AI 公司仍未兑现最大承诺：Dario Amodei 的信任危机回应

AI companies cannot rebuild public trust through positive messaging when their largest promised social benefits remain undelivered.

The strongest criticism of the AI industry is its failure to produce measurable public benefits at the scale of its promises.

People will judge AI companies by pricing, access, lobbying, opacity, participation, and the distribution of economic gains as well as technical breakthroughs.

Well-designed institutions can constrain frontier laboratories while exempting smaller competitors and preserving room for open-weight models.

Open weights reduce some forms of concentration but do not eliminate dependence on scarce compute, chips, infrastructure, and capital.

Strong Chinese open-weight releases should be evaluated as genuine technical competition rather than dismissed as distillation or benchmark optimization.

Cheaper open models do not invalidate infrastructure investment because enterprises still need compute, storage, security, access control, and operations.

The cost gap between leading Chinese models and efficient American models is narrowing even though open releases still offer deployment flexibility.

Keeping powerful models for internal use may support safety evaluation, but it also makes external scrutiny and competitive comparison harder.

Concise primary-source responses can improve AI discourse by letting people debate statements that are easy to inspect rather than rumors or clipped interviews.