AI Safety Researcher Argues a Global ASI Treaty Is Possible; Skeptics Push Back
davidmanheim · x · 2026-09-15
In an exchange with Luca DellAnna, davidmanheim outlines his plan for a global treaty managing ASI-race risks: argue publicly for it, show it aligns with everyone's incentives, and crucially, don't give up preemptively. DellAnnaLuca counters that negotiating with "bad guys" to stop won't work, and that "generally aligned, generally refuse" doesn't hold today—no argument shows scaling beyond human oversight makes AI safer. davidmanheim replies that winning a race to omnicide doesn't appeal to him, and letting bad actors scale freely doesn't reduce risk either.
More from AGI Musings
- Lagarde warns Europe faces 'unprecedented risk' of being cut off from AI — talkingatoms · 2026-09-15
- A four-tier taxonomy of mind: what AI lacks is self-consciousness, not consciousness — FalseDinner335 · 2026-09-15
- HBS classmate's take: 'Americans do business, but the Chinese do survival' — nijfranck · 2026-09-15
- LeCun: LLMs alone are not a path to human-level AI — ylecun · 2026-09-15
- UChicago AI panel: Doomsday Clock at 85 seconds, experts agree race-now-guardrails-later is risky — WillRinehart · 2026-09-15
- AI's $1.1T data-center bet: productivity must grow 2.7x by 2030 to break even — nordicinst · 2026-09-15