Mike Bishop comments on Why (and why not) Bayesian Updating?