Why don't frontier labs' models wow us? Three insider explanations
teortaxesTex · x · 2026-11-19
teortaxesTex 引用 Harry Waterman 的观点,讨论为什么公众看不到前沿模型在 AI 研发(RSI 相关)上的惊人能力。Waterman 给出三种可能解释:
- 模型确实主要针对 AI R&D 优化,对通用软件开发泛化很差;
- 前沿实验室员工大量使用这些模型,几乎「脑机接入」般熟悉它们,远比外部用户更擅长激发其能力;
- 我们拿到的并非真正意义上的内部模型。
teortaxesTex 补充称,实验室拒绝公开任何「惊为天人」的提示技巧,对外放出的大多是「加油你可以的」式话术,这反而说明上述可能性被低估了。
More from AGI Musings
- Andy Matuschak: Where's the Moral Desperation in Anthropic's Racing Stance? — KatjaGrace · 2026-11-19
- RSI debate: critics call recursive self-improvement a big-tech story, not a startup opportunity — teortaxesTex · 2026-11-19
- OpenAI researchers map three pathways for AI to enter organizations: promoting, interleaving, layering — devanshmehta · 2026-11-19
- 40+ Free AI Learning Resources: Videos, GitHub Repos and a Full Roadmap — nikola_mr64990 · 2026-11-19
- Panickssery rebuts 'AI pets' worry with 1900 self-employment analogy in Kelsey Tuoc thread — panickssery · 2026-11-19
- Dev predicts 2027 will be the year of 'outcome engineering' for agents — hwbhatti · 2026-11-19