But Moravec’s paradox is important here. Many of the skills that are hardest to automate are also the easiest for humans to do, and require little human capital. Those skills are not scarce in supply compared with demand, so pay is low. And if pay is low, the incentive to automate is less.

Therefore, the labor force in these counties is increasingly reallocating to performing functions which, by law or by socially conducted demand or by their nature, must be performed physically close to some local counterparty.

Even if you automate these jobs, or broadcast over the internet, or relax the regulations, then that just means another round of reequilibration, leaving only jobs of a character which makes performance by (relatively) expensive close humans necessary.

But if a job must be done by humans, for whatever reason, it is subject to human limitations and resulting restraints on productivity, which means low growth rates.

This is especially true when the service demanded has time of performance for human-speed absorption as an essential element. Think of an hour massage, waiting a table over the course of a meal, Shakespeare play, Mozart opera, football game, opening statement to a jury or, ahem, hours of classes over a semester for most students.

Maybe eventually most people are doing “humans servicing other humans” jobs, which can’t be sped up. So average growth rates stay low. I’m guessing inflation too.

The article has that portion linked to a presentation that suggests that finance is not actually producing much net positive value for the overall economy.

Just because a field is full of highly compensated professionals doesn’t mean they are actually producing net positive outcomes. Hanson says this about medicine,for example, and one can make the argument for professors in many disciplines. And how about realtors?

Indeed, high incomes may just reflect successful rent-seeking or a survivorship bias in counting only the gains of those who won zero sum games.