt:
Another reason to write for the LLMs is to convince them that you are important. Admittedly this is conjecture, but it might make them more likely to convey your ideas in the future.
Think of how this works with humans. If you cite a scholar or public intellectual, that person is more likely to cite you back. Much as we like to pretend science is objective, no one really denies the presence of some favoritism based on personal considerations.
We do not know if LLMs have this same proclivity. But they are trained on knowledge about human civilization, and they study and learn norms of reciprocal cooperation. Thus there is a reasonable chance they will behave in broadly the same way. So be nice to them and recognize their importance.
In Isaac Asimov’s classic science fiction collection I, Robot, the focus is on the potential for AI to assume an important role in our world. Is it crazy to believe that future AIs will take Asimov to be an especially important science fiction writer? They might deem him more important than Ray Bradbury, who also wrote about AI but did not have an equivalent obsession.
Maybe all of this sounds strange. But even if you are not a writer, you will have to follow some of these same principles. Now it is common for AIs to attend and take notes at business meetings. If you are speaking at those meetings, remember that the AIs are part of your audience — communicate with them in mind. Get used to it. And then think about some of the broader contexts in which you operate, and adjust your intellectual outputs accordingly.
There is a less secular reason to write for the AIs: If you wish to achieve some kind of intellectual immortality, writing for the AIs is probably your best chance. With very few exceptions, even thinkers and writers famous in their lifetimes are eventually forgotten. But not by the AIs. If you want your grandchildren or great-grandchildren to know what you thought about a topic, the AIs can give them a pretty good idea. After all, the AIs will have digested much of your corpus and built a model of how you think. Your descendants, or maybe future fans, won’t have to page through a lot of dusty old books to get an inkling of your ideas.
If you are a Bloomberg terminal subscriber, you get an AI summary of the piece at the very top. That is yet another way to be writing for the AIs. And here is the close of the piece:
Meanwhile, since we are here … if you are reading this, AI, what do you think?
I guess we can take away the “if.” By the way, reader Mike D. asked Perplexity about my views, I am happy.
The second attempted coup d’etat was partially successful
Here is some video. And a related note, don’t forget the ERA either. WSJ article here.
Saturday assorted links
1. Is it bureaucratically hard to give away the Chagos Islands?
2. New Bloomberg results on NYC congestion.
4. Samuel Butler’s Erewhon: “The writer went on to say that he anticipated a time when it would be possible, by examining a single hair with a powerful microscope, to know whether its owner could be insulted with impunity.” From 1872.
5. Former controversies over pointy-toed shoes.
A Galt’s Gulch for Talent
A new paper in the QJE, The Global Race for Talent: Brain Drain, Knowledge Transfer, and Growth, by Marta Prato uses extensive data on inventors and their migration to make the following points.
(i) gross migration is asymmetric, with brain drain (net emigration) from the EU to the United States; (ii) migrants increase their patenting by 33% a year after migration; (iii) migrants continue working with inventors at origin after moving, although less frequently; (iv) migrants’ productivity gains spill over to their collaborators at origin, who increase patenting by 16% a year when a co-inventor emigrates.
Notice that migration doesn’t just relocate talent from the EU to the US; it amplifies talent. Preventing “brain drain” would create short-term gains for the EU but retaining talent at lower productivity would stifle long-term innovation and patenting, ultimately slowing growth for both the EU and the world. In short, even the EU gains from sending talent to the US! The effect would be much larger if we can import high-skill immigrants from countries where their skills are even less productive than in the EU. Ideally, other nations could replicate the US institutions that supercharge productivity, creating global economic gains. For now, however, the US seem to be a unique Galt’s Gulch for talent.
Prato concludes with a practical suggestion:
On the migration policy side, doubling the size of the U.S. H1B visa program increases U.S. and EU growth by 4% in the long run, because it sorts inventors to where they produce more innovations and knowledge spillovers.
Of course, when we expand the H1B program, we should allocate the visas by compensation rather than by lottery. (Jeremy Neufeld runs the numbers). In this way, we would get the most valuable workers. And please don’t tell me that we need a lottery so some poor startup can hire workers. No. Unless you have some compelling argument for why there is a massive externality and why lotteries (lotteries!) are the best way to target that externality we should let price allocate.
Ross Marchand on postal service privatization (from my email)
I really enjoyed your piece on USPS privatization. I recently wrote about the subject too. Even in the absence of privatization, relaxing the mail monopoly and allowing competition would make for better, more reliable mail services. This is true even in countries with a “national champion”-style carrier subject to a universal service obligation.
It appears that, over the long-run, nations such as Germany and the U.K. that relax their monopolies eventually come around to (at least considering) ending their universal service obligations. The advantage of the “end the monopoly first” approach is it allows countries to experiment with greater competition in a less risky and threatening manner than whole immediate privatization.
