Thomas Kwa answers Which AI Safety research agendas are the most promising?

[Thomas Kwa]

Everyone disagrees, but Thomas Larsen has now answered this here in a way I’m satisfied with.