One should update on evidence according to Bayes’s rule [because/if.., what? you want to achieve best calibrated posteriors? Because Eliezer says so? Because you want to maximize your odds of something?]
These shoulds are all different. (..) and I am not sure about the last.
That’s why I mentioned it. It has an unconditional quality that Wei Dai’s examples lack.
That’s why I mentioned it. It has an unconditional quality that Wei Dai’s examples lack.