Bayeswatch 5: Hivemind — LessWrong