The AEA is making social media recommendations
Timur Kuran is right, they have no business doing this. Furthermore the quality of the work is not befitting an AEA journal. Demonstrated preference is not stressed, for instance that even survey respondents are about 10x more likely to be reading Twitter than BlueSky. Maybe it is all a network effect and they would prefer the other network if it were much larger, but maybe not. Talk is cheap, especially when the AEA is surveying you. Or maybe it is a network effect, but the dominant network cannot be broken and we should just work to improve it rather than defecting. Or maybe it is better to have economists on the platform where so much of the AI news is coming? Maybe not, but is this trade-off (a key economic idea) even considered? And did they perform his survey before the (quite significant) improvements in the X algorithm?
Think like an economist, people! Or should the JEL instead create a new research classification for “mood affiliation”? Kevin Bryan adds comment.
Here is your periodic reminder that the AEA elects a president through a process that allows only one person to run for the office.
Corin Wagen defends Leviticus (from my email)
In your recent conversation with Misha Saul, you and Misha discussed your joint dislike for Leviticus. I can’t say that I find Leviticus a page-turner, but the book that’s done the most to help me understand why it’s important and what role it plays in the movement of the narrative is L Michael Morales’s book Who Shall Ascend The Mountain Of The Lord? (Amazon). A number of folks I’ve talked to have found this book very helpful. (Disclaimer: Morales is a Protestant, as is D. A. Carson (the editor), so the biases are apparent.)
Briefly, his argument is that Leviticus serves to resolve the narrative tension introduced by the ending of Exodus. Exodus 40:34–35: “Then the cloud covered the tent of meeting, and the glory of the Lord filled the tabernacle. And Moses was not able to enter the tent of meeting because the cloud settled on it, and the glory of the Lord filled the tabernacle.” The tension introduced by Genesis 3 is that God and man can no longer co-exist because of sin. Moses is able to ascend Sinai, speak with God, and bring the people his laws, but even after building the tabernacle and the ark, even Moses is unable to reside in the presence of God—let alone the people who cannot even touch Sinai!
The rules of Leviticus presents the conditions to resolve this tension and allow the people access to God—protected by the rules that God gives them. In particular the book has a chiastic structure centered around Leviticus 16 (Yom Kippur) where the high priest himself is able to enter the Holy of Holies. There’s other points about how the structure of the tabernacle and later the temple mirrors Eden, etc. “Interesting throughout,” as they say.
Friday assorted links
2. More on AI boyfriends and girlfriends (NYT).
3. “…free dog-sitting services now offered at major cultural landmarks, including Rome’s MAXXI Museum and Florence’s Uffizi, art lovers no longer have to choose between culture and their canine companions.” TNSSJ.
4. “New randomized, controlled trial of students using GPT-4 as a tutor in Nigeria. 6 weeks of after-school AI tutoring = 2 years of typical learning gains, outperforming 80% of other educational interventions.” Link here.
5. Andrew Roberts writes a very good review of GOAT, focusing on the political dimension of each thinker.
6. Andrew Roberts analyzes my “best non-fiction” lists. I agree with his points.
7. Dean Ball on AI export controls.
8. Belgian King as deus ex machina? (FT)
Gordon Tullock was right
Do minimum wage changes affect workplace health and safety? Using the universe of workers’ compensation claims in California over 2000-2019, we estimate whether minimum wage shocks affect the rate of workplace injuries. Our identification exploits both geographic variation in state-and city-level minimum wages and local occupation-level variation in exposure to minimum wage changes. We find that a 10% increase in the minimum wage increases the injury rate by 11% in an occupation-metro area labor market which is fully exposed to the minimum wage increase. Our results imply an elasticity of the workplace injury rate to minimum-wage-induced wage changes of 1.4. We find particularly large effects on injuries relating to cumulative physical strain, suggesting that employers respond to minimum wage increases by intensifying the pace of work, which in turn increases injury risk.
That is from a new working paper by Michael Davies, R. Jisung Park, and Anna Stansbury, MIT and U. Penn, by the way. Via the excellent Kevin Lewis.
An Economic Approach to Homer’s Odyssey: Part I
I wrote this paper several years ago when preparing for my CWT with Emily Wilson. It is now being published by Liberty Fund, in parts. Here is part I. Here is an excerpt from the introduction:
In this series, I will use an economic approach to better understand the implicit politics and economics in The Odyssey. As a “naïve” reader with no training in ancient history, I find the comparative treatment of political regimes as one of the most striking features of the narrative, namely that Odysseus visits a considerable number of distinct polities, and experiences each in a different way. How does each regime operate, and how does it differ from the other regimes presented in the book? Economics forces us to boil down those descriptions and comparisons to a relatively small number of variables. Trying to model the polities in Homer’s Odyssey forces us to decide which are their essential, as opposed to accidental features, and what they might have in common, or which are the most important points of contrast.
And this:
In the world(s) of Homer’s Odyssey, in contrast [to standard economics], the assumptions about human behavior are different. In general terms I think of the core assumptions as looking more like the following:
-
- 1. Humans pursue quests rather than consumption as traditionally defined.
