More generally, I think this is a good argument for shorter timelines being higher-leverage. As time goes on, the safety community has to compete for influence with more and more actors who are waking up.
One of the really surprising things about post-2022 for me has been that I thought at least OpenAI/GDM level of concern about AI (both in terms of negative effects on society, such as job loss, and x-risk) would be pretty universal. But I think the last few years has shown that 1) general concern about AI seems peculiarly Western (see: the Pew survey from last year) 2) x-risk concerns are also western and their influence on the AI industry seems to have been entirely contingent on the historical oddity of the AI industry being founded by people worried about safety. Any spread of x-risk concerns seems to come from the influence of rationalists/EAs and doesn’t seem to be emerging independently in China.
More generally, I think this is a good argument for shorter timelines being higher-leverage. As time goes on, the safety community has to compete for influence with more and more actors who are waking up.
One of the really surprising things about post-2022 for me has been that I thought at least OpenAI/GDM level of concern about AI (both in terms of negative effects on society, such as job loss, and x-risk) would be pretty universal. But I think the last few years has shown that 1) general concern about AI seems peculiarly Western (see: the Pew survey from last year) 2) x-risk concerns are also western and their influence on the AI industry seems to have been entirely contingent on the historical oddity of the AI industry being founded by people worried about safety. Any spread of x-risk concerns seems to come from the influence of rationalists/EAs and doesn’t seem to be emerging independently in China.