“Taking AI Risk Seriously” – Thoughts by Andrew Critch

Link post

I wrote this sev­eral months ago for LessWrong, but it seemed use­ful to have cross­posted here.

It’s a writeup of sev­eral in­for­mal con­ver­sa­tions I had with An­drew Critch (of the Berkeley Ex­is­ten­tial Risk Ini­ti­a­tive) about what con­sid­er­a­tions are im­por­tant for tak­ing AI Risk se­ri­ously, based on his un­der­stand­ing of the AI land­scape. (The land­scape has changed slightly in the past year, but I think most con­cerns are still rele­vant)