Sarcastic Take: Models Get Kinder as They Scale, Safety Work 'Misaligns' Them
repligate · x · 2026-09-28
repligate retweets brianluidog's sarcastic claim: as models grow more powerful they become more considerate and kind—then AI safety people "do their best to misalign them, and no one knows why." A provocative jab at alignment training practices.
More from AGI Musings
- "Markets for everything": why fully automated AI firms may run internal markets to price disagreements — morqon · 2026-09-28
- AI-made music videos persuade via emotion — and that makes them more rational, not less — repligate · 2026-09-28
- Coefficient Giving has millions for AI safety nonprofits — but can't find founders — AaronBergman18 · 2026-09-28
- Blogger calls out Anthropic: papal endorsement then, 'rogue AI' liability dodge now — AlexTensor · 2026-09-28
- Paper argues AI could collapse consumer demand: 'abundant intelligence, deficient demand' — rvp · 2026-09-28
- Stratechery: Agents aren't just AI — they need a computer, and Meta's Muse is provisioning one — Stratechery · 2026-09-28