GiveWell conversation with Tom Dietterich about long-term AI safety http://files.givewell.org/files...