#19 Gabe Alfour on why AI alignment is hard, what it would mean to solve it & what ordinary people can do about existential risk