I agree with your top line point about being open to weirdness, but generally see most "AGI scene" people as being insufficiently economically aware and think they would strongly benefit from moving closer to Alex's position.
> Essentially, Jason gives three arguments:
> 1. Many of the jobs Alex lists have a smaller relational component than he suggests (see Trevor).
> 2. A small fraction are indeed purely relational, but since superhuman AI would dramatically shift consumption patterns, demand for those jobs could still collapse (see Philip).
> 3. Superhuman AI could reshape the human brain, drastically reducing our interest in relational services.
I don't really buy 1, and discussed with Trevor a bit here: https://x.com/herbiebradley/status/2046352086313668865. I haven't heard a satisfactory counterargument to the point that the AI-only product has to compete with AI + human, so the quality baseline is raised anyway. For therapists, don't you basically have to assume that there are no inherently-human tasks in the job of "therapist" to suppose that AI can be a full substitute (implicit in saying "better")? It's an *adjacent* product which fills some of the market demand but does not substitute.
On 2, not convinced it's a very small fraction. But on collapse, Alex's quote here directly argues against Phillip: "Many haven’t been invented yet, just as six out of ten jobs people hold today didn’t exist in 1940.". He's arguing that part of the expanding varieties is the expansion in relational service jobs. I haven't yet seen a good argument against the reasonable point that we should expect many more new such jobs to appear!
On 3, interesting take but seems far far too uncertain, even as someone AGI pilled, to really incorporate into a model here.
I don't think my intuition here rests on Alex's suggestion that many white-collar jobs have a core and underappreciated relational component, but I also think that's probably true and those who haven't worked in a large bureaucracy and significantly interacted with internal politics wildly, wildly underestimate it. Sometimes you can directly see the issue: people who have only ever worked as SWEs or AI researchers actually think that most jobs are highly intelligence loaded.
On the point that point that we should expect many more new such jobs to appear, I didn't respond to Alex in that comment thread because the point of my original comment was that the question of what will happen over time, as we invent new goods (both labor-intensive and not), is not answered one way or the other by what rich people today spend their money on. I kept trying to stick to that point, and it seemed to me that Alex kept changing the subject. I’m still not sure why we kept failing to reach agreement on that one point.
Some (compressed) thoughts on whether new human-intrinsic goods will keep the labor share high, offsetting the force of new capital-produced goods, are in footnote 12 of my own essay:
> “Likewise, we may invent new human-intrinsic goods, like new kinds of performances for people to engage in, with or without the assistance of capital-intensive technology. But the possible space of such goods is fundamentally constrained by what a human can do, whereas the space of goods producible by all possible superintelligent machines is practically unlimited. This asymmetry is in some sense the converse of the asymmetry that appears in many economic models of automation, in which labor can perform all tasks and capital can only perform a subset.”
One way to put the thought is, the way I see it, we’ll probably reach “technological maturity” on what can be done with the human brain and body long before we reach technological maturity in general.
To your point, isn’t #1 basically orthogonal to the last mile problem? Yes, in absolute terms, perhaps these jobs aren’t as relational as proposed, but there is a relational subset or capacity to them that will be a bit of a differentiator and will become more important (perhaps) as the rest of the job gets automated.
I agree with your top line point about being open to weirdness, but generally see most "AGI scene" people as being insufficiently economically aware and think they would strongly benefit from moving closer to Alex's position.
> Essentially, Jason gives three arguments:
> 1. Many of the jobs Alex lists have a smaller relational component than he suggests (see Trevor).
> 2. A small fraction are indeed purely relational, but since superhuman AI would dramatically shift consumption patterns, demand for those jobs could still collapse (see Philip).
> 3. Superhuman AI could reshape the human brain, drastically reducing our interest in relational services.
I don't really buy 1, and discussed with Trevor a bit here: https://x.com/herbiebradley/status/2046352086313668865. I haven't heard a satisfactory counterargument to the point that the AI-only product has to compete with AI + human, so the quality baseline is raised anyway. For therapists, don't you basically have to assume that there are no inherently-human tasks in the job of "therapist" to suppose that AI can be a full substitute (implicit in saying "better")? It's an *adjacent* product which fills some of the market demand but does not substitute.
On 2, not convinced it's a very small fraction. But on collapse, Alex's quote here directly argues against Phillip: "Many haven’t been invented yet, just as six out of ten jobs people hold today didn’t exist in 1940.". He's arguing that part of the expanding varieties is the expansion in relational service jobs. I haven't yet seen a good argument against the reasonable point that we should expect many more new such jobs to appear!
On 3, interesting take but seems far far too uncertain, even as someone AGI pilled, to really incorporate into a model here.
I don't think my intuition here rests on Alex's suggestion that many white-collar jobs have a core and underappreciated relational component, but I also think that's probably true and those who haven't worked in a large bureaucracy and significantly interacted with internal politics wildly, wildly underestimate it. Sometimes you can directly see the issue: people who have only ever worked as SWEs or AI researchers actually think that most jobs are highly intelligence loaded.
On the point that point that we should expect many more new such jobs to appear, I didn't respond to Alex in that comment thread because the point of my original comment was that the question of what will happen over time, as we invent new goods (both labor-intensive and not), is not answered one way or the other by what rich people today spend their money on. I kept trying to stick to that point, and it seemed to me that Alex kept changing the subject. I’m still not sure why we kept failing to reach agreement on that one point.
Some (compressed) thoughts on whether new human-intrinsic goods will keep the labor share high, offsetting the force of new capital-produced goods, are in footnote 12 of my own essay:
> “Likewise, we may invent new human-intrinsic goods, like new kinds of performances for people to engage in, with or without the assistance of capital-intensive technology. But the possible space of such goods is fundamentally constrained by what a human can do, whereas the space of goods producible by all possible superintelligent machines is practically unlimited. This asymmetry is in some sense the converse of the asymmetry that appears in many economic models of automation, in which labor can perform all tasks and capital can only perform a subset.”
One way to put the thought is, the way I see it, we’ll probably reach “technological maturity” on what can be done with the human brain and body long before we reach technological maturity in general.
To your point, isn’t #1 basically orthogonal to the last mile problem? Yes, in absolute terms, perhaps these jobs aren’t as relational as proposed, but there is a relational subset or capacity to them that will be a bit of a differentiator and will become more important (perhaps) as the rest of the job gets automated.
Yes, or one could also view it via Gans & Goldfarb's "focus effect" here: https://www.nber.org/papers/w34639
where the automation frees up time of the human to focus more on less automatable tasks, and so raises the quality of them.