[Work of large information systems].

The problem of application of informational ideas to the study of large informational systems, such as brain, is considered. Analysis follows some ideas of Winograd and Kowan's theory of reliable computations in the presence of noise. In the beginning of paper the "material analogy" for intuitive understanding of Shannon's results is proposed. Afterwards there is given the formulation of coding theorem for the case of "informational collective"--a large group of informational sources and receivers. In conclusion it is shown that such properties of neuronal structures as "stochastic" neural connections are reasonable from the point of information theory.