Ding — The Diffusion Marathon

Status

Stub via synthesis. This page captures Ding’s contribution as braided into geopolitics-global-ai-divide. The primary materials — Running the Right Artificial Intelligence Race: A National Strategy for AI Diffusion (RAND, Sept 2025) and Explaining China’s Diffusion Deficit (FPRI, Sept 2024), plus the ChinAI newsletter — have not yet been ingested directly.

Who

Jeffrey Ding. Assistant professor, George Washington University. ChinAI newsletter author. One of the most cited English-language analysts on US-China technology competition.

Central contribution

General Purpose Technologies (GPTs) are won by diffusion, not invention. Historical economic power transitions are rarely decided by the initial breakthrough; they are decided by which nation absorbs the technology fastest across its productive sectors.

See diffusion-vs-innovation for the concept page.

Two specific applications:

• The US is training for the wrong event. Framing AI as a sprint to AGI optimizes for frontier capability while neglecting domestic absorption. Ding advocates a “diffusion marathon” — widening the base of AI engineers, supporting public-private partnerships, helping SMEs adopt AI.
• China has a diffusion deficit. Despite massive innovation capacity (patents, publications), institutional bottlenecks and uneven STEM education historically isolated breakthroughs in elite institutions. The current Chinese pivot toward open-weight commoditization and B2B enterprise deployment (open-weight-asymmetry, lee-01ai-pivot) is a deliberate attempt to close this gap.

Why this source matters for the paper

• Reframes the US strategic posture as potentially misframed — a much stronger critique than the typical “national security export controls debate”.
• Provides the structural explanation for why China is bottom-heavy in divergent-value-stack-optima — not just because of embargoes, but because closing the diffusion deficit is the underlying state imperative.
• Connects to task-based-framework: if diffusion does not happen, the entire AI productivity surplus may be small enough that no value-stack shape redistributes meaningful gains.

To verify on primary read

• Ding’s exact framework for measuring “diffusion capacity” — is it operationalizable for the paper’s purposes?
• His current view (post late-2024) on US prospects for closing its own diffusion gap.
• Whether he treats open weights as a diffusion accelerant or as a separate strategic axis.

Related

• diffusion-vs-innovation — the concept page.
• open-weight-asymmetry — the mechanism by which China operationalizes diffusion.
• task-based-framework — the macro ceiling that diffusion is meant to unlock.
• geopolitics-global-ai-divide — the synthesis that braids him into the broader story.