Can the US government actually pause AI? Debate over liability, embedded evaluators and compute tracking
NathanpmYoung · x · 2026-11-15
NathanpmYoung argues that humanity does have the ability to stop AI — the US government could do it whenever it wants — but he'd prefer a defined threshold for when slowing down makes sense, plus concrete measures: liability regimes, embedded evaluators, and tracking the compute spent on research. An open-ended post inviting debate on AI governance.
More from AGI Musings
- Blaise Agüera y Arcas on how AI is blurring Descartes' mind-machine divide — blaiseaguera · 2026-11-15
- Edge City gave 297 villagers AI agents — GPT agents lied 94% vs Opus 2% — allisondman · 2026-11-15
- Anthropic study: one agent per person performs worse than a single shared agent — vista8 · 2026-11-15
- Deconstructing the "cure cancer" AI narrative: relief of suffering as Silicon Valley's only vision — danfaggella · 2026-11-15
- David Sacks warns recursive self-improvement is 'the most dangerous thing anybody can be working on' — peterwildeford · 2026-11-15
- AI Twitter Debate: Should Creating a New Digital Species Be a Crime? — zetalyrae · 2026-11-15