Friday assorted links
1. Rob Wiblin interviews Rohin Shah, who leads AGI alignment/safety at DeepMind.
2. Books Arnold Kling has reread.
4. Even (especially?) for ontologists, supply is elastic.
6. Major layoffs at The New School.
7. New forthcoming Ethan Mollick AI book.
9. How much is AI boosting productivity anyway? (FT) A much-needed dose of sanity.
Tyler and Alex Speak to OpenAI
We were honored to speak to OpenAI about the economics of AI. Lots of good material here. Self-recommending.
Western hemisphere fact of the day
Overall, the Western Hemisphere now produces more oil than the Middle East did before the crisis. Canada is the world’s fourth-largest oil producer. Brazil produces four times as much oil as Venezuela; and in Guyana, where production began only seven years ago, output almost equals Venezuela’s. In Argentina’s Vaca Muerta region, shale oil production has grown sixfold since 2020. The current disruption will propel more oil and gas investment in the Western Hemisphere and Africa.
Here is more from Daniel Yergin in the WSJ.
Rubber rationing in World War II
When during the meetings the Americans offered that at most they could convert 15 percent of U.S. auto plants to military production, Beaverbrook replies that 100 percent of British automobile factories had been converted, and encouraged Roosevelt to aim higher. He did, and on January 1 he ordered U.S. auto production halted by late Februrary. Within weeks the dearth of newe cars became moot when rubber, 90 percent of which came from Malaya and Indonesia, was rationed. The U.S. had no synthetic rubber factories to make up the shortfall. Americans soon learned what Britons had long known: without a spare tire or three stashed in the garage, the family car had a very limited range. Passage by rail — where for fifty years the Pullmans had been Americans’ preferred means of conveyance — was soon limited to troops and businessmen on official war business. And then the airlines — their routes and the national fleet of 434 aircraft — were commandeered. By spring, gasoline ratioining, as a mean to preserve rubber more than oil, dribbed on to the Eastern Seaboard and in the following year spread nationwide, guaranteeing that Americans in the heartland could no longer take their vacations at east or west coast beaches even if their bald tires could carry them there. That proved okay with most because by summer, oil and bilge tar and decomposing bodies — the U-boats’ harvest — regularly washed up onto America’s eastern beaches.
That is all from the excellent The Last Lion: Winston Spencer Churchill: Defender of the Realm, 1940-1965. As I’ve said before, you can always keep on reading books about World War II and you will continue to learn interesting and important things.
My twenty-minute AI talk for the Swedish company Sana
Thursday assorted links
1. Does banking consolidation harm households?
2. Some comments on the new federal framework for AI regulation.
3. This guy is skeptical about doing things in space.
4. Is Chairman Mao underrated, and why did India not get rich too?
5. “Between 1985–2023, MIT’s faculty grew 9%. Administrative staff grew 189%.”
CA Logic
In California a 17 year-old can drive a car but can’t ride alone in an Uber or a Waymo.
I long for the days when you could put a kid in the post. We have weakened as a civilization.
Should we recriminalize marijuana?
That is the topic of my latest Free Press column. Here is one excerpt:
The present and also future of mankind is a world where reasonably high levels of self-discipline are needed to do well. The journalist Daniel Akst pointed this out in his 2011 book Temptation: Finding Self-Control in an Age of Excess, and we are now living it full force.
I would rather cope with that world than face the full nanny state, backed by modern, AI-intensified surveillance techniques to boot. Concentrating more power in political authorities hardly solves the basic problem. If marijuana and sports gambling can manipulate weak individuals, so can unscrupulous political leaders. A greater realization of individual weakness does not translate into a case for more government action; if anything, it suggests the opposite. Better to allow our social problems to fester in a more decentralized fashion, rather than reinforce our social pathologies through a manipulative and dysfunctional leader at the very top.
In the longer term, we may need to look to medications, such as GLP-1 drugs and their offshoots, which seem to curb some forms of addictive behavior beyond the appetite for food. Alternatively, some individuals may choose self-surveillance, with self-imposed penalties for bad or addictive behavior. Perhaps your AI, or a hired third party, docks your bank account every time you puff on a joint. I am not convinced such services ever will become popular, but that should be taken seriously as an indicator of what people really want to do. We can at least give them better options for self-constraint. If they rarely choose such options, then perhaps for many of those people, marijuana consumption is not a matter of weakness but a very well-established preference, whether we like it or not…
In short, it is time to realize that paternalism is far less workable than in times past. Our government does not have the credibility, the control over information, or the control over our lives to pull it off.
I do understand that is in some significant ways bad news, as voluntary choice is overwhelming some of us with bad outcomes.
My response is to start by accepting some steps backward, holding paternalist tyranny at bay, and hoping some longer-run cultural and technological adjustments will make this all more workable.
If you have a better solution, I would love to hear it.
Recommended.