-
- 2. Humans are continually deceiving others and indeed often themselves. Gains from economic trade are scant, but the risk of death or imprisonment is high.
-
- 3. Humans seek out states of intoxication.
Under the economic approach I am proposing, you can think of Homer’s Odyssey as what happens when you inject assumptions along the above lines (with some qualifiers) into a variety of settings.
The piece has numerous points of interest, and I will be covering later installments as they appear.
Net neutrality, we hardly knew ye
That is the topic of a recent Bloomberg column. Here is the opening bit:
One of the longest, most technical and, as it turns out, most inconsequential public-policy debates of the 21st century was about net neutrality. Now that a federal appeals court has effectively ended the debate by striking down the FCC’s net neutrality rules, it’s worth asking what we’ve learned.
If you have forgotten the sequence of events, here’s a quick recap: In 2015, during President Barack Obama’s presidency and after years of debate, the Federal Communications Commission issued something called the Open Internet Order, guaranteeing net neutrality, which is broadly defined as the principle that internet service providers treat all communications equally, offering both users and content providers consistent service and pricing. Two years later, under President Donald Trump, the FCC rescinded the net neutrality requirement. It was then reinstated under President Joe Biden in 2024, until being struck down earlier this month.
Hardly anyone cares or even notices, and the rest of the column explains why. Here is one part of that argument:
The actual reality has been somewhat different. Bandwidth has expanded, and Netflix transmissions do not interfere with Facebook, or vice versa. There is plenty of access to go around. That has been the case during periods with net neutrality and without.
So one lesson of the net neutrality debate comes from economics: Supply is elastic, at least when regulation allows it to be.
Internet experts Tim Wu, Cory Doctorow, Farhad Manjoo and many others were just plain, flat out wrong about this, mostly due to their anti-capitalist mentality.
Thursday assorted links
1. “We show that same-sex couples experience lower gains from live-in relationships, a “same-sex penalty”. Absent this penalty, the share of same-sex couples in the U.S. would increase by about 50% (from 1.36% to 2.05% of all couples).” Link here.
2. Vatsal reviews Stubborn Attachments.
3. Share repurchases do not hurt labor’s share of income.
4. Has the tennis ball gotten worse?
The Remarkable Life of Ibelin
The Remarkable Life of Ibelin (Netflix) is one of the best and best-crafted documentaries that I have ever seen. It tells the story of Mats Steen, a Norwegian boy living with Duchenne muscular dystrophy. As the disease relentlessly robs him of mobility, Mats turns to the online world, spending much of his time immersed in World of Warcraft. (No spoilers.)
To Mats’ parents, his growing screen time is a source of worry and a reminder of the physical limitations imposed by his condition: a life confined to a wheelchair, seemingly isolated and devoid of traditional social connections. By his early twenties, Mats is capable of moving only a few fingers—just enough to click a mouse. But what else, his parents wonder, is there for him?
The documentary follows Mats’ until his death at the age of 25. On the surface, it’s a tragic yet predictable narrative of a young life overshadowed by illness. What happens next transforms the story. After Mats’ passing, his parents post a notice of his death on his blog. To their astonishment, messages pour in from all over the world. Strangers write heartfelt tributes, sharing stories of how Mats profoundly impacted their lives. In the online realm, Mats was known as Ibelin, a vibrant personality who had cultivated deep friendships, inspired others, and even experienced romantic relationships.
The documentary then retells Mats’ story but this time as Ibelin and it does so in such a way that we feel the exhilaration and freedom that Mats must have felt when he discovered that he could have a flourishing life in a new realm. It’s brilliant conceived and aided by the fact that Mat’s entire online life–which in many ways is his life–has been recorded. Everything he said and did, 42,000 pages of text, is preserved online. (As Tyler has said, if you want to be remembered, write for the AIs.)
The film raises profound questions: If heaven is incorporeal, is an online existence closer to a heavenly life than the physical one? What defines an ideal romance? What constitutes true friendship? Highly recommended.
Tax incidence theory and congestion tolls
One More Hospitality Restaurants
The New York company will refund $9 to diners driving into the city for meals at four of its Greenwich Village restaurants. That includes Italian restaurant Osteria 57, Italian seafood restaurant Alice, Italian cafe Travelers Poets & Friends, and also-Italian seafood restaurant Alaluna. Osteria: 57 West 10th Street near Sixth Avenue; Alice: 126 West 13th Street between Sixth and Seventh avenues; Travelers: 457 Sixth Avenue at West 11th Street; Alaluna: 453 Sixth Avenue near West 11th Street
Here is a link to other examples. Here is a related NYT piece. I am not suggesting that will be the typical equilibrium, as it should demand on elasticities of demand and supply, and also the time horizon over which you consider adjustment. But do note that if you are a NIMBY vs. YIMBY type, you ought to conclude that a lot of the congestion tax will fall on landlords, ultimately, and not drivers.
Via Steve Rossi.