x
Safety in Machine Learning — LessWrong