Category: Uncategorized
Thursday assorted links
1. Cato’s Vision for Liberty award for 50k.
2. Gross output signals an economic surge (WSJ).
3. Echo, a new AI site to mimic the styles of particular writers or writing styles. Thread on it here.
4. “Open USD (OUSD), the new stablecoin from Coinbase, Mastercard, Stripe, Visa and others launches…”
5. Are men or women more tolerant of differing views?
6. On AI desires.
What should I ask Moxie Marlinspike?
Yes I will be doing a Conversation with him, live at the Roots of Progress event next week. From Wikipedia:
Moxie Marlinspike is an American entrepreneur, cryptographer, and computer security researcher. Marlinspike is the creator of Signal, co-founder of the Signal Technology Foundation, and served as the first CEO of Signal Messenger LLC. He is also a co-author of the Signal Protocol encryption used by Signal, WhatsApp, Google Messages, Facebook Messenger and Skype.
There is much more at the link, for instance he is also an anarchist of some kind or another. So what should I ask him?
Merging LLMs and economics research
We introduce an open-source workflow that enables an LLM to reproduce, improve, and extend an economics article using the article’s published replication package. First, the workflow attempts to reproduce the original calculations, checks for discrepancies with published findings, and performs automated sensitivity analysis. Across 4,452 published replication packages for five economics journals, the workflow flags discrepancies in 3,460 articles or their appendices. Second, the workflow improves the original calculations by using a different implementation or algorithm. In 496 articles, the workflow is able to reduce a calculation’s computation time, at similar or greater accuracy, by more than a factor of 10. Third, the workflow extends the original analysis. In 923 articles, the workflow develops an extension that does not appear in the original article and that is aligned with the original article’s goals and assumptions.
That is from a new paper by
Wednesday assorted links
1. It seems there is no evidence for the concept of a fertility rebound.
2. We will tell children nasty stories, but mostly only show them positive images.
4. Weather risk is reflected in Florida home prices.
5. Have we discovered where Aristotle taught Alexander the Great?
Don’t let AI make you dumber
That is the topic of my latest Free Press column, here is one excerpt:
I do not think the skeptics would put it this way, but as I read Conti, I find he has a pretty bleak fundamental view of humanity. Are we all really just looking to veg out and abandon curiosity and inquiry, at least once the machines have taken care of both the basic functions of life and certain higher aims such as scientific research? I think some people are like that—indeed you might say many people—but it does not reflect what I take to be the general human condition.
If I look at most people who might fit into the “middle class” when it comes to intellectual pursuits or educational status, I observe they have a lot of strong interests. This might play with their pets, improve their performance at sports, or learn how to cook better. You do not have to identify those preferences with “the new Athens” or “the next Mozart” to think they are perfectly good and noble ways for people to spend their time.
Most of us want to do something interesting and stimulating with our leisure time, and if we do not, it is often because our jobs are so busy and stressful that we just wish to decompress. Of course, in this radical vision of our AI future, fewer jobs will be so all-consuming and so more of us will use vacations and leisure time to explore and learn rather than to just sit on the beach scrolling our phones. And to the extent some jobs do remain hectic, or become even more so (such as cybersecurity), they will continue to be challenging and intellectually stimulating.
A related worry is that humans may feel they simply cannot compete with the AIs, and thus they might turn away from creative pursuits. It is true that I, more than ever, have given up all hope of proving new theorems in mathematical economics. But many of my intellectual and creative pursuits do not involve competition at all. For instance, I use AI to understand classical music better, asking the models questions before I sit down to listen to a piece. (Such as “which are the best recordings?” and “what should I listen for in the second movement?”) As the models get better and smarter, I am not going to be discouraged in this endeavor, as I was not “competing” with the models to see which of us knew more. Rather, I will gratefully end up much better informed about classical music—my increasing knowledge has already induced me to see more live concerts.
Recommended, and AI saved me time on the proofreading and fact-checking (not the writing!), so I could return to reading China Mieville…
Tuesday assorted links
1. Comment section arbitrage. This is what the AIs do too, right?
3. Good Bryan Caplan post on Effective Altruism.
4. Reported female bisexuality is in retreat?
5. Sadly, now a fringey view, yes. But not with me.
Accounting for Cross-Country Income Differences Revisited
Also known as Why I Do Not Believe in the Housing Theory of Everything:
Development accounting is the search for proximate sources of cross-country income differences. This article describes how knowledge in this field has evolved over the two decades since the influential work of Caselli (2005). There have been large advances in the measurement of production inputs (labor, physical capital, and human capital). These advances have raised the estimated contribution of inputs, mostly human capital, in development accounting. Our preferred estimate is that inputs account for 55–70 percent of gross domestic product (GDP) per worker differences, versus 30 percent using the classic specification. The literature has also made progress in moving away from Cobb-Douglas production functions and measuring factors such as management quality that were previously bundled into total factor productivity (TFP). Our review highlights the new implications of these advances, areas where future research would be beneficial, and the limitations of development accounting.
That is from a new NBER working paper by
Monday assorted links
Bundesrepublik Deutschland
At times we forget what an amazing wonder the Bundesrepublik Deutschland was. At the end of the World War II, Germany was one of the sickest and cruelest human societies in history, ever. Not too many years later, it was one of the best and most successful societies ever.
By the 1980s, living standards had caught up to the United States, with the provision of public goods sometimes superior. The country was fully democratic, pro-Western, and largely pro-American.
Their rail system and postal service were amongst the best ever created.
For thinkers of note there were Hans Blumenberg, Habermas, Peter Weiss, Gadamer, late Heidegger and late Carl Schmitt, Reinhart Koselleck, the underrated Klaus Theweleit, Niklas Luhmann, and perhaps you value some of the other members of the Frankfurt School.
