Dwarkesh PodcastWhat remains scarce after AGI? – Alex Imas and Phil Trammell
At a glance
WHAT IT’S REALLY ABOUT
Post-AGI economics: scarcity, labor share, redistribution, and global inclusion risks
- The conversation frames “what remains scarce after AGI” as a question about which preferences and constraints still bind when production can be machine-run end-to-end, highlighting a possible “relational sector” where humans are valued intrinsically rather than instrumentally.
- Rather than confident point forecasts about labor share or unemployment, the speakers argue for scenario analysis because key empirical inputs—especially demand elasticities and willingness-to-pay for “human-in-the-loop” services—are missing or poorly measured.
- They explore why overall demand collapse and negative economic growth are unlikely under transformative AI, since displaced income tends to reappear as investment or other spending unless demand is implausibly hard-bounded and investment does not absorb savings.
- Policy discussion centers on how to tax and redistribute AI wealth (negative income tax/UBI vs universal basic capital and broader tax bases), emphasizing political economy risks, indexing/targeting challenges, and the possibility that returns concentrate in hard-to-access private assets.
- For developing countries outside the AI supply chain, the key risk is being left behind if AI rents concentrate; the best hedge may be financial indexing/sovereign wealth exposure to AI-driven returns, though this depends on whether AI behaves more like electricity (diffuse gains) or social media (platform rents).
IDEAS WORTH REMEMBERING
5 ideasScarcity after AGI may be preference-based, not production-based.
Imas argues that even if machines can produce most goods, humans remain scarce and may be valued intrinsically in “relational” services where replacing a human with an AI lowers willingness-to-pay (e.g., trust, empathy, authenticity).
Forecasting labor share is less useful than mapping scenarios and measuring the right parameters.
The speakers stress economists have historically disagreed and often been wrong; progress depends on collecting missing data (e.g., task-level substitution, consumer willingness-to-pay for human-in-the-loop, and demand elasticities by sector).
Variety growth can keep capital-demand high even if existing goods satiate.
Trammell’s “Mongolian economist” analogy: if you hold the set of goods fixed, you predict satiation and spending shifts to human performers, but historically new categories of consumption expanded, keeping “non-relational” spending large.
A ‘demand collapse’ recession from automation requires extreme, unlikely assumptions.
Imas argues negative growth needs rich capital owners to stop consuming and also not reinvest, implying a hard bound on demand plus failure of investment absorption—implausible in a world where AI makes new investments (fabs, datacenters, robotics) highly valuable.
Short-run disruption is more plausible as a political economy problem than a ‘no wealth to redistribute’ problem.
The “Messy Middle” risk is gradual displacement, underemployment, and wage declines that don’t trigger emergency policy the way a sudden unemployment spike would; the challenge is governance, targeting, and legitimacy of transfers, not the existence of resources.
WORDS WORTH SAVING
5 quotesIf you don't take anything out of this conversation from me, we don't have any data.
— Alex Imas
I like the pessimistic framing of Moore's law is every eighteen months, the value of computation halves.
— Phil Trammell
Right now, we have, we don't really have, uh, any evidence of a white-collar bloodbath.
— Alex Imas
For abundance to generate negative economic growth, that's really hard to calculate.
— Alex Imas
If it's... So think about ComEd or Com Edison, whatever, whatever the electricity provider here is. It's a monopoly. It provides a resource that everybody uses, but do we think about electricity as, like, generate, creating concentration of power and is ComEd, like, having, like, this huge amount of political power, social power or something like that? No, because a lot, with electricity, a lot of the downstream benefits actually came to, like, the users of the electricity rather than the, rather than the actual entity producing the electricity.
— Alex Imas
High quality AI-generated summary created from speaker-labeled transcript.
