The use of Bayesian networks is discussed and demonstrated. There are two main lessons: 1) information is accumulated even from "weak signals", ie several noisy syptoms 2) information dispapears quickly in complex models, even in case where uncertainly is low between single variables