Greg_Colbourn comments on Mediocre AI safety as existential risk