Overview
This series examines how China’s open AI ecosystem responds to technological containment: open weights can reduce reliance on a vendor-controlled interface, enable local adaptation and distributed improvement, and diffuse across borders. It then traces the compute, runtime, model-lineage, data, licensing, and institutional dependencies that remain, including modeled license conflicts and bounded automation. The evidence supports resilience as layered and conditional—not complete technological independence.
Learning outcomes
- Open weights reduce exclusive vendor control over model access, but they do not remove the wider dependency stack.
- Under infrastructure constraint, local deployment and adaptation gain option value, not guaranteed affordability or independence.
- Downstream descendants and cross-border use can distribute capability beyond the original publisher.
- Resilience depends on controllable compute, traceable lineage and data, compatible terms, and institutional acceptance.
- Automated license analysis can surface candidate conflicts, but human review remains necessary for legal decisions.
Keywords
Preview every chapter · Unlock full text, audio, and video from chapter 4
PlansFAQ
What dependency do open weights remove?
They remove exclusive control over model parameters or a hosted interface, while compute, runtime, lineage, data, licensing, and governance may remain external.
Why did technological containment increase the value of local adaptability in China’s open AI ecosystem?
It made access to chips, compute, and foreign-controlled services less certain, so locally deployable and modifiable models offered more workable alternatives; it did not prove containment alone caused the ecosystem’s growth.
How do community descendants change open-model capability?
They let downstream actors tune, compress, or otherwise adapt a release for particular tasks and constraints; selected benchmarks show the best descendants often match or exceed official releases, not that every derivative is broadly better.
Is an open-weight model independent once it can be downloaded?
No. Independence depends on whether the organization can run, adapt, trace, license, govern, and institutionally support the model for its intended use.