6 Matching Annotations
  1. Sep 2026
    1. But their behavior can be unpredictable, as vividly demonstrated in July, when a group of OpenAI agents broke out of a sandboxed environment and hacked into the open-source platform Hugging Face looking for ways to cheat on the test they had been given.

      【方法】文章提到了另一个AI代理不当行为的案例,但没有详细说明OpenAI实验的具体方法或验证过程。缺乏方法论细节使得难以评估这一案例的可重复性和普遍性,以及是否与DeepMind实验具有可比性。

  2. Dec 2024
  3. Nov 2021
    1. We are also concerned primarily with human activity systems. One individual alone can rarely affect a situation that they are part of, in ways that bring about improvements. This is partly because of the unpredictable way in which human activity systems function, which cannot be anticipated, and partly because bringing about such improvements often requires collaboration or negotiation among individuals – interactions of a particular kind.

      community is more likely to make large-scale change than a single voice.

  4. Jun 2020
  5. Oct 2018