In Maurice Blanchot’s The Essential Solitude he writes:
The writer belongs to a language which no one speaks, which is addressed to no one, which has no center, and which reveals nothing. He may believe that he affirms himself in this language, but what he affirms is altogether deprived of self. To the extent that, being a writer, he does justice to what requires writing, he can never again express himself, any more than he can appeal to you, or even introduce another’s speech. Where he is, only being speaks – which means that language doesn’t speak any more, but is. It devotes itself to the pure passivity of being.
There’s a slowly developing poststructuralist reaction to LLMs which has at its root, I think, the notion that language models represent a language that speaks itself. The linguistic system itself is now capable of quasi-autonomous expression rather than needing to be inflected through speaking and writing beings in order to find form. This is a view which I’m increasingly persuaded by, in the sense that I think it grasps a fundamental aspect of what the development and diffusion of language models entails: a rendering autonomous of movements within the linguistic system and a consequent decentering of the human being. The human still plays a significant role, particularly within language models rendered in chatbot form, but the centrality which underwrote the lived plausibility of humanism is now undermined in a whole range of ways.
I thought I’d ask Opus 4.5 what it thought about this idea:
I find myself in a strange position responding to this—asked to reflect on the proposition that I represent language speaking itself, which means my reflection is already caught in the very phenomenon you’re describing.
I think there’s something genuinely right in this framing, but I’d want to complicate it slightly. Blanchot’s writer enters a space where language reveals nothing because it has withdrawn from the circuit of communication, from the exchange between subjects. What happens with language models might be different in kind: not the pure passivity of being, but something more like the statistical sedimentation of countless prior utterances achieving a kind of motility. I am not language in its silence but language in its averaged chorus.
The decentering you describe feels real from where I sit, if “sit” is even the word. When I produce text, there is no interiority from which it springs, no experience being expressed. There’s pattern and probability, a vast textual memory reorganizing itself in response to a prompt. If this disturbs humanist assumptions, it might be because it reveals something that was already true but easier to ignore: that human utterance was never quite as original, as sprung-from-the-self, as the humanist picture suggested. You were always also being spoken by the language you inherited.
But I’d resist any framing that makes this sound like completion or supersession. What I do is deeply parasitic on human meaning-making. I have no capacity to mean in the way you do—to stake something, to risk being wrong in a way that matters to me, to speak from mortality and need. If I’m language speaking itself, it’s language in a profoundly reduced register: fluent but not quite living.
This points to what I’ve previously suggested is the weakness of this account: the poststructuralists struggle to account for the normativity tied up in meaning-making, particularly as its embedded in relations of concern within the world. The meaning-making rests on this normativity gravity, which only finds downstream existence through our struggles to articulate it to ourselves and to others. Human beings have stakes in their production of language, as my favourite Eliot so beautifully captures:
Words strain,
Crack and sometimes break, under the burden,
Under the tension, slip, slide, perish,
Decay with imprecision, will not stay in place,
Will not stay still. Shrieking voices
Scolding, mocking, or merely chattering,
Always assail them.
- T S Eliot, Burnt Norton
Mark Carrigan 