SIAI May report — LessWrong