Problem of fully updated deference - History — AI Alignment Forum