The AI labor apocalypse is on hold (for now) | Economy and Business | EL PAÍS English

TEXT START: There are three versions of Saint John’s Book of the Apocalypse, a text that never loses relevance and now seems to presage the impacts of artificial intelligence (AI).

The Dissection

This is a de-escalation memo disguised as analysis. It correctly attacks theatrical doomsday predictions and notes that mass layoffs have not yet appeared in aggregate data. Then it quietly substitutes a weaker question: “Has unemployment already exploded?” for the structurally decisive question: “Is human labor becoming less necessary and less powerful?”

The article’s own evidence exposes the transition it tries to soften: young people face a narrowing entry door, labor’s income share is deteriorating, work may be reduced to fewer hours, inequality is widening, and production is concentrating in a handful of firms. Those are not refutations of the labor apocalypse. They are its early, low-temperature symptoms.

Under the Discontinuity Thesis, the article is measuring the lag phase. Employment stocks remain high while task composition, bargaining power, and access to entry-level work deteriorate. AI-created jobs during an investment and deployment boom do not prove that equivalent human roles survive after systems become cheaper and more capable. The article observes the scaffolding and mistakes it for the completed building.

The Core Fallacy

The central error is treating historical precedent and current employment totals as laws of nature. Earlier technologies displaced tasks while generating new domains that still required large volumes of human labor. AI is aimed directly at cognitive coordination, analysis, writing, coding, administration, and oversight—the machinery that generated much of the modern white-collar economy.

“Employment has always recovered” is not an argument against P1, P2, and P3. It is an extrapolation from a different technological regime. If AI achieves durable cost and performance superiority across cognitive work, competitive firms cannot preserve human-only domains merely because society prefers them. Regulation can slow deployment, redistribute income, or subsidize employment. It cannot permanently restore productive necessity.

The article also confuses consumption with participation. Safety nets and universal basic income may keep people buying goods. They do not preserve ownership, bargaining power, status, or an economically necessary role. That is system maintenance, not survival of the postwar labor circuit.

Hidden Assumptions

• Aggregate employment equals healthy productive participation; hours, wages, job quality, underemployment, and bargaining power are treated as secondary.
• Job creation during the current investment wave will continue after AI deployment matures.
• Historical demand expansion will repeat even when the technology targets cognitive labor itself.
• Retraining can produce enough complementary work, quickly enough, to outrun AI capability gains.
• Governments can coordinate durable restrictions despite firms competing to automate and reduce costs.
• Human oversight roles are permanent rather than temporary transition scaffolding.
• The absence of decisive statistics means the mechanism is absent, rather than delayed or poorly measured.
• Redistribution can solve a problem fundamentally caused by concentrated ownership and control of productive systems.

The article recognizes centralization but treats it mainly as a regulatory defect. Under DT logic, it is the ownership structure of the replacement economy.

Social Function

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

Its partial truth is real: the immediate mass-layoff scenario has not materialized in the supplied evidence, and sensational forecasts are often deployed by firms seeking capital, influence, or regulatory advantage. But the article uses that valid criticism to launder a more comforting conclusion: that adaptation, upskilling, shorter workweeks, and safety nets can preserve the existing order.

That is the institutional lullaby. It converts a question of ownership and human necessity into a question of policy calibration. It reassures workers that the machine has not arrived while documenting that the entrance to the labor market is already being bricked shut.

The Verdict

The AI labor apocalypse is not “on hold.” The spectacular layoff phase is on hold. The displacement mechanism is advancing through task erosion, blocked entry, declining labor share, concentration, and weakened bargaining power.

The article does not disprove the Discontinuity Thesis. It supplies evidence for its lag defenses: physical deployment, institutional inertia, political hesitation, and statistical delay. P1 remains unresolved rather than defeated; P2 is acknowledged but underweighted; P3 is already visible at the labor-market entry point even if it has not yet appeared as aggregate unemployment.

This is a competent snapshot of the pre-collapse interval. Its headline mistakes delayed rupture for cancelled rupture.