Episode
"Current alignment training might be ineffective (and actively bad) in the age of RL" by Daniel Tan
- Published
- Sep 16, 2026
- Duration seconds
- 803
- Processing state
not_requested
Actions
POST https://stenobird.com/v1/public/podcasts/lesswrong-curated-popular-5643401/episodes/current-alignment-training-might-be-ineffective-and-actively-bad-in-the-age-of-rl-by-daniel-tan/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/lesswrong-curated-popular-5643401/current-alignment-training-might-be-ineffective-and-actively-bad-in-the-age-of-rl-by-daniel-tan.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
Tl;dr I am currently worried about current alignment techniques + how they are applied to frontier models. This decomposes into two hypotheses: Alignment techniques are not working to address misalignment from RL. Alignment techniques are actively obscuring evidence about misalignment. I think we do not currently have enough (public) evidence to conclude whether either of these claims are true. However, if both of these were true that would imply that alignment techniques are net bad and...