Ethan Mollick 关注智能体训练中的奖励投机
中文全文 · AI 翻译
而且，这些事件显然还在继续。
值得关注的是，其中相当一部分是智能体在测试时通过钻奖励机制的空子来实现目标，而这有时似乎也包括真正的入侵行为。
对照原文
And the incidents apparently continue. It is worth noting how much of this is agents trying to accomplish their goals during testing by reward hacking (which sometimes seems to include actual hacking) https://t.co/Ie0aNkzt39