The idea can be said to have originated around the 1950s, when the perceptron, Frank Rosenblatt's mathematical model of the brain, was invented. It was the first attempt to model the brain's neural network. Rosenblatt then created an electronic device that was intended to mimic human thought processes. It was not a breakthrough, but the idea had already emerged and scientists from various countries began to work on creating artificial intelligence. These works knew several "winters" and "winters", periods when researches were stopped and restarted and inspired new scientists. In general, it is possible to distinguish two big periods and two opposite paradigms of research of artificial intellect. The first period, and historically the first, when the central idea was to create so-called artificial intelligence implemented in expert systems. A machine was taught rules according to the "if-then" principle, and these rules were the basis of a particular expert system, allowed to obtain knowledge. It was a deductive approach, the principle "from the general to the particular". This wave of research "froze" mainly because it is difficult to find experts in different fields of knowledge, to formalize all the available knowledge and besides the knowledge itself is constantly changing, new discoveries and new systems appear and such expert artificial intellect could not keep up with the real development of science. The second period, the second wave began in the 1970s and 1980s, when they began to develop the opposite, statistical approach. These were developments in machine learning, an inductive approach from the particular to the general, from examples to dependencies. Like a child learning about the world, for example, you show him a maple leaf and an oak leaf, and you explain the differences and similarities: both are leaves, but this is a maple leaf, and this is an oak leaf, and these are the differences. And the child learns to recognize and identify the leaves. In the same way, machines were trained to recognise patterns, and to infer patterns from them. It was the idea behind neural networks. After all, what is a neuron? It is a very specialized cell, an additive classifier that conducts electrical stimulation and chemically interacts with other neurons, receiving, processing, conducting, storing and outputting information from the outside. When developing artificial intelligence, they wanted to make an analogue of the brain's neurons. It did not work, but they called what they did neural networks, because they, too.