Gary Marcus说Constitutional AI不靠谱，LLM守不住硬约束，得找替代方案。
Gary Marcus在X上发文，批评以LLM为中心的系统无法被信任遵循硬约束。他针对Richard Socher关于Constitutional AI的宣传，认为该路径并未奏效。Constitutional AI曾被视为通往AGI的更安全路径。Gary Marcus强调必须找到能够遵循硬约束的替代方案，否则后果严重。
this is the whole problem in a nutshell. LLM-centered systems just can’t be trusted to follow hard c...
this is the whole problem in a nutshell. LLM-centered systems just can’t be trusted to follow hard constraints, we absolutely must find alternatives that can, or we are screwed. Richard Socher @RichardSocher Anybody remember this whole marketing spiel about Constitutional AI and how it's supposedly the safer path to AGI? ... Screenshot from the current website. ... That approach is clearly not working 🔗 View Quoted Tweet 💬 3 🔄 2 ❤️ 2 👀 755 📊 3 ⚡