Infrafunctions and Robust Optimization — AI Alignment Forum