The central limit theorem in terms of convolutions — LessWrong