An approach is proposed meaningfully interpreting the concept of an autonomous agent - an element of the model of an “artificial society" as something “whole”, “alive” and “conscious”. The goal is to lay the foundation for computer modeling of the phenomenon of “life and conscious activity” of an agent in “society as a multi-agent system". For analysis the concepts “universal Turing machine (UMT)” and the principle of UMT self-applicability as a way of its self-programming were used. It is shown that as a result, the problem of algorithmic non-computability of the UMT self-management problem can be removed.