The problem of subjectivity of neural networks: humans and non-humans

Abstract

The article discusses the significant problem of how much the quality of subjectivity can be transferred to an AI neural network. The main goal is related to the analysis of the effects of the dual transformation of subjectivity, caused, on the one hand, by the fusion and hyper-interconnectedness of the modern individual and the digital world, on the other, by the func­tioning of artificial neural networks, demonstrating their autonomous desire to represent hu­man subjectivity. The focus is on a complex of interrelated aspects. Firstly, the fundamental characteristics of traditionally understood subjectivity with the predominance of reflexivity and intentionality of subjective experience are considered. Secondly, based on modern lite­rature, the concept of artificial neural networks is revealed, their specific features, structure and varieties are analyzed. The arguments “pros and cons” of the hypothetical recognition of the subjectivity of AI neural networks are discussed, and attention is drawn to the phe­nomenon of the information “global workspace”. Thirdly, the setting of neural networks for overperformance and its place in the analysis of hallucinations of artificial neural networks is revealed. Fourthly, the importance of the prompt (request) and the development of interac­tive AI, which can act not only as a customer, but also as a manager and controller, is as­sessed. It is concluded that the modern state of subjectivity acquires the qualities of an ac­tant who is forced to implement a sequence of functions necessary for “life in the digital world”.