How to impress students with recent advances in ML? — LessWrong