can someone explain why ppl now say that rationalist bayesian updates are wrong not just impractical?