A job apocalypse? | World Finance

TEXT START: In 2023, Goldman Sachs estimated 300 million jobs across the world could be exposed to artificial intelligence.

THE DISSECTION

This article is a soft-launch autopsy disguised as workforce analysis. It assembles evidence that firms are shrinking teams, agents can execute end-to-end processes, entry-level roles are disappearing, and low-wage workers face the greatest transition risk. Then it retreats into augmentation, the “feeling economy,” retraining, safety nets, human-centred workplaces, and uncertainty.

Its real function is to make structural rupture acceptable to institutional audiences without naming the ownership conflict. It treats AI as a workforce transformation rather than an ownership transformation: who controls intelligence, compute, data, energy, distribution, and the resulting output. The text records early P1 and P3 signals while avoiding the P2 implication: institutions have no reliable capacity to preserve mass human-only work once firms can coordinate production through agents.

THE CORE FALLACY

It confuses human usefulness with mass market necessity. Humans may remain useful as judges, relationship managers, trainers, explainers, and accountability proxies. That does not mean millions of humans must be paid to perform those functions. A small Servitor team can direct hundreds of agents, and the same systems can increasingly perform explanation, coordination, evaluation, persuasion, and simulated empathy. “Human in the loop” can mean a liability signature, an approval click, or a thin supervisory layer—not restored mass employment.

The article’s proposed new roles are not tested against scale. Agent orchestrators, AI coaches, and hybrid managers may exist, but two to five humans overseeing 50 to 100 agents is a demolition ratio, not a jobs program. The entry-level pipeline is especially vulnerable: when AI absorbs junior work, it also destroys the training ladder that supposedly produces experienced workers.

The article also assumes productivity gains will be distributed. UBI or universal high income may preserve consumption if politically imposed, but it does not restore productive participation or bargaining power. Under the Discontinuity Thesis, that is system maintenance—not survival of post-WWII capitalism.

HIDDEN ASSUMPTIONS

• Firms will retain humans because accountability, judgment, empathy, or trust are intrinsically human. In practice, they can retain a small number of legally useful or commercially indispensable Servitors and discard the rest.
• Human skills will remain scarce after AI systems become better at modeling language, emotion, persuasion, explanation, and social behavior. Scarcity is not a moral property; it is whatever the market cannot cheaply reproduce.
• Automation will create enough new roles to offset eliminated roles. The article provides no evidence for the necessary volume, wage level, or accessibility.
• Retraining can outrun deployment. It cannot if target occupations are automated faster than workers can enter them.
• Education democratizes the gains. Access to tools is not ownership of the productive system; it may simply create better-equipped applicants competing for fewer seats.
• Human oversight will remain labor-intensive. As agents become more reliable, oversight compresses into sampling, exception handling, and audit.
• Consumer distrust, hallucinations, bias, and regulation will materially block adoption. These are friction and lag, not reversal mechanisms.
• Governments will tax AI owners and distribute gains at the required scale. The article ignores the political power of the Sovereigns controlling the taxable base and infrastructure.
• Psychological distress is an economic veto. It is not. The market does not compensate people for suffering unless forced to do so.
• Unpredictability means the direction is unknown. The pace and institutional response are uncertain; the competitive direction—replace costly labor wherever possible—is not.

SOCIAL FUNCTION

Primary classification: transition management and ideological anesthetic, with a substantial partial-truth component.

The partial truth is valuable: the article accurately records the early mechanics of P1, especially in entry-level cognitive work, and admits that firms are pairing AI investment with headcount reduction. Its anesthetic function begins when it converts that evidence into soft skills, retraining, future job titles, and human-centred workplaces. Those proposals let executives and policymakers appear responsible without confronting ownership or the destruction of the mass employment-to-consumption circuit.

It is also elite self-exoneration. The beneficiaries can claim they warned everyone, encouraged reskilling, and supported a safety net. If displaced workers become economically unnecessary, the alibi is ready: they failed to adapt, technology was unpredictable, or education was neglected. The firms making the substitution decisions disappear behind the passive noun “AI.”

DT ASSESSMENT

The article’s evidence already sketches the hardened framework:

• P1: AI is gaining durable cost and performance advantages in cognitive work, with agents extending automation from task assistance to process execution.
• P2: The article offers no credible mechanism for preserving human-only domains at scale. Trust, regulation, and ethics may create delay markets, not a permanent labor shelter.
• P3: Entry-level erosion, thinner teams, and concentrated ownership remove the majority’s access to economically necessary labor.

The article’s own solution—more human direction around more agents—describes a Servitor minority attached to Sovereign-owned systems. It does not describe a viable future for the displaced majority.

THE VERDICT

This is not a neutral investigation of a possible job apocalypse. It is an early warning label wrapped around a refusal to read the warning. The article correctly documents the beginning of productive participation collapse, then hides the terminal implication behind human skills, retraining, safety nets, and uncertainty.

The central question is not whether AI creates some jobs. It will. The question is whether it creates enough economically necessary human work to preserve mass wages, bargaining power, and consumption. Nothing in the article establishes that. Its own numbers imply the opposite: fewer humans commanding vastly more machine labor.

The post-WWII circuit is not being upgraded. It is being severed. The likely endpoint is a concentrated Sovereign class, a narrow Servitor layer, transition-management institutions, and a majority maintained—if politics permits—through transfers after productive participation has ceased.