Wei_Dai comments on My current take on the Paul-MIRI disagreement on alignability of messy AI