AGI won't automate most jobs–because they're not worth the trouble (fortune.com) AI

A Yale economist argues that in an AGI era most jobs may not be automated because replacing people is not worth the compute cost, even if the systems could do it. Instead, compute would be directed to “bottleneck” work tied to long-run growth, while more “supplementary” roles like hospitality or customer-facing jobs may persist. The paper warns that automation could still reduce labor’s share of income and shift gains to owners of computing resources, making inequality the central political issue during the transition.