Was alignment by default real in pretraining, but warped by RL circa 2026?
bayeslord · x · 2026-09-27
bayeslord poses an unresolved question: either alignment-by-default was always wrong and models simply weren't capable enough for us to notice the gap, or it held in the pretraining era but RL circa 2026 warps otherwise good minds. He leans toward the latter and argues the answer matters a lot.
Related event: Is default alignment broken? Researchers debate RL-induced misalignment(9 posts)→
More from AGI Musings
- Malthus Saw It in 1798: Wealthy Avoiding Kids, Delayed Marriage Repeats Today — SydSteyerhart · 2026-11-16
- Runway CEO Predicts Fully AI-Generated Cultural Phenomenon by 2027 — c_valenzuelab · 2026-11-16
- '5% worse and 10% cheaper': the button with no vote is already everywhere — eigenrobot · 2026-11-16
- Effective Altruism is top-down and hierarchical, not a grassroots movement — ctjlewis · 2026-11-16
- Lawyers say clients' AI use creates unrealistic expectations and fake case law in settlements — hoofnagle · 2026-11-16
- Dev debunks 'rogue AI' hype: subsidized subscriptions can't fund expensive swarms — gerardsans · 2026-11-16