A representation of an untrained human brain, call it B(0), is encoded in the human genome -- its size can hardly exceed a few megabytes. In contrast, a representation of a trained brain, B(t), after big enough time t (say t=20years) must be very long (terabytes?) – it must include a representation of the brain's individual experience. How can a "simple" B(0) change into an extremely complex B(t) in the course of learning?