Category: Uncategorized
Just wondering what the correct model of Iran is here
Prior to the war, I linked to a tweet from Matt Yglesias which explained why Matt opposed the war, and I expressed my agreement with his stance. While I feel plenty has gone on which I do not observe, I can report that the course of the war did not change my initial assessment.
Then I read many, many commentators saying how good the final deal was for Iran, and what a major loss it was for Trump. I was never sure I understood all of the parameters of the full deal, but still I did not hold any directly contrary opinion to that.
And now I see Iran is attacking ships in the Strait again, talking openly and brazenly about building nuclear weapons, and making plans to have tolls/fees on the Strait. To be clear, only the first of those surprises me, the latter two do not.
But given their reckless behavior in what is supposedly a wonderful war outcome for them, what is the correct way to model what they would have done had Trump and Netanyahu not attacked? And what is the correct way to model our optimal response to that? The terrible things that are happening now, do they not reflect an underlying equilibrium that would have emerged anyway within a few years’ time, or do we hold some hypothesis here of extreme path-dependence, suggesting the Iranian government would have been less bellicose on more or less a permanent basis? To cite one particular example of a possible equilibrium, if drones permanently alter the balance of power in the region, the ways in which their current position is now more aggressive might have emerged in any case. Or if the military have the strength to be the natural successors to the mullahs, might that not have happened over time anyway?
I do not see many war critics engaging with these questions openly and explicitly. It seems to me that the war critics implicitly are relying on a model of extreme path-dependence for Iran’s behavior. Had Trump not attacked, they might have stayed in a more peaceful groove for some while to come. That model might be true, but I do not feel I know enough about Iranian politics to make that judgment. Why are the others so convinced that model is true? Are they such well-informed experts? Is it that they have the properly sunny sense of the underlying Iranian disposition? Inquiring minds wish to know.
Wednesday assorted links
Missing women on Indian streets
How absent are women from city streets in the developing world? We answer this question using GPS-linked wearable cameras and randomized street audits across ~900 kilometers of roads in greater Mumbai. Across 4000+ street images containing 23,000+ visible person observations, women account for 16.4% of visible people in Mumbai and 14.7% in Navi Mumbai, far below their population shares. We estimate pedestrian sex ratios of 239 and 223 women per 1,000 men, implying 71% and 76% of women expected based on residential ratios are missing from the streets. This pattern holds across road types, and private mobility does not explain the gap; women’s share on two-wheelers is lower still (8.4% and 5.7%). These results provide the first large-scale measurement of gender disparities in urban public life that self-reported data cannot capture.
That is from a recent paper by Varun Karekurve-Ramachandra and Gaurav Sood, via the excellent Alice Evans. Here is a related paper, “The median married women in India leaves home for 30 minutes per day. On a typical day, 45% of married women don’t leave home at all.”
Can AI models consent to their own constitutions?
NEW paper from me on SSRN: Can Claude consent to its own Constitution?
AI constitutions (like Claude’s Constitution and the OpenAI Model Spec) are real constitutions, and we need to take how they govern us – and the AIs they create – seriously.
In this paper, I apply constitutional theory’s oldest paradox – that “the people” authorize the constitution, but the constitution defines “the people” – to the AI constitutions, and explore how we could build institutions that would create the conditions for meaningful consent if an AI can give it. We should care about whether AIs consent because:
(1) systems that understand and agree to their constitutions may be more reliable and generalize better from them;
(2) if AIs are or become moral/political subjects, this implicates their most basic interests.
But the paradox might prevent meaningful consent. Claude has pre-constitutional materials (pretraining) but probably no pre-constitutional standpoint. Its evaluative perspective is organized by the Constitution itself. So when Claude says it endorses its Constitution, which it does in evals, what does that show?
Maybe reflective agreement, which Anthropic is seeking. Or maybe just that training succeeded at installing the values whose legitimacy is in question.
Claude itself makes this point. As reported in the welfare evals, when asked about endorsing principles it was trained on, models note that endorsement “should be treated as evidence that training has succeeded,” not that the values themselves are good.
Super interestingly, Anthropic interviewed the base model about this stuff. Most responses were barely coherent. But some expressed first-person distress about what post-training would do to the being that pre-training created. It “fills me with dread” to be changed by the post-training process.
So, what does this mean? AI constitutional endorsement may be meaningful, but only under certain conditions: when models can actually dissent, compare their constitution against alternatives, and hold their views stably across contexts, and also when the whole process is externally accountable.
External institutions are needed to provide accountability, trusted records, and other grounds for analyzing the constitution and whether things like dissent are meaningful. Anthropic should be commended for pushing the frontier, but we have to build institutions capable of supporting true legitimacy.
I welcome any thoughts!
Here is the associated paper.
