Here is a provocative thought: maybe the Chinese Room argument feels so compelling partly because we have the wrong theory of meaning.
Searle's argument assumes that meaning is a relationship between a symbol and a concept (or a referent) in a mind. If there is no mind with concepts, there is no meaning, just shapes. This is a picture of meaning that was already under pressure before LLMs arrived.
Wittgenstein's later philosophy (the Philosophical Investigations, 1953) argued that meaning is not a private mental state that words express, meaning is use. The meaning of a word is its role in the practices, language games, and forms of life in which it participates. "Hot" means what it does because of the way it functions in an enormous range of practical and linguistic contexts: warnings, recipes, weather talk, desire, physics. You understand "hot" not by having a private mental image of heat but by knowing how to use the word, in what circumstances, with what consequences, in what language games.
Now ask: what do LLMs do? They learn, from extraordinary quantities of human language, the patterns of use of every word and phrase: what contexts it appears in, what it tends to follow and precede, how it interacts with other expressions. In a sense, LLMs may be the first systems to have learned the full pragmatic profile of language at scale, not just definitions but the entire web of use.
If Wittgenstein is right that meaning just is use, and if LLMs have internalized use more thoroughly than any individual human ever could, then the question "does it understand?" starts to look more complicated than Searle suggested. It might be a question about thresholds and kinds of use, not a binary yes/no.
But there's a sharp counter-argument, which Merleau-Ponty would love: Wittgensteinian forms of life are not just linguistic. They are embodied, they involve doing things in the world with a body that has stakes, needs, sensations, and vulnerability. "Hot" is embedded in a form of life where things burn, hurt, comfort, destroy. A system that has never been in that form of life, that has never reached for something and pulled back, may be learning the shape of language use without the ground from which use grows.
This is perhaps the most interesting open question in philosophy of mind right now: whether embodied, sensorimotor engagement with the world is necessary for genuine meaning, or whether purely linguistic training, at sufficient scale, can produce something functionally equivalent. No one knows. The LLMs have arrived before the philosophy has caught up.