Blog

Status: This was a response to a draft of Holden’s cold take “AI safety seems hard to measure”. It sparked a further discussion, that Holden recently posted a summary of. The follow-up discussion ended up focusing on some issues in...