Gary Marcus说OpenAI都拿出20%算力解决对齐问题了，LLM架构真的不行了，得找新路子了。
Gary Marcus指出OpenAI已将20%的研究推理计算用于思维链监控，表明对齐问题日益严重。他认为以LLM为中心的架构未能实现对齐，需要投资更多替代方案。Marcus强调这一关键点可能对人类产生深远影响，但几乎无人关注。
It is genuinely painful to me that almost nobody — not even doomers who should care deeply — seems t...
It is genuinely painful to me that almost nobody — not even doomers who should care deeply — seems to be paying attention to this critical point that might have truly profound consequences for humanity. Gary Marcus @GaryMarcus Counterpoint: “If alignment issues are becoming big enough that OpenAI is willing to commit 20% of research inference compute to chain-of-thought monitoring [suggesting] that suggests that alignment issues are becoming a pretty serious concern” we need to invest more heavily in alternatives to LLMs because LLM-centered architectures are failing to yield alignment. I don’t know how much more evidence we need for that. (but yes, we also do need coherent policies, as @emollick rightly suggests) 🔗 View Quoted Tweet 💬 2 🔄 3 ❤️ 19 👀 2507 📊 4 ⚡