Do Large Language Models have cognitive abilities? Do Large Language Models have understanding? Is the correct recognition of verbal contexts or visual objects, based on pre-learning on a large training dataset, a manifestation of the ability to solve cognitive tasks? Or is any LLM just a statistical approximator that compiles averaged texts from its huge dataset close to the specified prompts? The answers to these questions require rigorous formal definitions of the cognitive concepts of "knowledge", "understanding" and related terms.

In this paper, I propose some new approach to determine the possibility of introducing artificial consciousness into information systems. The new rigorous formulations of the basic cognitive concepts outlined in the article are in themselves a valuable result. The implementation of basic cognitive concepts in the real AI software is impossible without adapting well-known, but verbose and vague formulations of such concepts to the requirements of Information Technology. In this article I propose possible “codable” formulations of such cognitive concepts, including “consciousness”.