Mark_Friedenbach comments on To contribute to AI safety, consider doing AI research