Alex Imas and Phil Trammell – What remains scarce after AGI?

What It Covers

Alex Imas (Google DeepMind / University of Chicago) and Phil Trammell (Stanford / EPoC) examine what remains scarce after AGI arrives, analyzing labor share stability, the "relational sector" where human involvement creates value, redistribution mechanisms including universal basic capital, and why developing nations should prioritize indexing AI returns over retraining programs.

Key Questions Answered

• •Labor Share Stability: Despite 200 years of industrial automation, human wages have consistently captured over 60% of total economic output. Some economists argue that when accounting methods are held constant, labor share has never meaningfully declined. This historical resilience suggests automation alone does not mechanically destroy labor's share — but AGI may represent a qualitative break if entire supply chains become fully automatable without any human input at any stage.
• •Relational Sector Valuation: Experimental data shows consumers pay significantly more for goods produced by a single human artist versus AI, but that premium collapses when 500 human-made copies exist. This suggests human-intrinsic value is tied to scarcity and connection, not just origin. To forecast which jobs survive automation, researchers need conjoint analysis measuring willingness-to-pay when specific tasks shift from human to machine — data that currently does not exist at scale.
• •Increasing Variety Prevents Satiation: The Mongolian economist thought experiment illustrates a core forecasting failure: holding product variety fixed while projecting automation effects. Just as horse-transport satiation never crushed singer employment because new goods emerged, AI may continuously generate new capital varieties that prevent demand satiation. GPU rental costs have actually risen despite massive compute expansion, because new AI use cases absorb supply faster than production scales — a direct parallel.
• •Messy Middle Risk: The most politically dangerous automation scenario is not mass unemployment but gradual displacement into lower-wage roles — mirroring the 1920–1940 telephone operator transition, which took 20 years despite available technology. A 2–3% unemployment spike triggers emergency fiscal response, but slow underemployment does not. Policymakers should monitor this drip scenario specifically, as it produces political instability without triggering the automatic stabilizers designed for acute economic shocks.
• •Redistribution Mechanism Trade-offs: Universal basic capital — distributing ownership shares rather than cash — avoids the political vulnerability of UBI, where benefit levels depend on who holds power. The core obstacle is indexing: if Anthropic collapses while an unknown robotics firm captures value, poorly targeted portfolios fail. A consumption tax funding broad equity purchases (similar to the original Social Security privatization proposal) offers one mechanism, but concentrated private AI companies make indexing harder than during the index-fund era.
• •Developing Nation Strategy: Countries outside the AI hardware and model production chain — not producing chips, HBM memory, EUV lithography, or frontier models — face two scenarios: AI diffuses broadly like electricity, making S&P-style indexing sufficient, or returns concentrate in private labs, requiring direct equity stakes in those specific companies. Purchasing diversified AI-adjacent equity now is a higher-priority strategy than retraining programs, though leapfrogging effects (as seen with mobile banking in Nigeria) remain a secondary possibility.