On Universal Prediction and Bayesian Confirmation