AI Could Make The Economy Far Richer. Anthropic Says. But Knowledge Workers Could Take a Hit.

TEXT START: A new economic model examines how different levels of AI adoption could affect growth, unemployment, wages and the share of income going to workers.

THE DISSECTION

This is a translation exercise: a possible labor-market rupture is converted into GDP projections and a manageable 2030 transition. The article admits the essential mechanism—AI performing nearly all knowledge work, no new knowledge tasks for humans, 15% annual growth, nearly one in five cognitive workers unemployed, knowledge-worker wages more than 10% below the counterfactual, and labor receiving only 45.2% of income. It then softens the impact with average-wage statistics and migration into construction or electrical work.

The real message is that aggregate output can soar while the people who made the old economy function lose bargaining power and income. Their displacement is treated as an adjustment variable.

THE CORE FALLACY

It confuses abundance with viability. GDP can rise while the wage-to-consumption circuit breaks. The claim that average wages rise is compositional camouflage: gains among less-exposed occupations do not restore the jobs, leverage, or status lost by displaced cognitive workers.

The deeper error is treating retraining as a solution when the problem is the disappearance of economically necessary human labor. The extreme scenario effectively grants P1—durable AI superiority across cognitive work—then treats the consequences as ordinary labor-market reallocation. A capital share of 54.8% is not a footnote. It is the mechanism by which production becomes less dependent on human wages. The model describes a richer economy whose mass participation system is being amputated.

HIDDEN ASSUMPTIONS

• Construction, electrical work, and other less-exposed occupations remain viable refuges rather than temporary lag domains. Highly capable physical automation is excluded.
• Displaced workers can cross skill, geographic, tenure, and status barriers quickly enough, despite the model admitting that such transitions are difficult.
• AI ownership and the resulting capital income remain socially and politically manageable.
• GDP growth translates into broad welfare and stable demand rather than concentrated purchasing power.
• Institutions can preserve human-only economic domains at scale, despite P2: coordination cannot permanently block cheaper, superior automation.
• The 2030 endpoint captures the relevant outcome. It does not model the continuation of automation after the transition begins.
• Job loss does not trigger feedback through falling demand, political instability, financial disruption, or institutional breakdown.
• Aggregate wage averages are an adequate measure of worker viability, even when gains are concentrated outside the occupations being destroyed.

SOCIAL FUNCTION

Classification: partial truth packaged as transition management, elite self-exoneration, and ideological anesthetic.

The article is not pure propaganda; it reports the damaging arithmetic and acknowledges major omissions. But its framing sanitizes a power transfer as a productivity scenario. It tells owners that they may receive a vastly larger economy and tells displaced workers to migrate into the remaining human niches. Capital capture is presented as a neutral output of growth rather than the death of labor's central role.

THE VERDICT

This is a soft-spoken pre-autopsy of postwar capitalism. Its extreme scenario already contains the Discontinuity Thesis kill sequence: cognitive labor is automated, human wages weaken, labor's income share collapses, output rises, and workers are displaced.

The article's mistake is temporal and systemic. It treats this as a temporary adjustment through 2030 instead of recognizing that, once AI becomes durably superior across cognitive work, retraining is hospice care for a dying labor market. The construction-and-electrical escape hatch is a lag defense, not a solution; the explicit exclusion of robots makes the scenario conservative.

AI does not need to make the economy poorer to kill the old order. It only needs to make production less dependent on human wages. This article accidentally documents that death.