Beyond Reason: Individuation of Gilbert Simondon in the Era of Computational Sovereignty and Automated Cognition

Digital technologies are radically changing the modern world, increasingly acting as autonomous agents. Algorithms governing social networks, neural networks creating content, or robots making decisions. Their actions are determined by their own internal logic, beyond human intentions. While many researchers in digital technology argue for the capability of machines to think based on their functional similarity to human reasoning when processing information, the outstanding philosopher and thinker in the field of technology and technological innovations, Gilbert Simondon, shifts the focus to the specificity of technical being and the collaborative nature of intellectual activity. The question "Can machines think?" transforms into a question of how we understand thinking within the complex interaction of the technosocial system "human-technology-society" with its own evolutionary dynamics. The methodological foundation of this article is based on methods of analysis, observation, and synthesis, consisting of the key ideas of philosophical individuation of technical systems (Simondon) and the ideas of genealogy, control, and machine cognition (Deleuze). The key task is to reveal the contingent process of the technosocial phase of individuation that produces operationally closed (autoetic) systems. Classical cybernetics, relying on the servomechanical model, interprets machines as passive executors of pre-set commands, denying the possibility of genuine "thinking." This article challenges this paradigm in relation to the modern machine learning process of neural networks. We assert that the servomechanism is incapable of adequately explaining the emergent, unpredictable, and creative aspects of complex models' behavior, especially deep neural networks, which go beyond simple responses to input stimuli. Drawing on Simondon's philosophy of technology, the article proposes an alternative ontological framework. The process of machine learning is interpreted as contingent technical individuation: the model arises ("individuates") from a "pre-individual" field (architecture, data, learning stochasticity) through the resolution of internal tensions (minimization of losses). The result is not a servomechanism but a unique operationally closed system. Its "thinking" is understood as an emergent property of this individuation – contingent (dependent on the learning path), immanent to the internal structure (weights), and manifesting in the ability to generalize, generate, and abduct.