On the Diplomacy AI — LessWrong