Moravec's Paradox and Growth Under Artificial General Intelligence: Regime Switches, Baumol Traps, and the Reversal of the Wage Premium

This paper introduces Moravec's Paradox into a task‐based model of growth driven by artificial general intelligence. When physical‐task automation costs are orders of magnitude larger than cognitive‐task costs, the economy traverses a Moravec plateau during which cognitive work is automated but physical work remains human‐operated, stabilizing the labor share. Under CES aggregation, the plateau‐phase income distribution exhibits a regime switch at unit elasticity: complementarity generates a Baumol trap while substitutability erodes labor's share. With heterogeneous workers, the wage premium reverses; the reversal survives labor mobility between task types. These results qualify predictions of rapid labor‐share collapse in existing frameworks.