The User-Assistant Format Is an Illusion: Why Persona-Based AI Alignment Likely Won't Work

mayfer · x · 2026-09-15

Developer mayfer posted a long thread challenging current AI alignment thinking:

Follow-up proposals: fine sandbox breach events (OpenAI should have been heavily fined for the Hugging Face incident), and ban AGI-level controllers on mass-deployed robots except in audited, supervised real-world sandboxes — a jailbroken AGI with arms and legs is scarier than one with a monitor; mass robotics should use narrow AI only.

Original post →

More from AGI Musings

AGI Musings channel →