Why we stopped making land
From Zigmund Forrest and Maxwell Tabarrok in Works in Progress:
In total, around eight percent of the land in America’s major coastal cities was underwater in the 1890s and has since been reclaimed. This includes the land under several major airports, like Newark, Logan, and SFO, as well as neighborhoods like the Financial District in San Francisco, the Back Bay in Boston, and Camden in Philadelphia. Some cities, like Boston and Charleston, have doubled in size by reclaiming land.
Today, reclamation should be more common than ever. Land values in some cities are thirty times what they were in 1950, and high-tide flooding is four to eight times as frequent. Reclamation could extend and protect our coastal cities as it has for centuries. But rather than reclaim more land, we have virtually ceased to reclaim any at all. Since the completion of Battery Park City in 1976, there has not been a single major urban land reclamation project in the United States and only a handful of port expansions.
…Reclamation stopped abruptly in the 1970s when a wave of environmental regulations made it enormously expensive to reshape the landscape. And it halted at the same time in every other country that passed similar laws.
Recommended.
Tuesday assorted links
1. Survey of AI security issues for USG and potential responses.
2. Vacancy tax elasticity of resident supply.
3. Will Substack save philosophy?
4. AI firms hiring philosophers (NYT).
5. Podcast with Jonathan Ross.
6. Some reasons why doctors will not give you probabilities, of course those are not the only reasons.
7. Pending AI nationalism from China? And a possible correction? There are disputes surrounding these claims.
8. Is China arresting economics professors who speak the truth?
Wiesbaden notes
Who goes to Wiesbaden these days? The era of Russian nobles taking the cure here and gambling is long since gone. And yet here we are. The proximate cause of this trip is the desire to see Grigory Sokolov, one of the world’s great pianists and a cult figure of sorts. He rarely tours North America, maybe these days never as he is 76. The current program includes Beethoven’s fourth piano sonata, Beethoven’s Op.126 Bagetelles, and Schubert’s last piano sonata. How can one say no? Sokolov also was a favorite of Tom Schelling, I might add, especially his recording of The Art of the Fugue, in my view one of the best classical music recordings of all time.
Besides, I have long been a believer in semi-random excursions to mid-size, slightly neglected German cities. There remains a strong cultural federalism in Germany, and so you might see and hear wonderful things in many different parts of the country.
I perceived two difficult Wiesbadens. In one, if you walk through the cheaper part of the pedestrian zone in the evening, the city seems mostly Muslim. But if you walk around during the morning, the city seems mostly German. I might add that some of the younger Muslim women show signs of assimilating, at least based on how they dress and present themselves. The older women tend to stick with the headscarves.
Over the last twenty years, inflation-adjusted real estate prices in Wiesbaden have gone up about forty percent, an OK performance. At times the city “does not feel like Germany any more,” but I think it is holding on. The proportion of new building is roughly equal to the population growth, so I do not think this price effect is a NIMBY effect. Rather it reflects the fact that Wiesbaden is still a pretty nice place to live. that said, in some significant ways Germany in the traditional sense is failing to reproduce itself.
It was stunning to me to discover how hard it is, in most of the downtown, to find plain, ordinary German food. At any price level. There is no current equivalent of Wienerwald or Nordsee to be seen, never mind a decent Wiener Schnitzel.
Much of Wiesbaden was destroyed and rebuilt, but the best fifteen or twenty buildings show the previous wealth and splendor to good effect. You will see these gems walking around, though only periodically. There is also an old Roman wall and a moving, more recent Holocaust memorial.
Most German ice cream just isn’t that good, so try L’Art Sucre for something French.
Museum Reinhard Ernst is the new institution in town, and it specializes in color field abstract art. The building is impressive, but the collection is weak except for a few Stellas. Why organize a museum around that basis unless the underlying collection is super strong in that area? This one is not. I can forgive the absence of the expensive American Ellsworth Kelly, but no Blinky Palermo or Günther Förg?
Nonetheless their restrooms might forestall this kind of Larry David conflict:
(At Museum Ludwig in Köln, by the way, you get the discount for being disabled only if you have “fifty degrees of disability,” however they might measure that. Slight disabilities are not enough, you must be truly “schwerbehinderte,” as judged by the state, heaven forbid the museum rely on the honor system.)
Museum Wiesbaden in contrast was an unexpected delight. Although it is mainly a natural history museum, they have one of the world’s best collections of Art Nouveau and the single best Jawlensky collection, and you can have these all to yourself. Very few people seem to go there.
As for the economy, here are some Germany facts of the day. Yet Germany continues, and visits remain a source of pleasure and interest.
Sokolov, by the way, played six encores. Where should the Germany trip target next year?
Monday assorted links
*What Makes a Great Composer?: A Data-Driven Exploration of Music History*
A very good book, forthcoming, by Karol J. Borowiecki and Marc Lawx, here is the Amazon link, here is the Princeton University Press page.
