“Third-wave AI safety needs sociopolitical thinking” by Richard_Ngo | LessWrong (30+ Karma) | Podwise