79 Comments

> But the fact that people are shifting their spending toward going out to eat more suggests that the burden is not as crushing as the WSJ’s viral chart suggests.

Not necessarily going out to eat more but likely paying more when going out. You didn’t assume that the slight rise in the percentage of income spent on groceries was people buying more groceries.

Expand full comment

Tribal lands as Charter Cities! :)

The next step would be to allow tribal groups to buy other land and take their tribal privileges with them. They could come in, develop an area using privilege to overcome NIMBY opposition and then sell out to go develope somewhere else!

Expand full comment

If AI reduces the productivity gap between low skilled and high skilled workers, wouldn't this increase the outsourcing of white collar professional jobs to countries like India and Vietnam?

Expand full comment

I read recently where one EU country, Denmark I think, is planning to give Ukraine all of Denmark's artillery. That seems like a sensible move to me, as artillery sitting in Denmark is pretty useless. Why not then send their pieces to the front now where it will do some lasting good? I presume that Denmark will then purchase new and improved artillery to replace its stocks. Denmark will have the luxury of being able to shop around for the best value and to wait for manufacturing to ramp up to meet this demand (after all, who is going to threaten Denmark's actual landmass, other than Russia, which is currently mired in eastern Ukraine). Ukraine does not have this luxury. If the Russians are soundly defeated this year, it will take them decades to be in a position to threaten Europe again. During that long lull, Europe can easily replace its stocks of weapons and munitions. Maybe France and Germany should consider doing the same, particularly France, which has a large mass of friendly neighbors between it and Russia.

Expand full comment

A near future sci-fi novel that has high tech companies owned by American indigenous tribes with cool ass names and new major cities on Tribal land sounds like such a neat world to explore

Expand full comment

With AI when I read “low-skill and high-skill” I just keep hearing “low-wage and high-wage”. I’m going to need a lot of convincing that AI, while not replacing people, is going to devalue them. Basic turning whole other sectors into McDonalds. People are necessary but largely disposable and thus low-wage.

Also, as someone in Residential Construction, the electrical panel example showed astonishing ignorance on the authors part, not really helping the case being made. Home Depot DIY culture has already made enough problems for professionals that have to come in and fix DIY disasters.

Expand full comment

> But this chart does two things for us. First of all, the y-axis starts at zero, so you can see how modest the recent increase is.

Great post as always, but the graph you show here does not start at zero, but at 0.1. If it starts at zero, I think your point will become even more visually clear.

Expand full comment

Canada's indigenous population is 4.5% of the population, compared to about 1.5% for the US. The difference in political power is even more pronounced, and that's led to greater FN self-governance and in some cases, greater business development maturity.

Expand full comment
Feb 28·edited Feb 28

So, I do extremely complex customer and field-tech support work.

I am somewhat worried that companies are going to think they can get away with replacing roles like mine with AI-assisted poorly trained n00bs. Right now the model is that you have less-educated, less-trained people working from scripts and flow-charts (many of which are being _written_ by people like me who have a deep theoretical understanding of the products we support and hard-won experience with tracing from patterns of symptoms to root causes) to resolve the more common problems, and then if the case defies solution by that kind of script-following, the case gets escalated.

If AI helps the lower-tier people triage things faster, that's great. I could even buy that AI may be useful in helping to sort through data on incidents across a large fleet (e.g. Tesla's many many thousands of PowerWalls) and identify classes of case where we need to refine the guidance for the lower-tier people.

But I think some companies may misguidedly think they can just take humans out of the loop entirely and let AI answer the calls, with no or minimal ability to escalate to a person. We already _see_ that kind of thing happening in some places, where there's some kind of robotic phone response, and negligible staffing if you try to escalate, so you sit on hold for an hour and then the system tells you, too bad, call volumes are higher than expected, call back some other time.

Expand full comment

Cross over with Slow Boring on parking: make descendents of formerly enslaved persons owners of urbans streets and roads so they can charge the parking fees and collect congestion tolls. win-win!

Expand full comment
Feb 28·edited Feb 28

Re: weapons. Does anyone know why this is? Last thing I read on this (which might have been here) European weapons factories were partly idle and waiting for work. Governments kept claiming to be ramping up aid, but somehow that wasn't translating into purchase orders. It was unclear why that is, though. If the UK is any guide it might be civil services that are refusing to implement orders they dislike?

Re: AI productivity. Economics is one of the better fields and I'm totally open to this possibility, but a gut reality check tells me to be careful here. Measuring productivity of individuals is notoriously difficult. The cited evidence boils down to two forms of work: writing English (customer support responses) and writing code. I absolutely 100% believe that gen AI can improve the productivity of customer support agents - people who churn out repetitive answers to things. It is politer, more helpful, more verbose, more knowledgeable and more verbally fluent than the average call center worker. Give the bottom rung a chance to outsource thinking to a machine and I can see how the results will improve because most support tickets are duplicates.

