The relational sector framing is the useful hinge, but it smuggles in an assumption worth testing: that buyers keep paying the human-in-the-loop premium once the automated version crosses good enough. Recorded music, ATMs, and self-checkout all show the premium erodes fast for most goods and survives only where the human presence is the product itself, not a quality signal. The scarce thing may not be human labor in the loop, it's credible proof that one was, which makes verification the part that accrues value.
"If the consumer is willing to pay more for a product or service where every single task is automated except for that one part where the doctor is delivering the diagnosis and providing support, we would call that job part of the relational sector. People are willing to pay more for the human to stay in the loop in the job."
In the current scenario, many people would be willing to pay more for an additional 5 minutes with the doctor but institutional constraints prevent bidding for the doctor's time. The doctor's time is scarce right now because they're doing all this other stuff. If the other stuff is automated, there is much less scarcity of the doctor's time - suddenly you're going to have all these doctors with a lot of time to consult with patients! The focus is always on elasticity of demand, but I think the supply side is also going to matter.
Imas, in contrasting someone who'll choose an AI with someone who'll prefer a human, says,
"Which of those two people are going to reproduce, find a mate, all of these sorts of things? I think the answer is clear. It's the second one that has the preference for other people."
He gets the selection backwards. The trait that actually predicts reproduction isn't a preference for humans over machines—it's the competitive status drive, because status is a key mate-value signal. A man maximizing status through resource acquisition will use whatever service helps him win this contest—AI or human. He may choose AI not because he's dropping out of the mating race, but because he wants to win it. We don't know if Elon Musk likes people—we do know he likes having children. Imas pinned reproduction on preference for human interaction when it actually runs on the status drive—two different traits, and he picked the wrong one.
I agree that AI will help you get ahead and maximize status in many ways, but one of the main ways to gain status, whether socially or monetarily, is through relationships. If everyone is using AI all the time, which is what we see happening with most jobs not even just software, then optimizing for relationships becomes an even stronger differentiating trait
Makes sense. I don’t think there’s one best choice between human and AI. My point was that picking AI doesn’t necessarily mean you’re asocial. In some cases picking AI might maximize social opportunities.
Like with increasing costs of basic stuff, the question is if there will be sufficient salary for labour. I m pessimistic with actual economic settings. I think the tax for wealthy, for robots and ai is much needed to secure decent living for people in general. Cause now it is really difficult time for ordinary people and it seems it is not improving. Basically, I think the we as people dont need billionaires/trillionaires. So yes, ai has potential to be amazing. We just need to limit the possibilities to increase probability of desired outcome. Which is what the tech billionaires now trying to achieve.
"It’s more difficult to imagine a good thing that doesn’t exist than losing something that exists. It’s much easier for somebody to go on a podcast and say, “These jobs that you like, they’re going away,” than for somebody to spin up a utopia which doesn’t exist yet."
I think this also points to negativity bias - we are much more inclined to pay attention to potential negative scenarios than positive.
idk why he's saying ppl are losing their jobs to AI that's not really true. a lot of companies have said we'll just do it now than later due to whatever reasons..we need to be more efficient..how do we know that a lot that haven't been cutting it haven't?
On the Messy Middle: it feels like Trammell and Imas dismiss this - or at least the political economy of this - possibility too quickly. AI could automate enough jobs such that the pie is growing fast and there is a lot of wealth creation, but the gains in the pie accrue largely to holders of capital. Even if AI becomes commoditised, those gains might accrue mainly to those holding land or other forms of capital (e.g. IP) that AI cannot readily substitute for. While in theory, the gains could be redistributed from those holders of capital to the "losers" from AI to get a Pareto-optimal improvement, the political economy of many countries may mean that redistribution will not flow readily to the "losers".
There are, after all, many potential Pareto-optimal improvements that do not survive contact with real world politics.
The relational sector framing provides empirical grounding to something I've been exploring narratively. The art print experiment — higher willingness to pay when a human made it — is Rory Sutherland's doorman question in miniature.
A consultant sees the doorman. Defines the job as "opening the door." Replaces him with an automatic door. Claims the cost saving. Leaves. Five years later, the hotel has declined. Nobody connects the two.
We know relational value exists — and not just for 'luxury goods'. Will organizations and policymakers recognize this before it gets automated away?
Hi Dwarkesh, Haven't listened yet but dealing with extreme inequality raises the issue of fair and efficient ways to redistribute income at scale from capital to labor. The theoretically best answer, long known to econnomists at least since Keynes, is via something called a graduated expenditure tax. Here is one version, maybe the best possible version for pretty obvious reasons: https://shorturl.at/a6wqw
The link to the QJE paper doesn't work - here it is :) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3722562
The relational sector framing is the useful hinge, but it smuggles in an assumption worth testing: that buyers keep paying the human-in-the-loop premium once the automated version crosses good enough. Recorded music, ATMs, and self-checkout all show the premium erodes fast for most goods and survives only where the human presence is the product itself, not a quality signal. The scarce thing may not be human labor in the loop, it's credible proof that one was, which makes verification the part that accrues value.
