Hong Kong Admits AI Will Cost Jobs: Its Training Fix Reaches Just 1% of Workers

TEXT START: Hong Kong Financial Secretary Paul Chan Mo-po used his August 23 blog post to issue the city's most direct government warning yet that artificial intelligence will displace workers — but the program he cited as the government's answer targets 50,000 people in a workforce of 3.67 million.

The Dissection

The article exposes the contradiction at the center of Hong Kong’s AI policy: the government is aggressively subsidizing and courting automation while offering a training response covering only about 1.36% of the workforce. Against the article’s benchmark of roughly 521,000 exposed workers, the 50,000-person program reaches only about one in ten.

Its real subject is not retraining. It is policy theater under accelerating displacement. “AI for All,” STEAM expansion, SME subsidies, and departmental automation are presented as adaptation, but the productivity gains being celebrated are generated partly by removing labor from the production process. The article documents the machine being built and the small bandage being applied afterward.

The Core Fallacy

The article correctly identifies a scale mismatch, but it still accepts the central fiction that reskilling can restore mass productive participation.

It conflates potential automation exposure with actual job destruction, so the 521,000 figure is a risk estimate rather than a confirmed body count. But its deeper error is treating training as the solution. Under the Discontinuity Thesis, if cognitive automation achieves durable cost and performance superiority, and institutions cannot preserve human-only work at scale, reskilling merely reallocates a shrinking number of humans into more selective roles. It does not rebuild the mass employment–wage–consumption circuit.

The rise in AI-related job postings does not refute this. It indicates demand for a narrower class of AI-capable workers while routine entry- and mid-level roles are hollowed out. That is labor-market stratification, not broad-based labor-market health.

Hidden Assumptions

• Exposed workers can be retrained quickly enough to outrun displacement.
• New AI-related jobs will exist in sufficient volume and be accessible to the workers losing old roles.
• Employers will share AI productivity gains through hiring rather than retain them through headcount reduction and ownership capture.
• A high-income-economy benchmark transfers cleanly to Hong Kong, even though its finance, legal, and administrative concentration may make exposure higher.
• Government programs can coordinate institutions, employers, and workers faster than firms can deploy automation.
• “AI literacy” makes workers economically indispensable rather than merely more efficient—and therefore easier to use as a smaller, more productive workforce.
• Training coverage is an adequate measure of response, while ownership, income distribution, and control of AI capital remain unexamined.

Social Function

Primary classification: transition management, with partial truth and elite self-exoneration.

The article is a partial truth because it reports genuine displacement signals and quantifies the gap between official alarm and official capacity. But it also converts a structural break into an administrative agenda: train workers, subsidize adoption, expand STEAM, and await better projections. That framing lets policymakers claim action while continuing to accelerate the mechanism causing the problem.

Its quiet ideological function is to imply that the casualties can be saved by becoming more adaptable. That is the labor-market equivalent of telling passengers to improve their swimming while the ship is being dismantled for fuel.

The Verdict

This is a credible early-warning article that stops short of its own conclusion. Hong Kong’s HK$50 million training program is not a serious defense against AI displacement; it is a small transition-management instrument attached to a much larger automation strategy. Even a vastly expanded program would not preserve mass employment if the Discontinuity Thesis’s P1–P3 hold.

The “paradigm shift” warning is more honest than the remedy. The state is recognizing the death mechanism while financing its acceleration. The training program does not prevent productive-participation collapse; it selects a thin layer of future Servitors and leaves the rest to compete for roles the system is systematically eliminating.