Matthew Barnett's article, "AGI Could Drive Wages Below Subsistence Level", explores the potential economic consequences of artificial general intelligence (AGI) on human wages. Unlike previous technological advancements that increased productivity and improved living standards, AGI could disrupt labor markets by fully substituting for human workers across all tasks. Using economic models such as the Cobb-Douglas production function, Barnett argues that an abundance of AGI-driven labor would reduce the marginal productivity of human workers, driving wages down. The shortage of essential but non-scalable resources, like land and energy, could further exacerbate the issue by creating bottlenecks in production. While AGI may initially complement human labor and spur innovation, Barnett predicts that diminishing returns and physical constraints on progress will eventually lead to human wages falling below subsistence levels.

Barnett estimates a 1 in 3 chance that wages could drop below subsistence by 2045 and a 2 in 3 chance by 2125, noting that AGIs may operate with lower energy costs and greater efficiency than humans. This could render humans unable to afford basic necessities through wages alone. While some argue that comparative advantage will sustain human employment, Barnett contends that this principle doesn’t guarantee livable wages. He concludes that while humans may still achieve well-being through alternative income sources like welfare or investment returns, the long-term effects of AGI on human welfare remain speculative. As Barnett states, “If AGIs can fully substitute for human workers, this wage floor will no longer be determined by the human-subsistence level.” Ultimately, the article underscores the urgency of addressing AGI's potential economic impacts to ensure human welfare in an increasingly automated future.