Three (or Maybe Four) Ways of Thinking AI Philosophically

This essay develops a critical, non-reductive account of large language model (LLM) AI by examining how this technology reshapes language, labor, and life. It considers four philosophically distinct approaches to AI, including: the classical cognitivist debate running from Turing to Searle; the discourse of existential risk, developed by Nick Bostrom and others; Marxist theories of technology; and Michel Foucault's account of the relations between power, knowledge, and subjectivity. It questions whether agentic misalignment warrants conceptualization in terms of existential risk, suggesting that the problem needs to be interpreted from the standpoint of what we know about the alignment of human reason under late capitalism. It proposes a definition of generative AI as "fixed capital accumulated until it becomes discourse," and uses this definition to sketch a synthesis of the Marxist and Foucaultian approaches. In general, this essay argues that AI should be understood not merely through debates about the meaning of intelligence or questions of applied ethics, but as a problem for a critically engaged social philosophy.