Gary Marcus直接点出AI不听指令比幻觉更危险，还引用了越具体越不听话的现象。想了解AI安全风险，看这条。
Gary Marcus在推文中指出，AI不遵循指令的问题正被许多人忽视，正如早期对幻觉问题的轻视。他认为，这类核心问题无法通过现有方法快速解决，除非开发出根本性新架构。Marcus警告，以LLM为中心的架构对AI安全构成深刻威胁。他还引用Huw Ringer的观点，称指令越具体，模型遵循的可能性反而越低。
Many people’s blindness to this problem — AI failing to follow instructions — is like their initiall...
Many people’s blindness to this problem — AI failing to follow instructions — is like their initially blasé attitude towards hallucinations, with the same fantasy that core, endemic problems would rapidly be solved. These problems won’t actually be solved, until we have fundamentally new architectures. Which is why LLM-centered architectures are a profound threat to AI safety. Huw Ringer @HuwRinger @GaryMarcus It’s actually slightly worse than this: the more specific set of instructions you give them, the less likely they are to follow them: dev.to/minatoplanb/i-… 🔗 View Quoted Tweet 💬 5 🔄 8 ❤️ 37 👀 3305 📊 8 ⚡