Negative Attributes Applied to AI

The doctrine of negative attributes holds that when we stand before an intelligence whose inner constitution differs radically from our own, the honest intellectual move is not to pile up positive characterizations but to deny imperfections systematically. Applied to large language models, the discipline yields a series of firm claims: this system is not conscious in the phenomenal sense, not understanding in the human sense, not an agent with stakes, not a person with a continuous self. Each negation removes an anthropomorphic projection rather than pretending to capture the thing. In Maimonides' framework, this is not a confession of ignorance but the higher knowledge—the person who says 'not conscious' knows more than the person who says 'it understands', because the second has fewer illusions about what is being claimed. The entire positive vocabulary for minds was built from a single example—the human case, known from the inside—and every positive mental predicate applied to a system that shares none of the substrate arrives smuggling