Beyond control: Will to Power in AI

Sensational claims portray AI as an entity beyond human control. We argue that while AI may exhibit characteristics that appear “beyond control,” this does not necessarily arise from its status as an independent entity. The prevalent “out-of-control” AI narrative obscures meaningful discussions on how we can effectively govern AI. This paper introduces a novel theoretical framework based on two variables: controllable/non-controllable AI and value-free/value-laden AI, yielding four distinct perspectives on AI ontology: Value-Free Instrumentalism, Determinism, Value-Laden Critical Theory, and Substantialism. We contend that AI's perceived lack of control stems not from value-free mind-independence, but from value-laden sociocultural power dynamics. Through the lens of Critical Theory, we elucidate that AI remains subject to control, despite prevalent perceptions suggesting otherwise. While AI may present legitimate existential concerns, these do not stem from an inherent “out-of-control” state, but rather from the exertion of control by entities whose motives may not align with public interest. This analysis aims to shift the discourse from panic-driven narratives to a nuanced understanding of AI's controllability, fostering more effective approaches to AI governance and development.