<p>This article rethinks the foundations of artificial intelligence (AI) by challenging both the logocentric assumptions of “good old-fashioned AI” and the ungrounded empiricism of contemporary large language models. Instead of treating language as an abstract, universal code, it reconstructs a plural genealogy of empirical semantics that runs through Wittgenstein and Margaret Masterman into Yuen Ren Chao’s continuity-based Chinese linguistics and the cybernetic reconfiguration of language as signal and information. Chao’s analyses of tone, vernacular writing, and script reform reveal a Chinese-inflected empiricism in which categories are graded, usage-based, and technologically mediated. The paper then turns to contemporary Chinese thought—translingual practice and cosmotechnics, figure–landscape semiotics and classifier-driven ontology, verb-centred models of existence and <i>tianxia</i> politics—to show how different linguistic and cosmotechnical traditions carve up continua of experience in non-identical ways. On this basis, it proposes plural intelligence as an alternative horizon for AI: not a single, disembodied AGI, but an ecology of situated, culturally responsive intelligences whose operations are continuous, figural, and relational rather than anchored in a universal <i>logos</i>.</p>