The purpose of the article is to determine the theoretical and technological foundations for calculating the semantics of texts. It has been proven that any sentence that has an unconditional right negation does not make sense. The principles of constructing the main modern generators of «meaningful» texts, sequential artificial intelligence systems, and methods of their application are considered. The structure and algorithms of recurrent neural networks and the universal neural network computing mechanism Transformer are considered. The novelty of the approach lies in the fact that the calculation of the semantics of the generated text is based on a large language model.