Some day fabricated neural networks will potentially become genuinely self-organizing and self-learning, but thus far this goal remains elusive. Most networks rely on supervised knowledge. Their education process is guided by human judgment concerning what the “right answers” are to a set of education cases. The network tries to reproduce human judgment - first for the education cases and then for cases which have not been analyzed by humans. It is relatively easy to watch if the network is succeeding in replicating human judgment concerning the test cases and additionally practical with respect to spot checking the alternative cases. However, if the “right answers” are not known, then it is tough to evaluate the output of the simulated neural network.