This essay explores an unexpected affinity between contemporary large language models (LLMs) and the tradition of ordinary language philosophy associated with Wittgenstein and Cavell. On the philosophical side, meaning is understood not as an inner representation or a rule-governed mapping, but as something grounded in use, judgment, and shared criteria within a form of life.
Against this background, the essay examines LLMs as systems that achieve remarkable linguistic fluency by learning patterns of use at scale. Their successes and characteristic limitations illuminate a distinction between reproducing linguistic regularities and participating in the normative practices that make meaning binding. Framed this way, the essay engages with current debates about understanding in modern AI, and reconsiders central questions concerning meaning, alignment, embodiment, and self-improvement.
📄 Download the full essay (PDF)
Final essay.
Yufa Zhou — December 19, 2025