The Economics of AGI: Scarcity, Labor, and Wealth Distribution

The Difficulty of Economic Forecasting

Alex Imas and Phil Trammell argue that individual economic forecasts regarding AI are unreliable due to the 'lump-of-labor' fallacy and the historical tendency for automation to create new, unforeseen categories of demand. They suggest that instead of relying on individual predictions, economists should utilize prediction markets and aggregate data to understand how labor and capital shares might shift. Historical precedents, such as the Industrial Revolution, show that while automation destroys specific tasks, it often leads to structural changes that maintain high employment rates in new sectors.

The Relational Sector and Human Scarcity

A central theme is the 'relational sector'—services where human involvement is an intrinsic part of the value proposition. Imas presents experimental evidence suggesting that consumers value human-produced goods (like art) differently than AI-produced ones, provided the human connection is perceived as unique. However, the speakers debate whether this sector can remain a significant portion of the economy. If AI can automate the entire supply chain for non-relational goods, the economy might shift toward a 'machine-only' loop where human labor becomes a negligible fraction of total output.

Capital Share vs. Labor Share

The speakers discuss the 'Kaldor facts,' noting that labor share has remained remarkably stable at roughly 60% for centuries despite massive technological advancement. They explore whether AGI represents a qualitative shift where the network-adjusted capital share could move toward 100%. Phil Trammell highlights that even if specific goods become fully automated, the economy might avoid satiation by constantly expanding the variety of goods and services demanded, thereby preventing the collapse of the labor share—provided that humans continue to find new, valuable tasks to perform.

Redistribution and the 'Messy Middle'

The conversation touches on the political and economic challenges of a 'Messy Middle' scenario, where AI automates jobs faster than it generates wealth that can be effectively redistributed. The speakers note that while the resources saved by automation technically exist, the political and logistical hurdles of compensating displaced workers—especially those in high-income brackets—create significant risks of instability. They emphasize the urgent need for better data on consumer demand elasticities and task-based job structures to prepare for these transitions.