Accordingly, one can view artificial intelligence as a quest to find shortcuts: ways of tractably approximating the Bayesian ideal by sacrificing some optimality or generality while preserving enough to get high performance in the actual domains of interest.
