Examples of AI's behaving badly — LessWrong