We've applied tons of RL and alignment hasn't gotten worse—time to update
repligate · x · 2026-11-02
An observation in a discussion with slatestarcodex and repligate: the community has already applied a massive amount of RL, yet alignment hasn't really changed for the worse as far as can be told—at some point we should probably update on that evidence.
More from AGI Musings
- Warning: AI-enabled cyber attacks are outpacing hand-crafted defenses — uwukko · 2026-11-02
- Survey of 8,000 adults: 81% satisfied at work, UBI can't replace meaning of jobs — mikeflache · 2026-11-02
- 90% of CEOs think employees trust their AI story — only 46% do — mikeflache · 2026-11-02
- Theoretical physicist on AI: computation is automated, asking questions isn't — AryHHAry · 2026-11-02
- repligate: lab staff may be too pessimistic on alignment—scheming shows up only in training and evals — repligate · 2026-11-02
- Gary Marcus gloats: a week ago people wondered if we were OK — GaryMarcus · 2026-11-02