Is it worth making a database for moral predictions? — LessWrong