DARPA accepting proposals for explainable AI — LessWrong