New Hackathon: Robustness to distribution changes and ambiguity — AI Alignment Forum