How to teach people how to work with AI
I have a very concrete and specific proposal for teaching people how to work with AI. It is also a proposal for how to reform higher education.
Given them some topics to investigate, and have them run a variety of questions, exercises, programming, paper-writing tasks — whatever — through the second or third-best model, or some combination of slightly lesser models.
Have the students grade and correct the outputs of those models. The key is to figure out where the AIs are going wrong.
Then have the best model grade the grading of the students. The professor occasionally may be asked to contribute here as well, depending on how good the models are in absolute terms.
In essence, the students are learning how to grade and correct AI models.
Just keep on doing this.
Of course the students need to learn subject matter as well, but perhaps this process should be what…one-third of all higher education?
You might think the models are too good for human grading and correction to matter at all, whether now or in the future. That may be true at some point. That is the same scenario, however, where it does not matter what we teach the students. Might as well teach them for the world-states in which they are of some use at all.
This is all apart from the PhD you might get in gardening, but even there I think you will need to spend some time learning how to correct and judge AI models.
Income inequality in the Nordics
Our analysis suggests that income equality in the Nordics is primarily driven by a significant compression of hourly wages, reducing the returns to labor market skills and education. This appears to be achieved through a wage bargaining system characterized by strong coordination both within and across industries. This finding contrasts with other commonly cited explanations for Nordic income equality, such as redistribution through the tax-transfer system, public spending on goods that complement employment, and public policies aimed at equalizing skills and human capital distribution.
That is from a new NBER working paper by Magne Mogstad, Kjell G. Salvanes, and Gaute Torsvik.
Thursday assorted links
Steel Tariffs in Two Pictures
Recall Principle 2 of Three Simple Principles of Trade Policy, Businesses are Consumers Too. Case in point, steel. Justin Wolfers summarizes an analysis of Trump’s 2018 steel tariffs:
Going back further we have a good analysis from Lydia Cox of the Bush steel tariffs. Even though the tariffs were temporary, they led to a rearrangement of supply chains which led to long-lasting declines in exports and employment in steel using industries.
Place Effects on Fertility Decision: Evidence from Mover Design
This paper investigates the causal impact of place-based factors on fertility decision using mover design and data from the Panel Study of Income Dynamics (1968-2019). We find that moving to a state with a 1 percentage point higher birth rate increases the probability of childbirth by 0.9 percentage points, with cumulative effects reaching 3.8 percentage points three years post-move. The response demonstrates concentration among first births and exhibits systematic variation across demographic characteristics—with particularly pronounced effects observed among white women who are married, younger, and have higher income levels. Our variance decomposition shows the contribution of place effects to fertility variance increased from 4.7 percent to 26.0 percent before and after the Great Recession, with geographical variation in contraceptive access and healthcare infrastructure showing the strongest correlations with these place effects. This research emphasizes the importance of considering contextual factors in fertility research and policy interventions.
That is from a new paper by Hantao Wu and Man Zhu. Via the excellent Kevin Lewis.
Will stablecoins herald a broader-based dollarization?
That is the topic of my latest Bloomberg column. Here is one excerpt:
Stablecoins are programmable crypto assets that promise conversion into some currency, typically US dollars. Currently, they are the fastest-growing sector of crypto. Stablecoin usage is up 84% since August 2023 and is now at a peak of $224 billion. The sympathetic stance of President Donald Trump’s administration toward crypto is likely to help growth further.
It is noteworthy that, measured by market capitalization, perhaps as much as 99% of stablecoins are denominated in dollars. That is a much higher share than is found in standard international trade and finance. This shows that, if monetary institutions were started all over again from scratch — which is part of what crypto is doing — the market would opt largely for dollars.
Traders in many less well governed countries want to partake in dollar-based economies, but they do not always have ready access to dollar-based banking in the way that Americans do. Their domestic banking systems may be unreliable or be regulated to discourage dollar dominance. Traders may also be afraid of US regulations, which operate through sanctions and restrictions on dollar-denominated transactions. “Know your customer” regulations, for example, can make interacting with US financial institutions very costly.
