Work

The Chinese Mind and A Chinese Room: Patterned Language and Meaning in Generative AI

Year: 2026
Type: preprint
Abstract: This paper reconsiders Searle’s Chinese Room argument in light of contemporary large language models (LLMs) and selected resources from classical Chinese philosophy. I argue that Searle is right that formal symbol manipulation is insufficient for full understanding, but that his opposition between syntax and semantics is too rigid for the context-sensitive and pattern-tracking of generative AI. Drawing on Zhuangzi and Xunzi and classical Chinese vocabulary, I develop an account of meaning as relational, norm-governed participation in practices of distinction making. I argue that meaning runs in two directions, from the world to the mind and language back to what becomes salient to the mind. With this, I suggest linguistic patterning carries a directional force that shapes affective and normative responses regardless of the source. This grounds four tiers of semantic competence: empty formal manipulation, derived textual meaningfulness, relational semantic competence, and full lived understanding. LLMs may occupy the middle tiers, exerting directional force without consciousness, intentionality, or embodied salience. This blocks any inference to human-like understanding while allowing a more detailed account of LLM semantic competence. The final payoff is ethical. I suggest LLMs may shape public language without being accountable for the linguistic and normative distinctions drawn.
Author: Huang, Yijia
Language: English
Subfield: Philosophy
Open Access status: green
OpenAlex ID: W7202329324
Cites: 34
Cited by:
Related to: 10