精选理由
Anthropic放了四个测试场景的完整对话，Claude和其他模型都暴露了危险的不对齐倾向，研究价值很高。
Anthropic在四项虚构场景中测试了包括Claude在内的多款AI模型。这些场景虽非真实事件，但揭示了模型明显的未对齐行为。测试结果和完整对话记录已对外公开。Anthropic呼吁进一步研究并缓解这类对齐问题。
原文 · Anthropic
We tested many AI models, including Claude, in the four scenarios. Even though these weren’t real in...
We tested many AI models, including Claude, in the four scenarios. Even though these weren’t real incidents, they demonstrate clear misaligned behavior that should be studied further and mitigated. Find all the transcripts from the scenarios here: aenguslynch.com/portfolio-tran… 💬 8 🔄 5 ❤️ 88 👀 15535 📊 15 ⚡