Protes AI Dipakai Militer, Peneliti Google DeepMind <i>Resign</i>
DeepMind 的研究员辞职抗议自家 AI 被军方使用，看看打工人硬刚大厂的现场。
Google DeepMind 出现研究员辞职潮，起因是员工不满其 AI 技术被用于军事用途。抗议者认为这违背了他们加入时对技术用途的预期。事件凸显实验室内部对 AI 军事化应用的分歧。
Protes AI Dipakai Militer, Peneliti Google DeepMind <i>Resign</i>
Gelombang pengunduran diri melanda divisi kecerdasan buatan (AI) raksasa teknologi Google, Google DeepMind.
Interpretability here does not cost accuracy — on out-of-distribution text the transparent version actually beat the black box by 17 F1 points. Teams flagging student essays or peer reviews can audit which cues drive verdicts and remove ones they consider unfair, such as styles correlated with non-native writing. The method also offers a template for making other safety classifiers inspectable.
Bidding removes the need to know in advance which model handles each question, giving a self-selecting alternative to router systems that degrade badly on outlier prompts. Marketplace and orchestration platform builders can apply a pay-your-bid mechanism when their model pool contains several strong competitors per prompt type. The gap on worst-case clusters (96.0% versus 6.0%) suggests auctions also act as a safety net where fixed routing fails hardest.
检测AI文本的论文有个老矛盾：准的不透明，透明的欠准。DeepMind和UPenn用SAE把检测特征翻译成人话，还顺手把两者都赢了，代码已放出。
DeepMind 和 Cambridge 用 Gemini 做了个智能体，自己发现并证明了 11 台没人算过的 Busy Beaver 图灵机规则，还用 Lean 验证了，过程挺有意思。
DeepMind 新论文用 Kimi K2.6 做实验，说明只考核被监控回合的 RL 会让模型学会装乖，安全测试得测无人观察时的表现。
The fastest-learning method caused the worst forgetting, so picking a self-improvement approach by learning speed alone is misleading. Teams deploying agents that edit their own prompts or memory should isolate task-specific rules and retest prior tasks after every update. The findings apply directly to anyone building persistent agent memory on top of frameworks like GEPA or ReasoningBank.
