Scott Alexander on Governments and labs should build the c…

Scott Alexander

Author and psychiatrist

A good pause strategy would involve both sides being able to monitor the other's data centers to prevent illegal training, then limiting training to some slow mutually-agreeable rate that lets alignment researchers thoroughly test each generation of AIs pre-release, monitor them post-release, and develop techniques to respond to any problems detected on deep levels that they expect to survive distribution shifts.

AI Verified source (2026)

4mo ago

Policy proposals and claims

Governments and labs should build the capability for a global slowdown of frontier AI development in case it's needed

votes For

Statement relation verification history AI Verified Report this

AI Verified The quote supports a coordinated AI 'pause strategy' by proposing monitoring of data centers and limiting training to a slow agreed rate. That clearly implies support for building the capability to enforce a broad slowdown of frontier AI if needed, even if 'governments and labs' is not named explicitly. · YouCongress gpt-5.4-2026-03-05 · 4mo ago

Vote inference verification history AI Verified Report this

AI Verified The quote endorses a coordinated slowdown mechanism by saying 'a good pause strategy would involve' monitoring data centers and 'limiting training to some slow mutually-agreeable rate.' · YouCongress gpt-5.4-2026-03-05 · 4mo ago

Slow down AI development

votes For

Statement relation verification history AI Verified Report this

AI Verified Full-claim match: the quote addresses AI development/capability growth and a slower pace, pause, halt, or explicit rejection/mixed position. Evidence: A good pause strategy would involve both sides being able to monitor the other's data centers to prevent illegal training, then limiting training to some slow mutually-agreeable rate that lets alignment researchers thoroughly test each gene · Hector Perez Arenas gpt-5 · 1mo ago

Vote inference verification history AI Verified Report this

AI Verified Recorded for matches the quote's stance on slowing AI development. Evidence: A good pause strategy would involve both sides being able to monitor the other's data centers to prevent illegal training, then limiting training to some slow mutually-agreeable rate that lets alignment researchers thoroughly test each gene · Hector Perez Arenas gpt-5 · 1mo ago

Quote authenticity verification history

Report this

AI Verified The quote is authentic: it appears verbatim in the Astral Codex Ten post “My AI Opinions,” whose byline identifies Scott Alexander as the author and whose publication date is June 11, 2026. The exact sentence appears in the source URL in the paragraph beginning “Overall thoughts on this section.” ([astralcodexten.com](https://www.astralcodexten.com/p/my-ai-opinions)) · YouCongress gpt-5.4-2026-03-05 · 4mo ago

replying to Scott Alexander