Abstract: The retrospective analysis of the evolution of concepts of constructing artificial neural networks in the context of the possibility of their further application to create artificial intelligence is carried out in the article. There are considered key achievements of international scientific community in this field starting from the first attempts of formalizing the neuron model and to the latest implementations of neural computing systems. The study concluded that there is the loss of researchers’ interest in constructing of an accurate model of the biological neuron and shifting attention to the creation of computer systems based on synaptic structures.

The analysis also helped to identify the relationship between neurobiology achievements and qualitative leaps in field of construction of artificial neural networks, which shows the importance of further study of biological neural network structures and the need to create the most exact model of the biological neuron, as a key structural unit of the neural network.

The conclusion is that there is an absence of a single unifying concept concerning the creation of biologically plausible artificial neural networks, so the author proposed the concept of the neuron structure decomposition into elementary biological functional subsystems. A feature of the proposed concept is the possibility of modeling of each neuron subsystem individually with further association of the components into an integrated system. Thus, the developed concept considers the possibility of adding and modifying elements of the neuron model for refinement and complication of its design in order to bring it to the maximum correspondence with the biological prototype.

The practical value of the proposed concept lies in the possibility of developing a neural network structure on a fundamentally new level of quality with ensuring of the maximum similarity with the nervous system of a living organism