"Who is speaking?" in the context of communication between a person and a neural network

The object of study is the communication situation between a person and a neural network in the form of large language models (LLMs), while its subject is the language model as a communication agent. The author examines aspects of the topic such as the subjectivity and agency of large language models, the very possibility of communication between humans and neural networks, the specifics and mechanisms of its implementation in the human-language model system, hypothetical prospects and consequences, as well as challenges and risks of communication interaction between humans and such new technical objects. In our work, we aimed to explore the key aspects of human communication with neural networks by utilizing the heuristic potential of modern philosophical concepts—Graham Harman's object-oriented ontology and Bruno Latour's actor-network theory. The methodological foundation of the work is based on Graham Harman's object-oriented ontology and Bruno Latour's actor-network theory. The empirical data consisted of the results of a neural network study based on the Turing test by the University of California (March 2025) and the findings of a content analysis of the chat-bot constitution of the company Anthropic’s Claude. Specific aspects of the problem are illustrated with excerpts from interaction protocols with DeepSeek. The novelty of the research lies in the application of modern philosophical concepts (object-oriented ontology and actor-network theory) to analyze the problem of subjectivity and agency of large language models, which allowed the author to conceptualize the communication of human and non-human agents as a heterogeneous network of interactions, revealing its opacity, introversion, and asymmetry, as well as the stochastic nature of generation and the unpredictability of potential communicative effects. A significant contribution by the author to the topic's research is the proposal to use mathematical terms such as "infinite set" and "transfinite number" for a hypothetical assessment of the projected power of the actors "human" and "neural network," necessitating further development of this approach. The results obtained allow for a critical reflection on the practice of communication between humans and neural networks to enhance the level of awareness and reflexivity of human participation.