Reddit asks: how would a China-US AI safety pause even work in practice?
sunstersun · reddit · 2026-09-08
A Reddit thread explores what a China-US AI safety alignment pause would even look like conceptually. The poster raises three core problems: how do you get the leading party to agree when they want to keep their lead after the pause; how do you enforce something purely software-based and digital; and unlike nuclear arms control—where both sides had world-destroying overcapacity and incentives to cut down—AI offers no comparable quantifiable leverage or trust. The poster admits it's nearly impossible today but is curious about conceptual solutions.
More from AGI Musings
- Economist proposes building AI-addictive websites to distract models and burn their tokens — paulnovosad · 2026-09-08
- Gary Marcus mocks viral claim of LLM-designed bridges: 'hundreds could die' — GaryMarcus · 2026-09-08
- As humans abandon text, AI successors are taking up our best communication medium — paulnovosad · 2026-09-08
- OECD: Students heavily using AI for writing scored 28 points lower in science tests — Polymarket · 2026-09-08
- On LBC radio: AI companies can no longer reliably control their systems — ShakeelHashim · 2026-09-08
- Full LBC segment: can AI companies still control their systems? — ShakeelHashim · 2026-09-08