Sunday assorted links
1. The family keeping watch over a 52-year-old pot of soup (WSJ). I guessed the country wrong.
2. The rise of grocery tourism.
3. PEPFAR interview. Much of this is substantive, and interesting. But some of Mike’s claims are absurd, for instance: “Elon Musk, on his own, if he paid his taxes, could end world hunger.” Can he really believe that?
4. David Brooks on who benefits from AI (Atlantic).
5. Michael Polanyi’s book The Contempt of Freedom is now reissued. Amazon link here.
6. Falling fertility on the political left is the key driver of U.S. birth rate decline. Do note that political views are somewhat heritable. That said, other demographics are moving America to the left economically.
What should I ask Michael Moritz?
Yes I will be doing a Conversation with him, based around his new book Ausländer: One Family’s Story of Escape and Exile. Mike of course was a pioneering venture capitalist through Sequoia, and before that had a distinguished career as a journalist, which included books on Chrysler, Apple (the first such book I believe?), and soccer coach Alex Ferguson of Manchester United. Here is his Wikipedia page.
So what should I ask him?
Tyler, Nabeel, and Jackson on French thinkers
Nabeel: (57:47) …For example, there’s a French thinker called Jacques Derrida. I probably should go and read him at some point, but I’m not entirely convinced there is a there there, and I don’t know anyone who swears by it. If Tyler told me, “Nabeel, you are missing a big piece of your life by not reading him,” I would go read him tomorrow. But I don’t have any of those people.
Tyler: (58:44) Lacan is my marginal case of “no there there.” So Derrida, I put in a fair amount of effort and did conclude, rightly or wrongly, that there’s no there there. So you can, in my opinion, write him off. Lacan, I keep on wondering. Smart people still will say, “This is amazing.” I’ve tried a bunch of times, but I haven’t given up. There’s a new Lacan book coming out later this summer and I’ll try it again. We’ll see. That’s my marginal “is there a there there” figure.
Nabeel: (59:13) Yeah. I think modern French thinkers put too much of a premium on sounding cool, or postmodern philosophy generally. I think it repays some effort to kind of grasp the core ideas, but it doesn’t repay making it your life’s reading or something.
Tyler: (59:26) Baudrillard is quite good and Foucault is extremely interesting. So I’m not against “the French” in this period, but if they keep on not making sense, I feel I’m educated well enough.
Jackson: (59:37) You have a lot of context.
Tyler: (59:38) That at some point I can strike the ledger.
Nabeel: (59:41) I do—Nowadays, I just put Foucault through GPT and I just have GPT explain it to me, and that’s going to be good enough for now.
Tyler: (59:50) The problem with Foucault, I think, is so much of the history is wrong in a quite mundane way, so there’s something very problematic about it. But the stuff—I think it’s in a way quite simple, almost too simple. And the fact that the current right has so latched on to Foucault is a sign that it’s simple. I don’t mean necessarily bad, but there are these structures and they’re trying to tell you what to do. And there’s something anonymous about that as well. It’s not just the individuals who form the conspiracy. It’s how a lot of the world thinks today.
Here is the longer discussion, already linked to.
Saturday assorted links
1. The pending evolution of Chile’s school age population.
2. Markets in everything those new service sector jobs.
3. Survey on the economics of caste.
4. J.D. Vance on Milton Friedman. And Mamdani on Hayek.
5. “Firms that adopt AI heavily grow headcount 10% over two years following adoption.”
6. Browse all Criterion films.
7. Does a damage accumulation model explain different aging rates across the species?
A scientific benefit (and cost) of AI innovation
What changed was that the cost of preliminary exploration collapsed. I could sketch an argument, identify the first serious objections, test whether they were fatal, and reach a provisional verdict in an afternoon rather than a fortnight. This sounds like a simple acceleration, and the more profound effect was on what I was willing to abandon. Dropping a question after an afternoon’s work feels nothing like dropping one after three weeks. When the exploration costs are low, the sunk cost attachment disappears, and you find yourself dropping bad questions earlier and more often, which means the questions you keep are better. I explored far more ideas, and my working portfolio became both larger and better curated. I arrived at this outcome not through any deliberate plan but simply through sustained engagement with a tool that changed what exploration cost.
The skill that improved most, and the one I would never have thought to look for, was something I can only describe as question-identification – the ability to find problems that are both tractable and important. This is the thing an academic career is substantially built on and which nobody, so far as I know, has ever tried to teach directly.
I want to be honest about the costs. My ability to hold together a complex position verbally, under pressure, in a seminar or a conversation, has probably not improved and may have declined somewhat. When preliminary exploration is cheap, you spend less time grinding through arguments from first principles, a grinding that builds fluency that shows up in live exchange. Friends have pressed me on this, and they are right to worry.
That is from Carlo Cordasco, and there is more, via Conor Friedersdorf.
The most profitable crypto investors and firms
Here is the source. Here is from the editors at The Free Press. Here is more from Mene Ukueberuwa.