paulfchristiano comments on The Learning-Theoretic AI Alignment Research Agenda