AI keeps stubbornly refusing to take our jobs - Noahpinion

TEXT START: It’s Labor Day, so here’s a post about how human labor is alive and well in the age of AI.

The Dissection

The article converts a snapshot of labor-market resilience into a rebuttal of AI-driven obsolescence. It catalogs complementary tasks, infrastructure jobs, AI-management roles, and the persistence of messy occupations, then treats their existence as evidence that human employment remains structurally secure.

Its strongest point is narrow: task automation does not immediately equal occupational extinction. AI currently creates new implementation, oversight, infrastructure, and demand-side work. That is a transition effect, not proof that the employment system survives the transition.

The Core Fallacy

The article confuses delayed displacement with defeated displacement.

The Discontinuity Thesis does not require every AI advance to reduce employment immediately. It requires AI to achieve durable superiority across cognitive work, after which institutions fail to preserve stable human-only economic domains. New tasks can temporarily reinstate labor, but the article never demonstrates that those tasks will remain economically necessary once AI can perform the higher-order functions of directing, checking, integrating, and improving AI systems itself.

“Humans will always have jobs keeping AI agents on track” is not an argument. It is an unsupported assumption about a permanent human comparative advantage. If supervision becomes automatable, the supervisory layer is merely the next carcass.

The article also treats increased output, product variety, and AI infrastructure spending as durable employment engines. They may increase labor demand during the buildout. They do not guarantee that labor captures the gains, retains bargaining power, or remains necessary after the capital stock is installed.

Hidden Assumptions

• Human oversight will remain cheaper and more reliable than machine oversight.
• New AI-created tasks will expand faster than AI can absorb them.
• Productivity-driven demand growth will translate into mass employment rather than greater returns to AI capital.
• Physical bottlenecks will remain durable enough to anchor broad human livelihoods.
• Corporate complexity makes workers indispensable rather than merely temporarily useful.
• Current employment levels reveal the terminal trajectory rather than institutional and infrastructural lag.
• AI companies’ messaging reflects economic reality rather than adaptation to political resistance and deployment constraints.
• “Jobs created” is treated as equivalent to productive participation, even when those jobs are temporary construction, capital-complementary roles, or narrow transition niches.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The article performs a useful empirical correction against simplistic claims that every exposed occupation will vanish overnight. But socially, it reassures the threatened class that visible employment is evidence of permanent necessity. It turns the early construction phase of an automation regime into a story of labor’s resilience.

Its optimism also exonerates the system’s winners. If AI creates enough new tasks, then no structural redistribution or ownership shift is required; displaced workers merely need to wait for the next niche. That is the comforting fiction. The decisive question is not whether AI creates work, but whether humans retain ownership or indispensable control over the productive machinery.

The Verdict

The article is accurate about the lag and wrong about the destination. AI has not yet killed mass employment because the system is still in the complementarity, deployment, and infrastructure phase. New tasks are hospice care for the old labor circuit, not evidence that the circuit cannot be severed.

It documents the first delay in the death process, then mistakes the delay for survival.