"If the consumer is willing to pay more for a product or service where every single task is automated except for that one part where the doctor is delivering the diagnosis and providing support, we would call that job part of the relational sector. People are willing to pay more for the human to stay in the loop in the job."
In the current scenario, many people would be willing to pay more for an additional 5 minutes with the doctor but institutional constraints prevent bidding for the doctor's time. The doctor's time is scarce right now because they're doing all this other stuff. If the other stuff is automated, there is much less scarcity of the doctor's time - suddenly you're going to have all these doctors with a lot of time to consult with patients! The focus is always on elasticity of demand, but I think the supply side is also going to matter.
Imas, in contrasting someone who'll choose an AI with someone who'll prefer a human, says,
"Which of those two people are going to reproduce, find a mate, all of these sorts of things? I think the answer is clear. It's the second one that has the preference for other people."
He gets the selection backwards. The trait that actually predicts reproduction isn't a preference for humans over machines—it's the competitive status drive, because status is a key mate-value signal. A man maximizing status through resource acquisition will use whatever service helps him win this contest—AI or human. He may choose AI not because he's dropping out of the mating race, but because he wants to win it. We don't know if Elon Musk likes people—we do know he likes having children. Imas pinned reproduction on preference for human interaction when it actually runs on the status drive—two different traits, and he picked the wrong one.
I agree that AI will help you get ahead and maximize status in many ways, but one of the main ways to gain status, whether socially or monetarily, is through relationships. If everyone is using AI all the time, which is what we see happening with most jobs not even just software, then optimizing for relationships becomes an even stronger differentiating trait
Makes sense. I don’t think there’s one best choice between human and AI. My point was that picking AI doesn’t necessarily mean you’re asocial. In some cases picking AI might maximize social opportunities.
Loved this conversation and did a face off @Alex Imas vs @Will Manidis a few weeks back: https://www.thecuriousmind.org/p/unrepeatable
Like with increasing costs of basic stuff, the question is if there will be sufficient salary for labour. I m pessimistic with actual economic settings. I think the tax for wealthy, for robots and ai is much needed to secure decent living for people in general. Cause now it is really difficult time for ordinary people and it seems it is not improving. Basically, I think the we as people dont need billionaires/trillionaires. So yes, ai has potential to be amazing. We just need to limit the possibilities to increase probability of desired outcome. Which is what the tech billionaires now trying to achieve.
On this quote by Alex Imas:
"It’s more difficult to imagine a good thing that doesn’t exist than losing something that exists. It’s much easier for somebody to go on a podcast and say, “These jobs that you like, they’re going away,” than for somebody to spin up a utopia which doesn’t exist yet."
I think this also points to negativity bias - we are much more inclined to pay attention to potential negative scenarios than positive.
idk why he's saying ppl are losing their jobs to AI that's not really true. a lot of companies have said we'll just do it now than later due to whatever reasons..we need to be more efficient..how do we know that a lot that haven't been cutting it haven't?
On the Messy Middle: it feels like Trammell and Imas dismiss this - or at least the political economy of this - possibility too quickly. AI could automate enough jobs such that the pie is growing fast and there is a lot of wealth creation, but the gains in the pie accrue largely to holders of capital. Even if AI becomes commoditised, those gains might accrue mainly to those holding land or other forms of capital (e.g. IP) that AI cannot readily substitute for. While in theory, the gains could be redistributed from those holders of capital to the "losers" from AI to get a Pareto-optimal improvement, the political economy of many countries may mean that redistribution will not flow readily to the "losers".
There are, after all, many potential Pareto-optimal improvements that do not survive contact with real world politics.
The relational sector framing provides empirical grounding to something I've been exploring narratively. The art print experiment — higher willingness to pay when a human made it — is Rory Sutherland's doorman question in miniature.
A consultant sees the doorman. Defines the job as "opening the door." Replaces him with an automatic door. Claims the cost saving. Leaves. Five years later, the hotel has declined. Nobody connects the two.
We know relational value exists — and not just for 'luxury goods'. Will organizations and policymakers recognize this before it gets automated away?
Wrote something adjacent: https://rajeshachanta.substack.com/p/the-doorman-and-deepseek
If abundance doesn't generate negative economic growth, then is the problem simply one of redistribution of wealth?
Hi Dwarkesh, Haven't listened yet but dealing with extreme inequality raises the issue of fair and efficient ways to redistribute income at scale from capital to labor. The theoretically best answer, long known to econnomists at least since Keynes, is via something called a graduated expenditure tax. Here is one version, maybe the best possible version for pretty obvious reasons: https://shorturl.at/a6wqw
Enjoy.
04:27 A second peak is a double top.