The visual arts were very strong, with Richter, Polke, Baselitz, Beuys, Penck, Palermo, and much more.
Music produced Stockhausen, Henze, Lachenmann, Rihm, Zimmerman, Kraftwerk, Can, and all of Krautrock, later techno, though right at the time of unification rather than during the BRD per se. The list of vocalists, instrumentalists, and conductors is strong. Was there anywhere better for hearing opera?
Perhaps I prefer the fiction from Austria and Switzerland, but at the very least Germany provided a major market for those authors and it was an extraordinarily literate country with amazing bookstores. For domestic authors there were Böll, Patrick Süskind, Siegfried Lenz, Wolfgang Koeppen, can I count Uwe Johnson?, and Arno Schmidt maybe? I do not like Grass, but it seems wrong not to list him.
The food could be very good, especially in the southwest. There were Michelin star restaurants all over the country (still are, to be clear on this point). So many well-functioning cities, with many of the world’s best transit systems. Lots of nuclear power and a strong industrial base. West Berlin was an exciting city with an air of mystery. Some might say no speed limit on the Autobahn, though I am less sure that was a virtue. Plenty of beautiful women and reasonable attitudes toward sex.
If you had to choose, what was the worst thing about the country? No shopping on Sundays? Workplace and shopping hours discrimination against women? Too much smoking? Obsession with Waldsterben?
Are there features of post-unification Germany that can compare to this earlier era? So much seems not to work well. So many policy mistakes have been made. So much leadership lost in the areas mentioned above. So much pessimism, sadly a lot of it seems to be justified.
Where did all the good performance go? And why did it leave? Lack of a communist enemy? Absorption of East Germany? The simple accretion of distance from pre-WWII German creativity?
The wonder that was the Bundesrepublik Deutschland. Johannes, we hardly knew ye.
AI in science
Scientific progress is a key driver of economic growth and prosperity. There is great excitement- but also concerns- about the impacts of AI on science, but so far little data. We provide early insights on this from three data sources: a sample of 15 million Gemini interactions, an inventory of over 2,600 specialized AI models across disciplines, and a survey of over 600 scientists. We map these data to a new taxonomy of scientific tasks to study how scientists are using AI. Four main findings emerge. First, we find broad adoption and coverage: scientists use AI more than most other occupations. Specialized AI models have broad disciplinary coverage and are highly cited. Nearly half of the scientists surveyed report using some form of AI every day. Second, we document evidence that LLMs (proxied through Gemini usage) and specialized models act as complements—LLMs are used for general analysis, coding, and manuscript preparation, while specialized models provide domain-specific predictions, data generation and classification. Third, scientists report large productivity gains from using AI: a saving of nearly 7 hours per week, time which is primarily re-invested in more research. Finally, we show that AI is already changing the scientific process. As some stages of scientific research become easier, bottlenecks shift downstream. Scientists report an increased backlog of untested hypotheses and substantial demand for output verification. Our findings suggest that AI holds significant potential to increase scientific productivity. However, as with other sectors, its ultimate impact will be governed by complex task interdependencies and investment into the elimination of emerging bottlenecks.
That is from a new paper by Mihai Codreanu, et.al.
Sunday assorted links
Earth fact of the day, #2
The shortages have gone on for so long that they are aggressively driving down how much carbon is being released into the atmosphere, a Washington Post analysis of data from the International Energy Agency shows. People worldwide are using significantly less oil and gas, which means less climate pollution…
Such an annual decline has not happened since the height of the coronavirus pandemic. Fossil fuel consumption dropped significantly more then, and consumption was lower in absolute terms, too. Crude oil demand averaged 91 million barrels per day in 2020, compared with 102 million barrels per day under the latest IEA forecast.
But this year’s drop — especially given the sharp increase that was initially forecast — is substantial.
Here is the full story. Not a good thing overall, but there is a lesson in that to…
Saturday assorted links
A doomsday scenario for American AI
That is the title of my latest Free Press column, here is the closing bit:
Sick and elderly Americans will go to Chinese companies for their AI-invented and AI-tested medical devices and drugs. America still will be a wealthy country, so China will charge the highest prices possible, yet prioritize Chinese citizens for treatment. Large numbers of Americans will die prematurely, at least compared to a world in which many of those innovations came from the U.S. My colleague Alex Tabarrok has coined the phrase invisible graveyard to refer to these lost lives, invisible because we do not observe the state of the world where they get treatment readily and cheaply. Over time, this invisible graveyard will swell into the many millions.
Finally, we will ask what went wrong.
The postmortem will be this. Many people panicked about the possibility of strong AI models killing us all. That fear was not based on peer-reviewed scientific research showing a high chance of doom, nor was doom indicated in any market prices of the time, including measures of risk. It was a story, just like this is a story, and it was spread on social media. The key point of the doom story was that, if America keeps the No. 1 spot in AI models, the models will be so strong they will do us all in, or lead to unimaginable catastrophes. We were too afraid to have America keep the lead, forgetting that if truly destructive AI is our fate, the doomer scenario can come from Chinese AI as well.
Today, we cannot say for sure that the AI doom scenario is false. But is it a story we wish to live by? Is belief in it a good way to protect and extend life, liberty, and the pursuit of happiness? Will it help us much, or for long, if it is Chinese AI that turns on us and does us in?
In my view, successful societies accept challenges and meet them. Solving problems, bit by bit, is the best way to ensure that we have the capabilities to meet big and truly existential risks, should those risks come along. Debating the chances of our doom, ex ante, on a highly speculative basis, is unlikely to provide the same kind of expertise and talent cultivation. It is instead more likely to demoralize and immobilize us.
So which America are we going to choose?
Recommended, do read the whole thing.