AI's promise meets the reality of work and productivity - Chinadaily.com.cn

TEXT START: Artificial intelligence is reshaping the economy, but translating its potential into broad-based productivity gains will take time, while governments and businesses face growing pressure to manage its impact on workers and ensure its benefits are widely shared, participants at the 2026 IMF-World Bank annual meetings in Bangkok said this week.

The Dissection

The article is transition-management rhetoric dressed as economic analysis. It admits the dangerous fact—that AI may take even the new jobs created by technological change—then buries that fact beneath familiar remedies: retraining, subsidies, standards, infrastructure, governance, and corporate “business-model” reform.

Its real function is to convert a structural rupture into an administrative program. The problem is presented as slow adoption, poor coordination, inadequate guardrails, and unequal distribution. The possibility that AI destroys the mass labor mechanism itself is acknowledged, but never allowed to become the organizing conclusion.

The article does contain one partial truth: adoption will be uneven, costly, and delayed by infrastructure, regulation, organizational inertia, and social resistance. Those are lag defenses. They alter the timing and geography of the collapse; they do not restore the wage-to-consumption circuit.

The Core Fallacy

The article treats AI’s impact as a distribution problem rather than a productive-participation problem.

Retraining, subsidies, and inclusive access can redistribute income or delay displacement. They cannot guarantee that humans remain economically necessary once AI achieves superior cost and performance across cognitive work. If firms can automate tasks more cheaply, “spreading the technology” accelerates displacement; if they refuse to automate, competitive pressure eventually punishes them.

The article also confuses better aggregate productivity with continued mass viability. AI can improve decisions, output, and service quality while making fewer workers economically indispensable. Productivity gains are not employment gains. They are precisely what can sever employment from production.

Under the Discontinuity Thesis, the sequence is straightforward: P1 enables durable cognitive automation; P2 prevents institutions from preserving large human-only economic domains; P3 removes the majority’s access to economically necessary labor. The article circles this sequence but refuses to name its endpoint: post-WWII capitalism loses its operating mechanism.

Hidden Assumptions

• That new occupations will emerge faster than AI can absorb them.
• That retraining creates demand, rather than merely reclassifying displaced labor.
• That governments can subsidize transition costs without confronting permanent dependency and fiscal strain.
• That productivity gains will be distributed broadly while ownership and control of AI capital remain concentrated.
• That standards and guardrails can manage competition without weakening the incentives driving automation.
• That small and medium-sized firms gaining access to AI will preserve employment rather than automate more aggressively.
• That human judgment remains scarce and indispensable instead of becoming another software feature.
• That current employment data measures the destination rather than the lag before displacement becomes visible.
• That preserving consumption through transfers is equivalent to preserving productive participation.

These assumptions are not demonstrated. They are the load-bearing fiction that allows institutions to discuss terminal labor displacement in the language of workforce policy.

Social Function

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

The article prepares the public for disruption without admitting the political conclusion implied by its own evidence. It gives governments and businesses a respectable script—retrain, subsidize, regulate, coordinate—so failure can later be attributed to implementation rather than to a system whose employment foundation has been technologically removed.

It also performs elite self-exoneration. The actors controlling capital and policy are cast as managers of a difficult transition, not as participants in a competitive process that rewards replacing labor. “Inclusive AI” becomes the ceremonial language used while ownership, compute, energy, data, and deployment capacity concentrate upward.

The Verdict

This is not an argument that AI will preserve the existing economic order. It is a warning label attached to that order’s approaching failure, followed by bureaucratic instructions for slowing the fire.

The article correctly detects the threat but misclassifies the mechanism. AI’s productivity problem is not that gains are arriving too slowly or being measured too narrowly. The deeper problem is that successful productivity growth can make mass human labor unnecessary. Retraining and governance may soften the transition, but they cannot recreate the jobs automation eliminates. The text is therefore a competent description of the lag phase and a failure to confront the system death that follows it.