2 Matching Annotations
  1. Jun 2026
    1. Restoring global trust in American AI is another thing entirely. No matter how long the shutdown lasts, it shined a light on how fragile access to US frontier AI models is.

      大多数人可能认为美国AI技术的优势地位是稳固的,但作者认为,这次事件暴露了美国AI访问权的脆弱性,可能永久性地损害了全球对美国AI技术的信任。这一观点挑战了美国AI技术主导地位的稳固性假设。

  2. Apr 2026
    1. As part of this investigation, we ran more ablations (removing lines from the system prompt to understand the impact of each line) using a broader set of evaluations. One of these evaluations showed a 3% drop for both Opus 4.6 and 4.7.

      大多数人认为微小的系统提示变更只会带来微不足道的影响,但作者展示了一个看似微不足道的提示变更(限制字数)却导致了3%的性能下降。这挑战了'小变更小影响'的直觉认知,揭示了AI系统中微小变化可能带来的非线性影响。