11 heuristics for choosing (alignment) research projects — LessWrong