Longreads + Open Thread

Longreads

• Matt Beane in IEEE Spectrum on how robotic surgery makes it harder for surgeons to get effective training. This problem is much broader than in healthcare: in any industry that requires specific, narrow skills, expectations around training can get formed with one level of technology and then get applied with another. In some professions, that leads to well-paid workers doing fairly rote work (junior investment bankers get paid pretty good money to format powerpoints, at least if you calculate their compensation by the year and not by the hour). In other professions, technology means that people miss learning some of the fundamentals instead. It can be a useful exercise sometimes to take a problem you're working on and ask how someone would have solved it thirty years ago. If nothing else, you'll get a renewed appreciation for how much things have improved.
  
  (Via Liberty's Highlights.)
• Justina Lee in Bloomberg has the story of how crypto fund Three Arrows Capital blew up. One interesting thing in this story is that the public narrative about Three Arrows was that they were calling the crypto supercycle, but many of their trades were less about directional bets and more about arbitrage. As it turns out, many of those arbitrages were, in fact, dependent on some variant of the supercycle theory: when prices went down, the low-risk trades blew up, too.
• Saba Rahimian in Eater on what it's like to run a restaurant post-Covid. It is, oddly enough, yet another story of businesses reluctant to invest in more capacity because of concerns about long-term demand. ("Amid the struggle to ramp up, media recognition and accolades were a double-edged sword. Weekends that followed a highly viewed media post were mayhem. A line formed out the door from 10 a.m. until 3 p.m., wait times exceeded an hour, and phone orders rang in every minute; I had to shut off online ordering.") It's easy to see why companies would struggle with a low absolute level of revenue, but given the large swings in consumer spending between goods and services, the question of how they deal with unpredictable revenue will remain important.
• Cedric Chin in Commonplace on how to read and misread history. The trouble with treating history as a science is that your sample size is always 1: you're either looking at a single event, or looking at events that are similar but that all have numerous confounding variables. This doesn't make it worthless to study what's happened in the past, but does make it tricky. Chin makes a case for a specific way to extract repeatable lessons.
• And on the topic of history, Tanner Greer of Scholars Stage asks why the share of undergraduates majoring in history dropped from ~5% in the 60s to ~1.5% today, with several theories. One of the strong ones is that majoring in history was popular when more college students were well-off and not particularly worried about their career prospects (and one piece of evidence for this is that Harvard, Yale, and Princeton students are disproportionately likely to choose it as a major). And the darker possibility is that history requires lots of reading, and the average person, even the average college student, is less and less willing to do it.

Books

• Procter and Gamble: The House That Ivory Built: Corporate histories tend to be obsequious, and salespeople who work with a single big company tend to be, too. So a history of the world's largest advertiser, published by Ad Age and based heavily on interviews with P&G’s ad agencies, is definitely on the favorable side. Still, it's a good look at the company's history, and in particular their early adoption of new media. It also has some fun anecdotes. For example, in the 50s, the company wanted to measure how many households in Venezuela were tuning in to radio and TV over a specific fifteen-minute period. They hired out-of-work bullfighters to sprint down the street conducting rapid-fire door-to-door interviews. This book was a key source for this week’s subscribers-only post on P&G ($).

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Something I've been thinking about recently: usually, service jobs have lower measured productivity growth than manufacturing or agriculture. Some of this is because they're hard to measure (the American healthcare system can do very impressive things it couldn't do in 1970, but whether we're getting our collective money's worth is debatable). But some of this is related to the structure of the jobs themselves; it's hard to imagine a Moore's Law for therapists, teachers, or personal trainers, allowing them to double the number of people they work with every eighteen months. On the other hand, each of those jobs does have a partly-automated version—meditation apps, MOOCs, and connected fitness. So: which service jobs are most likely to get mostly or entirely replaced by software, the way travel agents were? And which things that are currently un-digitized service jobs are next up for automation?

A Word From Our Sponsors

Here's a dirty secret: part of equity research consists of being one of the world's best-paid data-entry professionals. It's a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.

If you've ever fired up Excel at 8pm and realized you'll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.