The Machine that Refutes Descartes

The real epistemic shock of modern language models lies not in their size, not in their fluency, not even in their ability to sustain entire discourses. It lies deeper: in the quiet dissolution of a philosophical assumption that has shaped our understanding of machines for centuries. That assumption is that mechanics and meaning belong to different orders. Here dead structure, there living sense. Here res extensa, there res cogitans. Here signal, there mind. In the classical symbolic machine, that separation could be maintained with remarkable plausibility. In high-dimensional neural vector spaces, it begins to collapse — not because mathematics has suddenly become magical, but because mathematics itself takes on a form in which relational structure already carries semantic function. Meaning is no longer something projected from the outside onto a form that is, in itself, meaningless. It appears instead as the inner geometry of form itself. This is not a minor technical detail. It is a shift in the ontological status of representation. This essay shows why the transformer forces this shift, why it is not a metaphor, and what follows from it — philosophically, empirically, and for our understanding of meaning, values, and intelligence as such.