But it doesn't follow that this will rebuild the middle class; that's the sort of grandiose over-generalization that gives intellectualism a bad name. The problem here is that if AI can answer support requests then the very next step is to just replace the lower-than-average workers entirely. Today that isn't feasible because inferencing latency is too high to make AI fully fluent in [phone] conversation, and companies don't trust AI to make decisions yet (see how Air Canada just got burned by a hallucinated policy). But TTS is now human quality or higher, and Groq is demonstrating real-time low latency inference, and the hallucination problem is becoming better understood/more solvable. At this point I think we're only a couple of years away from callcenters that handle almost all calls/chats purely with AI and only a handful of highly trained/trusted managers, whose job is exclusively resolving obscure edge cases. This will increase western productivity numbers because their cost base will go down whilst output remains the same or goes up (people will use support more), and decrease poor country productivity numbers because the agents that get laid off will stop producing entirely.

W.R.T. generalizing this to all jobs, I feel deeply uncomfortable about that. I can believe the general thrust is true _at the moment_, because the easiest way to integrate AI into a workflow today is to give existing workers suggestions, and that will boost weak performers in the short run. But again, the question is what happens next? I read a lot of coding gen AI papers and the way things are going here seems clear enough to me - AI will become increasingly able to plan out and execute whole sub-projects independently, and then whole projects. That means the work of using gen AI will shift from reviewing small and simple auto-completions to figuring out what's possible and what exactly should be built, whilst simultaneously reviewing/fixing the hardest parts where the AI training set is of no use. That's usually considered the most senior work! So again, I think we see that whilst currently AI is a boost to lower performers, very near term and small improvements could easily just obsolete the lower performers leaving all the remaining work to the people on the very leading edge of the profession.

That said, I'm not a doomer about these things. There's just so much freaking work to do out there, so much potential that goes unharvested, and so much of the dysfunction and effort in the software industry goes into working around the very high cost of labor, that we could easily 100x our productivity and still have plenty of employment in the sector. The way the job looks might change a lot, but the existence of the job won't.

Expand full comment
Feb 28·edited Feb 28

So what's the hold-up with Europe? Where is the kink in the line between words & actions? Is it that their gov'ts are not following through with purchase orders? Or that the manufacturing capacity simply no longer exists?

If the former, why? Is it because their welfare states are so expensive, that with a population significantly older on average than it was in 1989 they can no longer afford to maintain their welfare states AND spend more on defense? If this is the case, I hate to break it to you, but not a single German or Frenchman is going to agree to any reduction in lifestyle in order to fund the military when Russia almost certainly cannot and will not conquer past Poland. Hell, just look at the revolts that break out in Europe now periodically every time they try to remove subsidies for farmers. The attitude is, if the Russians take eastern Europe, so what? They had it all during the Cold War and we did just fine then.

If the holdup is point #2, lack of physical plant, then you're still up shit creek because the physical capital no longer exists and will have to be rebuilt, necessitating lots of money (which takes you back to point #1). In this scenario, Denmark is particularly foolish for donating all of their artillery. Even if they place the orders for new kit today, when is it going to be delivered? In five years? Taiwan is already waiting for back-filled orders!

When the West decided that heavy manufacturing was gauche and that civilized countries should financialize and move their economies over to services, it basically decided to gamble on there never being large-scale great power conflict ever again. Gambled and lost, it would seem.

Expand full comment

Have you seen Klarna's data yet on customer service job replacements? AI replaced 700 live chat agents completely and is getting better scores. It's possible online customer service as a job doesn't exist in 5 years.

Expand full comment

It’s fascinating to watch France say a lot while doing so little

Expand full comment

I’m bullish on the idea that AI ends up being more helpful for the middle class/consumers than the worst case scenarios that get a lot of focus, but complex medical procedures like (vascular) catheterization are one of the worst examples I can think of. I perform medical procedures part time and can tell you that a lot of procedural competency is just reps, not lack of information or cognitive issues. I explain to new trainees that the learning process js more like driving than textbook learning or even other areas of clinical practice.

What’s odd to me is there are many areas of clinical medicine that have become algorithmic in a way that would lend itself to an AI takeover. But while I could see some patients preferring the expediency and convenience of an AI doctor for straightforward decisions, I think most will continue to want the more nuanced expertise and plans developed by humans.

Expand full comment

Any chance of getting Econ 102 (and Hexapodia) into written form? Like have an AI assistant listen and write up a summary? :)

Expand full comment