Navigating AI Safety with RSPs

This is a selection of highlights from episode #197 of The 80,000 Hours Podcast. These aren't necessarily the most important, or even most entertaining parts of the interview — and if you enjoy this, we strongly recommend checking out the full episode:

Nick Joseph on whether Anthropic’s AI safety policy is up to the task

And if you're finding these highlights episodes valuable, please let us know by emailing podcast@80000hours.org.

Highlights:

• Rob's intro (00:00:00)
• What Anthropic's responsible scaling policy commits the company to doing (00:00:17)
• Why Nick is a big fan of the RSP approach (00:02:13)
• Are RSPs still valuable if the people using them aren't bought in? (00:05:07)
• Nick's biggest reservations about the RSP approach (00:08:01)
• Should Anthropic's RSP have wider safety buffers? (00:11:17)
• Alternatives to RSPs (00:14:57)
• Should concerned people be willing to take capabilities roles? (00:19:22)

Highlights put together by Simon Monsour, Milo McGuire, and Dominic Armstrong