Extraction of human preferences 👨→🤖 — LessWrong