So foreigners are increasingly turning to stablecoins, which they can access quickly and directly through apps. Stablecoins are not much regulated now, but eventually regulation will emerge in many countries. That said, there will likely be more and less regulated versions of stablecoins for a long time to come. For that reason, another advantage of stablecoins — at least for individual traders, if not always for broader society — is that traders will be able to choose the level of regulation they desire.
Several countries have dollarized in recent years — including Panama, Ecuador and El Salvador — and none seems to regret it. President Javier Milei in Argentina has pledged dollarization, and that proved to be a popular campaign promise, although it remains to be seen if he can summon the resources to pull it off.
The simplest scenario is that people outside developed nations use stablecoins more and more. Their economies will become partly “dollarized” — or if you wish to use an even less elegant term, “stablecoinized,” with the stablecoins backed by dollars. People in those economies will get more used to thinking and calculating in terms of dollars, even for their domestic transactions. Dual-currency economies may become more common, with both a domestic currency and dollar-backed stablecoins. Over time, fearing the redemption risks associated with stablecoins, many nations will opt for outright dollarization, either full or partial. In some cases, the dollar might end up predominating.
Worth a ponder, and this is yet another scenario where crypto proves useful.
Wednesday assorted links
Why it is hard for the Executive to disobey the judiciary
An excellent thread from Dilan Esper. Excerpt:
I have pointed out already that Elon Musk has massive economic interests in cases currently before the federal courts. That is reason enough he would have to obey an order or resign if Trump demanded he not do so. A federal court could default him on those suits.
But more generally federal courts can issue writs that can be levied on bank accounts and properties. These could effectively freeze the assets of ANY noncompliant government official. Yes, even the President. And no, President Trump couldn’t order then unfrozen.
There is more at the link. Retweeted by Alex T.
The danger of Trump disobeying court orders
Ilya Somin covers this question over at Volokh Conspiracy. I receive many queries about this, some of them panicky and anguished. I haven’t covered it, mostly because I don’t feel I have enough insights into the relevant matters of constitutional law, or for that matter what is going on inside the administration (for instance, how should one interpret those Vance tweets?)
I can tell you what I would find useful. If you are especially pessimistic on this front, which are the securities prices that would indicate an actual constitutional problem? Particular equities? Interest rates? The value of the dollar? Measures of volatility? Something else? Don’t restrict yourself to the absolute level of share prices, surely there are favored and disfavored companies and sectors, right?
I am allergic to the view that “fascism could come and market prices would not even budge.” In fact, I think it is extremely skeptical and subversive of democracy, or shall I better say a constitutional republic. I think fascism, or a constitutional collapse, would be a terrible outcome in a variety of very practical ways, in addition to its moral failings. At the very least it would matter for many particular enterprises.
In a variety of other contexts, such as tariffs, market prices have been super-sensitive to the actions of the Trump administration. So people, on this question, which exactly are the measurable, market price indicators? After all, you don’t want to be like those doomster AI skeptics who think no one can trade on the (supposed) truth.
In the comments section, I am not interested in your blah blah blah opinion full of adjectives. Just tell me which prices please. I do see this issue as constituting a real risk, if perhaps a sometimes exaggerated one. So I will follow those market prices with great interest. I just need to know what they are.
Addendum: In an excellent Substack today Matt Yglesias notes: “Republicans, meanwhile, are making very little forward progress on their legislative agenda.”
What should I ask Austan Goolsbee?
Yes I will be doing a Conversation with him. So what should I ask? Here is his Wikipedia page. Here is Austan on Twitter. Here is scholar.google.com. Note that Austan is now president of the Chicago Fed. So what should I ask him?
The three levels of AI understanding
If you are trying to contextualize someone’s opinion on current AI, I suggest asking three basic questions about their perspective. In particular, you should know which of the following three realities they are aware of. Here goes:
1. How good are the best models today?
Most people do not know this, even if you are speaking with someone at a top university.
2. How rapidly are the best current models are able to self-improve?
In my view, their output is good enough that REinforcement Learning can work, synthetic data can work, time scaling can be scaled, they can grade themselves, and they are on a steady glide toward ongoing self-improvement. You can debate the rate, but “compound interest” gets you there sooner or later.
