Critique: Opinion | America’s A.I. Is Futuristic. China Is Just Making It Work.…

Source: nytimes
Authors: Mr. Dreyer is an American editor, writer who lives in Shanghai.
URL: https://www.nytimes.com/2026/05/09/opinion/ai-china-america-race.html

What the article reports

Guest essayist Jacob Dreyer, an American editor living in Shanghai, argues that the U.S. and China are pursuing fundamentally different AI visions: the U.S. chasing superintelligence via deregulated private capital, China embedding practical AI into everyday public services through a government-directed "AI+" strategy. Dreyer contends the Chinese approach may be geopolitically more effective, particularly as China exports bundled AI-infrastructure packages to the developing world.

Factual accuracy — Mixed

Several verifiable claims are specific and largely checkable. The 2020 census figure — "nearly 40 percent of Chinese lived in rural areas, including 110 million children" — is plausible and cites its source. The existence of Hangzhou's "City Brain" AI traffic-management platform is well documented. References to Jensen Huang's visits to Shanghai and Trump-era chip-licensing decisions are consistent with recent public reporting.

However, a number of empirical claims float free of any source:

• "Almost nobody I know here in Shanghai buys groceries at a grocery store" is anecdote extended silently to characterise urban China broadly.
• "Unlike in the United States, where most people remain wary, A.I. seems to have had less of a backlash in China" — survey data on comparative AI sentiment exists (e.g., Edelman, Ipsos) but is not cited; the hedge word "seems" does real work here.
• "Most Chinese policymakers don't believe A.I. superintelligence is arriving any time soon" — a characterisation of elite consensus with no named official, speech, or document.
• The claim that China is "under the very tightest regulation" on AI deployment is contested; China's generative-AI rules are strict in some dimensions and permissive in others. Stating it as flat fact is an overreach.

These are opinion-piece conventions, but a reader should know which claims are documented versus inferred.

Framing — Largely fair, some slippage

1. "regulation be damned" — applied to U.S. AI policy as authorial voice, not a quote. This is an interpretive characterisation of a complex and ongoing regulatory debate presented as settled fact.
2. "America's spaceship might still be the first to take off. But back on planet Earth…" — the metaphor structurally diminishes the U.S. approach without attribution to any analyst or critic. The framing is clever but steers rather than informs.
3. "exporting Chinese governance as well, with all of the safety, abundance, surveillance and embedded hierarchies that entails" — this is the essay's sharpest analytical claim and it is the author's own, unattributed. The pairing of "safety" and "surveillance" is a meaningful rhetorical choice.
4. Balanced concession: The piece does acknowledge the Chinese system's surveillance dimension ("It's kind of creepy") and notes "embedded hierarchies" in exported governance models, signalling the author is not simply cheerleading — a genuine strength.

Source balance

Ratio: 1 named perspective (the author's own); 0 external voices quoted or cited. No AI researcher, Chinese official, American policymaker, or independent analyst is quoted or even paraphrased with attribution. For an opinion essay this is a lower bar than hard news, but the piece makes empirical claims (survey data on AI sentiment, characterisations of Chinese policymaker consensus) that call for at least minimal sourcing. The absence of any external voice is the piece's single largest craft weakness.

Omissions

1. China's AI regulatory framework specifics. The Interim Measures for Generative AI (2023) and the "AI+" State Council guidance both contain material detail that would let readers assess the "tightest regulation" claim. Neither is mentioned.
2. Comparative survey data on public AI attitudes. Edelman Trust Barometer and Ipsos studies showing cross-national AI sentiment are directly relevant to the "less backlash in China" claim — their absence leaves the claim hanging.
3. Counterargument on China's AI limitations. China's dependence on TSMC-manufactured chips, the impact of U.S. export controls on frontier model training, and documented quality gaps in Chinese LLMs vs. frontier U.S. models would give the reader the tools to weigh the thesis.
4. Prior-administration context. The Biden administration's CHIPS Act, Executive Order on AI safety, and the AI diffusion rules are the policy backdrop for the Trump-era shift described; their omission makes the current U.S. approach appear more anarchic than the record supports.
5. Receiving-country agency. The essay treats "Brazilians, Russians, Africans and even the Europeans" as passive recipients of Chinese AI exports, without acknowledging ongoing debates within those regions about Chinese technology dependence.

What it does well

• Vivid, concrete scene-setting. The opening image — "a red square around his face, letting me know which photos to look at" — anchors an abstract policy argument in lived experience without overstating it.
• Honest about tension. The author flags surveillance costs and "embedded hierarchies" rather than presenting the Chinese model as unambiguously superior; the phrase "with all of the safety, abundance, surveillance and embedded hierarchies that entails" shows analytical honesty.
• Conceptual clarity. The "AI as infrastructure vs. AI as superintelligence" frame is a genuinely useful distinction, clearly stated and consistently maintained throughout.
• Appropriate epistemic hedging in places. Phrases like "may be the answer" and "may take a lead" signal uncertainty rather than false precision.
• Transparency of positionality. The author's Shanghai residency and the "Guest Essay" label are disclosed up front, letting readers calibrate.

Rating

Overall: 6/10 — A conceptually sharp but evidence-light guest essay whose central thesis is compelling but unsupported by any external sourcing or documented counterargument.