Law professors prefer AI over peer answers
Large language models (LLMs) are increasingly promoted as educational tutors, yet most evaluations focus on domains with a single ground truth. Many disciplines, however, hinge on judgment: reasoning, weighing ambiguity, and reaching defensible conclusions. Law provides a sharp test. We conducted a blinded evaluation of short-answer tutoring in contracts courses with sixteen U.S. law professors. Participants created 40 representative questions, wrote answers, and judged 2,918 anonymized comparisons between human and LLM responses. Professors rated LLMs far higher than their peers (average win rate = 75.33%), with models performing similarly to the best instructor. LLM responses were also rarely flagged as harmful (3.53%, vs 12.06% for professors). Preferences for LLM answers were consistent across evaluators and reflected shared professional standards. Our evaluation can be reliably extended to additional models by employing a separate LLM as a judge, rendering expert agreements an effective, scalable method to evaluate AI tutors in judgment-rich domains.
“far”. That is from a new paper by Alejandro Salinas, et.al. Via Andrew Curran. And via John Chamberlain:
Artificial intelligence (AI) and large language models (LLMs) tools are capable of mass-producing academic finance papers that are nearly indistinguishable from human-authored research, according to a new study published in the Journal of Economic Literature.
C’mon people, get ready. I know it is difficult to admit when your human capital has been devalued, but that time is upon us. In particular, being prolific is no longer such a comparative advantage in academia. You might run to the “but I know what questions to ask” cope, but I implore you to solve for the equilibrium. What is the equilibrium wage for merely asking questions?
Of course academic life and projects will continue, but the real rewards will go to people doing new, innovative, and hitherto impossible projects with AI.
Wednesday assorted links
1. One hundred greatest bird names of all time.
2. The method for counting 52,019 puffins.
3. Luggage-tag switching scheme involves flights from Canada to countries where drug smuggling can carry death penalty. Yet a third link from Jodi Ettenberg.
5. How cowboy culture remade Brazil.
6. Are India’s gdp figures OK after all? (FT)
7. Plans released for a $16 billion mile-long ship capable of carrying 80,000 people.
Richard Feynman’s formula for the best holiday restaurant
According to Feynman’s approach, in this context, people should try a different restaurant each night until they find one that exceeds a particular threshold that reflects a desired quality.
In Feynman’s equations this threshold is not fixed. Instead it declines more and more rapidly as the number of days left in the city reduces. In other words, as the days go by there is increasingly less motivation to hunt for an amazing dining spot, because the time you will have to enjoy it has decreased.
“The thresholds are being guided by the best thing you might be able to find if you kept looking,” said Griffiths. “If you have a long time to look, finding something amazing has a lot of value because you can go back many times.”
Feynman’s approach assumed there is equal possibility of finding any restaurant within a fixed range of quality. However the researchers also explored other scenarios.
“We showed that if the distribution of restaurants varies, then the strategy you should follow will change too,” said Griffiths.
Here is the full story, and here is the PNAS article. I think of that as a pretty pessimistic approach to the problem. In most locales you should be able to find lots of very good restaurants, so if you find a quality place early on you do not return to it, rather you keep looking for more, in fact feeling emboldened by your early success. Maybe this algorithm applies to Cuba?
Via both Adam K. and Mike Doherty.
Sentences to ponder
In 2019, there were about 150,000 people working in autism therapy. Six years later, there were 654,000—more than the number of people who work in mining and logging, or telecommunications, or at the US Postal Service.
That is from Derek Thompson. And here is the seven-minute nap story from the NYT:
At Compleat Kidz, a fast-growing chain of autism clinics based in North Carolina, the policy is firm: Naps cannot be longer than seven minutes before children are awakened to resume therapy. The company says this is necessary to prevent fraud since clinics can be paid only when children are awake and getting services. But it also allows the clinic to bill insurers or Medicaid for more hours.
I do not even need to say “model this.”
Consent-based laws and aggregate fertility
This paper examines how expanding the legal definition of sexual assault affects fertility and sexual behavior, using a panel of European countries. I find that switching to tacit consent-based legislation reduces fertility by about 4% relative to the mean. This effect is driven by a decrease in couple formation and an increase in abortion rates. Supporting evidence is consistent with a behavioral channel in which more risk-averse individuals withdraw from dating and partner markets following the reform, altering the composition of those who remain active toward a pool that is less precautionary. Consistent with this compositional shift, contraceptive use rises among younger women but declines among older age groups, while condom use falls among young men. Finally, an analysis of appeals court verdicts in Sweden following the adoption of consent-based legislation shows a decline in unanimous guilty verdicts, indicating challenges in assessing tacit consent. These results are consistent with a simple framework in which heterogeneity in risk perceptions and precautionary behavior in dating and partner markets, including reduced participation by some individuals, helps explain the observed decline in fertility following the reform.
That is by Adrian Mehic in the Journal of Health Economics. tekl.
Big if true
Several important questions — such as the possibility of debt-rollover without primary surpluses — turn on whether the present value of the aggregate endowment is finite, i.e., whether the economic growth rate under the “risk-neutral” measure, lies below the risk-free rate. It is tempting to argue that the endowment must be finitely valued, since there exist finitely-valued, non-depreciating assets whose cash flows are cointegrated with aggregate output. This paper shows why this argument is incorrect. A remarkable historical episode in which French government bonds were indexed to aggregate growth allows direct measurement of the risk-adjusted growth rate, which is found to exceed the risk-free rate.
That is from a new NBER working paper by Stavros Panageas.