Top RL researcher: stay paranoid about AI until GPT-9's alignment is formally verified
AaronBergman18 · x · 2026-11-11
Prominent ML researcher tszzl (Leo Gao) argues we should genuinely internalize that the 'dinosaurs' are months away from eating us — i.e., AI risk may be imminent — and only stop being paranoid after formally verifying the alignment of GPT-9. A tongue-in-cheek but pointed take on alignment urgency under short timelines.
More from AGI Musings
- Musk: Superintelligence and Robotics Era Coming, But Birth Rate Collapse Is the Greatest Threat — XFreeze · 2026-11-11
- NBER model: jagged super-intelligence may not transform scientific progress and growth — JMateosGarcia · 2026-11-11
- Harvard physics professor leaves OpenAI to join Neolab, betting on AI-driven physics — ns123abc · 2026-11-11
- David Manheim: verifiable tasks scale fast, but 'squishy' real-world tasks hit sparse-reward limits — davidmanheim · 2026-11-11
- Is Anthropic's alleged 2x AI R&D speedup real? A thread stress-tests the 'benchmark-only' counterargument — dfrsrchtwts · 2026-11-11
- Forecast track records favor Kokotajlo's AGI timelines over critics, argues Manheim — davidmanheim · 2026-11-11