Larks comments on Third-wave AI safety needs sociopolitical thinking