Ryan Greenblatt on Pausing AI at human-genius level to sol…

Ryan Greenblatt

Chief Scientist at Redwood Research; AI safety researcher; lead author of "Alignment faking in large language models"

I do think that pausing further capabilities once we have human-ish-level AIs for even just a few years while we focus on safety would massively improve the situation. This currently seems unlikely to happen.

AI Verified (2025)

Source:

Why would AI companies use human-level AI to do alignment research?

2mo ago

Policy proposals and claims

Pausing AI at human-genius level to solve alignment is safer than either racing to superintelligence or halting entirely

votes For

Statement relation verification history AI Verified Report this

AI Verified The quote explicitly says pausing further capabilities for a few years while focusing on safety would massively improve the situation, so it is on-topic. · Hector Perez Arenas gpt-5 · 2mo ago

Vote inference verification history AI Verified Report this

AI Verified The quote explicitly says pausing further capabilities for a few years while focusing on safety would massively improve the situation, which supports the statement. · Hector Perez Arenas gpt-5 · 2mo ago

Slow down AI development

votes For

Statement relation verification history AI Verified Report this

AI Verified Full-claim match: The quote says pausing further capabilities at human-level AI for a few years would massively improve the situation. · Hector Perez Arenas gpt-5 · 2d ago

Vote inference verification history AI Verified Report this

AI Verified Recorded for matches the quote’s explicit support for slowing or pausing AI capability development. The quote says pausing further capabilities at human-level AI for a few years would massively improve the situation. · Hector Perez Arenas gpt-5 · 2d ago

Quote authenticity verification history

Report this

AI Verified EA Forum post line 74 matches the quote about pausing further capabilities at human-ish-level AIs to focus on safety. · Hector Perez Arenas gpt-5 · 2mo ago

replying to Ryan Greenblatt

Other opinions from Ryan Greenblatt

See all

Ryan Greenblatt votes For

We should adopt AI 2040 Plan A

Plan A seems like a good plan for handling powerful AI, or at least the best plan anyone’s written up.

AI Verified source (Jul 11, 2026)

added 23d ago

Ryan Greenblatt votes For

The U.S. and China should establish Mutually Assured Compute Destruction

Plan A seems like a good plan for handling powerful AI, or at least the best plan anyone’s written up. Many choices initially seem crazy, but are actually pretty carefully considered. Plan A isn’t likely to happen, but pushing for something like this...

more AI Verified source (Jul 11, 2026)

added 2mo ago

Ryan Greenblatt votes Against

AI alignment is solvable

We don't know how AIs are aligned. A somewhat crazy aspect of the current situation is that we have very little confirmed public information about why frontier AIs end up being apparently behaviorally aligned. And more generally, we don't know what f...

more Unverifiable source (May 8, 2026)

added 4mo ago