Sunday, November 12, 2023

Sunday Night Links

  • It is no accident that the American military is in a recruiting crisis. Who in their right mind would want to fight for a regime regarding which the primary controversy is whether it is merely stupid or actively evil, and of which it is not controversial at all that the stupid and/or evil people managing it hate you? [Postcards From Barsoom]
  • Whenever addictive and destructive neuro-chemical pathways are introduced to a society it takes generations for said society to learn how to adapt to these temptations. China has learned the hard lesson of opiates at least twice, and I think that has some influence on their cultural (in)tolerance of those substances. With the invention of distillation in the 18th century, 19th century America was drowning in hard liquor, 9 gallons per person annually of 80 proof. It took a century and the temperance movement (largely driven by women sick of their deadbeat husbands) to help us regulate to a point, but alcohol is still many’s greatest demon. The same process must happen with screens. I’m pretty sure it won’t just be with screens though. All invasive technology will have to be rejected (on a personal and moral basis just like other substances which activate dopamine artificially). We must see that screens (and their connection to the internet) are like cigarettes for our soul, rotting us from the inside, making us less capable of interacting with the real world. [Chaos Ordered]
  • Like Scott, I do not want to preach radical skepticism. I want to preach scientific reasoning. If you’re interested in the research on $topic_x, you should familiarize yourself with the methods of that field, and especially with the field’s most critical voices. You should know what’s right and what’s wrong and be able to recognize all of the issues that are common enough for the field’s researchers to see them with a sideways glance. Most people are not equipped to do this. When they are, they may not know they’re capable; when they’re not, they may wrongly believe they’re capable. Scott’s recommendation to decrease your confidence in claims is good, avoiding biased people’s conclusions is also good, although I would like to add that biased people are good to read to understand flaws in the arguments of the people they oppose, and the need to look at all of the evidence is still quite obvious. I want to add that thinking about causal inference by focusing on designs is necessary to build a proper understanding of science in general. One well-designed, high-powered study is often much more valuable than a vast number of more poorly-identified and lower-power studies. This is so true that the man of one study who knows he’s the man of one study because the rest are garbage is often much less wrong than his peer men of many studies.
    [Cremieux Recueil]
  • When humans observe a task, mirror neurons simulate the feeling of physically performing that same task. Musicians in the audience are playing their instruments in their mind, and if all we could see was their synaptic activity, we would struggle to differentiate between the performers on stage and their fans. Likewise with literature. When we look upon this ancient text, our neurons fire in a particular way that no human brain has experienced since the last reader carefully wrapped the scroll and returned it to its alcove, minutes or years before looking out the window and noticing that the volcano had come to life. Recovery of ancient writing is cognitive archaeology. [Casey Handmer]
  • Sovereign finance should be viewed simply as a form of banking. Sovereigns raise funds for unspecified purposes and promise risk-free returns they may be unable to provide in real terms. When things go wrong, bondholders think taxpayers should be on the hook, and taxpayers think bondholders should pay. As usual, everyone has a patsy, someone else was supposed to take the hit. Ex ante everyone was assured they have nothing to fear. [Interfluidity]
  • Looked at the right way, this last decade was essentially about copying China without admitting it, from the Atlantic pushing for China-style censorship to NYT calling for China-style industrial policy. All of that is precedent for what a Democrat/Communist pact might look like. Think about how Hollywood came to rely on Chinese money, preaching civil rights while practicing selective censorship…but at country scale. A Democrat/Communist pact could be a return to Obama-era Chimerica, but on different terms, where China becomes at least an equal partner and perhaps eventually the unofficial driver. With that said — yes, the silver lining of such a rapprochement is that we could avoid WW3. But it’s a monkey’s paw outcome, because detente would allow Democrats and Communists to focus on their many other enemies within and without — from American conservatives to Chinese liberals, from India to the Internet. And so the implications for everyone that isn’t a Democrat or Communist are ominous. [Balajis]
  • These bond losses remain “unrealized” so long as people don’t come looking for their money. But when they do — like during a bank run when banks must pay out, or during hurricane season when insurance companies must fork over cash — then bad assets are sold and losses become “realized.” And then we all “realize” that the poor dumb institutions that bought US government bonds in 2021 are actually insolvent — unless the Fed prints to paper over a bond crisis the Fed itself caused. Which it will do, over and over again, until the crisis devalues the dollar itself. Anyway, this is what happened in March 2023 when five huge banks died in quick succession, followed by a quick print called BTFP to cover up the Fed’s failure. [Balajis]
  • Not a single corporate journalist, politician, regulator, or policeman thought to investigate SBF until Erik Voorhees smelled a rat and Ian Allison found the rat. In fact, even the least self-aware corporate journalist in modern history admitted that citizen journalists “outshine traditional media on coverage of FTX implosion.” So: yes, the only reason SBF was even exposed, let alone convicted was because of people posting on Twitter. Twitter is important! That’s why the regime didn’t want Trump to post on there, doesn’t want you to post on there, and doesn’t want Elon to let you post on there. [Balajis]
  • Because digital glasnost is upon us. We now have truly free speech. And so millions of people have now been able to match up their individual observations with each other in public, thereby discovering that much of what the US establishment says is manipulated, misleading, or erroneous in some fashion — in a word, fake. Of course that doesn’t mean the US establishment is always lying. Sometimes they’re in error. Sometimes their data is merely noisy, or quietly revised. Sometimes they are telling the truth, but spinning it. Sometimes they are taking a position for tribal political reasons. And sometimes they are, of course, outright faking it. [Balajis]
  • It is important to emphasize the word available because poor credit is obviously not in itself a cause of poor loss experience. In this sense, it is analogous to territory. Presumably credit is predictive because it reflects varying levels of "stress", planning and organization, and/or degrees of risk-taking that cannot be directly measured by insurers. These specific conjectures have been offered many times and they are intuitively plausible. However it is less conjectural to say that whatever credit might be a proxy for, it is not a proxy for any other variable (or combination of variables) practically available to insurers. In our data mining projects we explicitly set out to generate the most comprehensive universe of predictive variables possible. In this sense, we therefore use credit in the "ultimate" kind of multivariate analysis. Even in this truly multivariate
    setting, credit is indicated to have significant predictive power in our models. It is beyond the scope of this paper to comment on the societal fairness of using credit for insurance pricing and underwriting. From a statistical and actuarial point of view, it seems to us that the matter is settled: credit does bear a real relationship to insurance
    losses. [Cheng-Sheng Peter Wu]

No comments: