Alex Imas and Phil Trammell – What remains scarce after AGI?
Economists Alex Imas and Phil Trammell discuss what will remain scarce after AGI, covering labor share stability, the 'relational sector,' wealth redistribution mechanisms, and implications for developing countries. They explore historical parallels to industrial automation, the plausibility of various economic scenarios, and why negative economic growth from AI abundance is theoretically very difficult to achieve.
Summary
The conversation opens with a central question: what kinds of things will remain scarce in a world of increasing automation and advanced AI? The economists identify the 'relational sector' as a candidate — services where human involvement is intrinsically part of the value (e.g., a human barista, doctor, or performer). However, they note that whether this sector can sustain significant labor share depends on empirical data about consumer demand elasticities that largely does not yet exist.
A significant portion of the discussion focuses on the historical stability of labor share — the roughly 60%+ of GDP paid out as wages — which has remained surprisingly constant through previous waves of automation including the industrial revolution. They reference the 'lump of labor fallacy' via David Ricardo's failed predictions, arguing that economists have been consistently poor at forecasting labor market outcomes because new job categories and goods emerge in unpredictable ways. They advocate for scenario-based modeling and prediction markets over individual forecasts.
The economists discuss the concept of 'network-adjusted factor shares,' noting that even seemingly automated sectors still embed significant labor value when you trace the full supply chain. They acknowledge a qualitative shift is coming where some goods may have their entire supply chain automated, but argue the implications for overall labor share are ambiguous because of increasing variety in capital goods and potential satiation dynamics.
On current evidence of AI-driven unemployment, they find little empirical support for a 'white collar bloodbath,' citing Yale Budget Lab research showing only marginal signals like slightly below-trend growth in junior developer hiring, with demand for senior engineers actually increasing. They attribute much anecdotal evidence of AI layoffs to narrative-driven behavior by firms wanting to appear AI-forward.
The 'messy middle' scenario — where AI automates enough jobs to cause political disruption but not enough wealth creation for broad redistribution — is discussed as theoretically possible but practically narrow. They argue that if AI can automate enough white-collar work to cause mass unemployment, the productivity gains would also be large enough to fund redistribution, making the scenario where automation occurs without significant pie growth quite unlikely.
On redistribution mechanisms, they discuss negative income tax, UBI, universal basic capital, consumption taxes, and wealth taxes, each with tradeoffs. A key concern raised is the political economy risk of dependency on government transfers versus asset-based redistribution. They also note the difficulty of indexing the AI economy, comparing scenarios where AI resembles electricity (broad diffusion of gains) versus social media (rents captured by platforms).
For developing countries like Nigeria or India, they argue the most promising strategy is likely buying into the AI economy through indexing rather than retraining programs, though they acknowledge the difficulty if AI gains remain concentrated in private companies. They express optimism that commoditization of frontier AI and eventual public listings would make broad-based indexing more feasible.
The conversation closes with reflections on the long-run evolution of preferences — whether the 'greedy optimizer' archetype of wealthy individuals who don't satiate in capital accumulation will come to dominate the economy through selection effects — and the safety tradeoffs of concentrated versus commoditized frontier AI development.
About this episode
<p>Economics of AGI episode w <a href="https://www.aleximas.com/" target="_blank">Alex Imas</a> and <a href="https://philiptrammell.com/" target="_blank">Phil Trammell</a>.</p><p>There’s a bunch of important questions about how we deal with AI that only economics can answer.</p><p>What is the optimal way to tax and redistribute the wealth that will be generated? How should countries not in the AI supply chain index into the gains? Is there any world where inequality doesn’t explode?</p><p>It might seem like these questions have obvious answers, but the first thing economics teaches you is that your intuitions can often be entirely wrong.</p><p>It was very helpful to chat through these things with Alex and Phil.</p><p>Watch on <a href="https://youtu.be/Jj-kBHzUohs" target="_blank">YouTube</a>; read the <a href="https://www.dwarkesh.com/p/alex-imas-phil-trammell" target="_blank">transcript</a>.</p><p><strong>Sponsors</strong></p><p><a href="https://janestreet.com/dwarkesh" target="_blank">Jane Street</a> invests heavily in turning smart people into exceptional researchers and engineers. In addition to their apprenticeship model, Jane Street runs lectures and bootcamps in their in-office classrooms -- managers clear their teams’ schedules to encourage attendance. If you’d like to work at a place that takes learning this seriously, Jane Street is hiring. Check out their open roles at <a href="https://janestreet.com/dwarkesh" target="_blank">janestreet.com/dwarkesh</a></p><p><a href="https://gemini.google" target="_blank">Google’s Gemini Omni</a> has incredible video editing capabilities -- you can upload a video and have Omni change the background, adjust lighting, or add specific elements. But Omni is also a preview of how future frontier models will be trained -- fully multimodal on both input and output. You can try it yourself in the Gemini app at <a href="https://gemini.google" target="_blank">gemini.google</a> or in Flow at <a href="https://flow.google" target="_blank">flow.google</a></p><p><a href="https://cursor.com/dwarkesh" target="_blank">Cursor</a> used targeted RL with textual feedback to help train their Composer 2.5 model. One of their researchers, Sasha Rush, gave me an impromptu blackboard lecture to explain how this form of on-policy self-distillation works -- I posted the full thing on X. If you want to try Composer 2.5, go to <a href="https://cursor.com/dwarkesh" target="_blank">cursor.com/dwarkesh</a></p><p>Timestamps</p><p>(00:00:00) – Will capital share increase?</p><p>(00:19:36) – Messy Middle scenario</p><p>(00:25:57) – How to tax and redistribute AI wealth</p><p>(00:30:02) – Why demand collapse is unlikely</p><p>(00:39:26) – Human employees would be hard to integrate into the machine economy</p><p>(00:43:08) – What if some humans (or AIs) value wealth accumulation intrinsically?</p><p>(01:01:28) – What should developing countries do?</p> <br /><br />Get full access to Dwarkesh Podcast at <a href="https://www.dwarkesh.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4">www.dwarkesh.com/subscribe</a>
Key Insights
- Imas argues that labor share has remained above 60% through all prior waves of automation, which he calls 'incredibly surprising' and notes some economists suspect it may be an accounting artifact rather than a genuine economic constant.
- Trammell points out that nothing has yet been 'completely automated' in a network-adjusted sense — when you trace the full supply chain of even highly automated goods like electronics, labor still contributes roughly 50% of value.
- Imas uses the Mongolian economist thought experiment to argue that forecasters consistently fail because they hold the variety of goods fixed, missing that automation enables entirely new categories of consumption that prevent satiation in capital goods.
- Trammell argues that the conditions required for negative economic growth under AGI — specifically that wealthy capital holders would stop investing and their demand would be hard-bounded — are extremely implausible, since investment in data centers and infrastructure would continue regardless of consumption satiation.
- Imas argues the 'messy middle' scenario of gradual automation without mass unemployment is politically the most dangerous, because it produces underemployment and wage decline without triggering the kind of fast unemployment spike that historically prompts large fiscal responses.
- Imas and Trammell both argue that the scenario where AI creates just enough automation to displace workers but insufficient wealth growth to fund redistribution requires a very narrow and unlikely set of conditions — specifically that automating entire white-collar sectors somehow doesn't expand the economic frontier.
- Trammell introduces the concept of investment-specific technical change — where the price of capital falls relative to consumption goods faster than the capital stock grows — as the mechanism by which explosive AI growth might not straightforwardly translate into high real interest rates or growing labor share.
- Imas argues from experimental data that human-produced goods command significantly higher willingness-to-pay than AI-produced goods, but only when scarcity is preserved — a single human-made art print is valued much higher than AI-made, but 500 human-made prints lose that premium entirely.
- Imas suggests that an evolutionary argument supports persistence of human-preference for human-intrinsic services: people with strong preferences for human social interaction are more likely to find mates and reproduce, potentially selecting for rather than against these preferences over time.
- Trammell argues that the historical failure of wealthy dynasties to maintain disproportionate economic control is largely due to 'dissipation shocks' — heirs squander wealth, foundations spend it down — and that without mortality and without these dissipation mechanisms, selection for capital-accumulating agents could be much more persistent.
- Imas and Trammell argue that for developing countries, buying into the AI economy through indexing (sovereign wealth funds purchasing AI company shares) is likely more promising than retraining programs, but this strategy only works cleanly if AI gains diffuse broadly like electricity rather than concentrating in a few private platforms like social media.
- Trammell raises the point that commoditization of frontier AI — while potentially beneficial for broad distribution of gains — creates safety tradeoffs by intensifying competitive dynamics and making it harder for any single lab to slow down for safety reasons, though he ultimately suggests the safety benefits of diffusion may outweigh this risk.
Topics
Transcript
Today, I'm chatting with Alex Emas, who is Director of AGI Economics at Google DeepMind and Professor of Economics at University of Chicago, and Phil Trammell, who is Head of Economics at Epoch and Research Scholar at Stanford. In general, in this interview, what I want to understand is what economics tells us about what we can expect in a world with more and more automation, more and more advanced AI, what that tells us about what will happen to wages, to labor share, what the best way to tax and redistribute the wealth that we generated as a result of AGI will be, and what kinds of things will be scarce, because what is scarce kind of tells you…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Dwarkesh Podcast
Why smarter AI models could drive up compute prices 10x
As AI models become more capable and valuable, compute capacity growth (3x yearly) cannot keep pace with revenue growth (10x yearly), forcing labs to either increase margins, raise compute prices, or shift more resources to inference. The speaker argues compute prices will likely increase significantly as AI models approach human-level capabilities, making compute a scarce resource similar to skilled labor.
Adam Brown – Einstein's happiest thought: General Relativity from scratch
Adam Brown explains Einstein's General Relativity from first principles, showing how the equivalence of inertial and gravitational mass led Einstein to conceptualize gravity as curved spacetime rather than a force. The lecture progresses from special relativity through the geometric interpretation of gravity to black holes, demonstrating GR's explanatory power across vastly different scales.
Grant Sanderson – AI and the future of math
The discussion centers on the rapid advancements of AI in mathematics, exploring its implications for the future of math and related fields. The conversation highlights how AI's capabilities impact traditional mathematical roles, the process of knowledge creation, and the potential for new insights in various domains.
The next big breakthrough will be AIs learning on the job
The speaker discusses how AI labs are betting on reinforcement learning from verified rollouts (RLVR) to achieve AGI, but argues this approach has fundamental limitations. He contends that true general intelligence requires continual on-the-job learning through weight updates, which current scaling paradigms don't adequately address.
The data black hole at the center of AI
The transcript argues that AI's primary driver of progress is data quantity and quality rather than architectural improvements or scaling, highlighting a massive gap in sample efficiency between humans and AI models. The speaker contends that current AI systems are fundamentally different from human intelligence, requiring orders of magnitude more data to learn skills. Despite this inefficiency, AI can still automate white-collar work due to the economics of scale and parallelism.