Rational Agents Cooperate in the Prisoner's Dilemma — AI Alignment Forum