It is fine if someone disagrees with that, but at the very least you want them to have considered this possibility. Most observers really have not.
3. How will the best current models be knit together in stacked, decentralized networks of self-improvement, broadly akin to “the republic of science” for human beings?
This one is far more speculative, as it is not possible to observe much directly in the public sphere. And most of what will happen does not exist yet, not even as plans on the drawing board. Still, you want a person to have given this question some thought. I believe, for one, that this will move the AIs into the realm of being truly innovative. Stack them, have some of them generate a few billion new ideas a week, have the others grade those ideas…etc.
I find it is difficult to have good AI conversations with people who are not conversant with at least the first two realities on this list. The very best conversations are with people who also have spent time thinking about number three as well.
China fact of the day
Marriages in China plunged by a fifth to the lowest level on record last year, a setback to efforts by the government to reverse a demographic crisis threatening the world’s second-biggest economy.
The number of marriage registrations fell to 6.1 million, according to statistics released by China’s Ministry of Civil Affairs on Saturday, after a post-pandemic increase to nearly 7.7 million in 2023. The tally for last year marks the fewest marriages since public records began in 1986 and is less than half the peak reached in 2013.
Here is more from Bloomberg News.
What should I ask Jennifer Pahlka?
Yes, I will be doing a Conversation with her. From Wikipedia:
Jennifer Pahlka (born December 27, 1969) is an American businesswoman and political advisor. She is the founder and former executive director of Code for America. She served as U.S. Deputy Chief Technology Officer from June 2013 to June 2014 and helped found the United States Digital Service. Previously she had worked at CMP Media with various roles in the computer game industry. She was the co-chair and general manager of the Web 2.0 conferences. In June 2023, she released the book Recoding America: Why Government Is Failing in the Digital Age and How We Can Do Better.
Recently she has been working on the Niskanen Institute state capacity project. So what should I ask her?
Tuesday assorted links
Lift the Ban on Supersonics: No Boom
Boom, the supersonic startup, has announced that their new jet reaches supersonic speeds but without creating much of an audible boom. How so? According to CEO Blake Scholl:
It’s actually well-known physics called Mach cutoff. When an aircraft breaks the sound barrier at a sufficiently high altitude, the boom refracts in the atmosphere and curls upward without reaching the ground. It makes a U-turn before anyone can hear it. Mach cutoff physics is a theoretical capability on some military supersonic aircraft; now XB-1 has proven it with airliner-ready technology. Just as a light ray bends as it goes through a glass of water, sound rays bend as they go through media with varying speeds of sound. Speed of sound varies with temperature… and temperature varies with altitude. With colder temperatures aloft, sonic booms bend upward. This means that sonic booms can make a U-turn in the atmosphere without ever touching the ground. The height of the U varies—with the aircraft speed, with atmospheric temperature gradient, and with winds.
….Boomless Cruise requires engines powerful enough to break the sound barrier at an altitude high enough that the boom has enough altitude to U- turn. And realtime weather and powerful algorithms to predict the boom propagation precisely.
Here is the crazy part. Civilian supersonic aircraft have been banned in the United States for over 50 years! In case that wasn’t clear, we didn’t ban noisy aircraft we banned supersonic aircraft. Thus, even quiet supersonic aircraft are banned today. This was a serious mistake. Aside from the fact that the noise was exaggerated, technological development is endogenous.
If you ban supersonic aircraft, the money, experience and learning by doing needed to develop quieter supersonic aircraft won’t exist. A ban will make technological developments in the industry much slower and dependent upon exogeneous progress in other industries.
When we ban a new technology we have to think not just about the costs and benefits of a ban today but about the costs and benefits on the entire glide path of the technology.
In short, we must build to build better. We stopped building and so it has taken more than 50 years to get better. Not learning, by not doing.
In 2018 Congress directed the FAA:
..to exercise leadership in the creation of Federal and international policies, regulations, and standards relating to the certification and safe and efficient operation of civil supersonic aircraft.
But, aside from tidying up some regulations related to testing, the FAA hasn’t done much to speed up progress. I’d like to see the new administration move forthwith to lift the ban on supersonic aircraft. We have been moving too slow.
Addendum: Elon says it will happen.