Learning representations - History — AI Alignment Forum