Mechanisms of Generation of Fake Information by Artificial Intelligence in Modern Media Discourse

Aim. To try to investigate the impact of the main mechanisms of fake information generation on society in the modern media discourse. The hypothesis of the study is that with the development of artificial intelligence (AI) capabilities, the generation of fake information occurs with the help of deep cognitive-linguistic models, because "strong" AI is based on the generative architecture of neural networks, which is able to "learn", to solve complex tasks without any human intervention, which is an existential threat to civilization. Methodology. In analyzing the empirical material, we used a number of methods and techniques to analyze the texts generated using neural networks, such as contextual analysis, generalization and systematization method, component and comparative analysis. Results. We have analyzed fake information generated by neural networks, and also conducted an experiment to check the degree of truthfulness of information generated by ChatGPT which turned out to be 99% fake. In addition, we conducted an experiment using the "argument" option of the GPTunnel neural network aggregator, the purpose of which was to reveal the quality of argumentation on a given topic. The experiment showed that the claimed "ability to think logically and consider arguments" was not confirmed. However, rapidly developing AI and increasing capabilities of neural networks, might in a very short time create a more advanced neural network capable of generating logically constructed argumentation. Research implications. The results of the research contribute to a deeper understanding of the mechanisms of fake generation using generative neural networks in modern media space, which in the future will help linguists solve problems related to the identification of texts generated by AI.