Foucault, AI, and the Death of the Author Reimagined
Reflections from an Author in the Age of AI
Because of my learning curve with the core features of Substack, I believe this never got published as a post. But if you’ve seen it before, my apologies. Nonetheless, here are few thoughts of Foucault’s philosophy and AI…
We tend to think of writing as something that comes from deep within—a moment of insight, inspiration, or genius. But what if it doesn’t? What if, instead, writing is more like tuning in to a conversation that’s already been going on long before we arrived?
That’s where Michel Foucault’s ideas might resonate in the age of AI. He suggested that we stop picturing the author as a solitary genius and start seeing writing as something shaped by culture, history, and community—or what he called discourse.
“The author is a certain functional principle by which, in our culture, one limits, excludes and chooses.”
— Michel Foucault, “What is an Author?”
In his seminal 1969 lecture “What is an Author?”, Michel Foucault could not have know that he was laying the theoretical groundwork for understanding one of the most significant technological developments of the 21st century: LLMs or large language models. His radical problematization of authorship exhibited a prescience that anticipated the mechanisms by which AI systems like ChatGPT generate text.
The Death of the Author, Rebooted
Foucault challenged the Romantic idea of the author as a unique creative originator. Instead, he proposed that discourse itself—the episteme—operates through us, not from us. The author is not a godlike source of meaning but a channel through which pre-existing structures of language and knowledge crystallize into text.
Foucault connects this disappearance with the Nietzschean declaration that God is dead:
“It is not enough, however, to repeat the empty affirmation that the author has disappeared. For the same reason, it is not enough to keep repeating that God and man have died a common death. Instead, we must locate the space left empty by the author's disappearance, follow the distribution of gaps and breaches, and watch for the openings this disappearance uncovers.” Michel Foucault
For this post, I’ll bracket his atheism—with the intent of returning to it. Still, this reframing of authorship fits remarkably well with how large language models work.
The AI System as Discourse Conduit
When ChatGPT generates a response, it doesn’t “create” in the traditional sense. It draws upon the entire archive of human discourse—the training data it was built on—composed of millions of texts, styles, and voices.
The AI becomes a sophisticated conduit, not a conscious agent. Like Foucault’s author, it is a part of a vast network, culling discourse. It is not an originator of meaning.
The LLM doesn’t express intention. It expresses pattern. It doesn’t know—it mirrors the shape of knowledge.
If Foucault’s author is a cultural construct used to organize meaning, then LLMs strip away the illusion entirely. In doing so, they expose the machinery of discourse itself.
Foucault questioned how authorship functions socially. But before proceeding further (in this text I seem to be authoring): it isn’t that I have to agree with Foucault entirely, but by extricating my hermeneutical framework from a Romantic authorial creation, I can see how AI seems to author texts analogously.
The loss of an individual author is not an entirely new idea to me. It was striking when I studied Homeric Greek in college, and specifically The Odyssey, that scholars don’t know if there was one individual named “Homer” who authored this epic. But, in a way, the ancient Greeks didn’t really care. It represented their culture and their values, the configuration of knowledge and discourse that Foucault called an episteme.
In a weird way, where once we granted authority through human authorship, AI-generated texts now can command authority because they lack individual authorship. A strange inversion of Foucault’s approach is occurring. Assigning authorship becomes distributed. It can become not a person, but a process.
So, why do AI outputs feel eerily “almost human”? Foucault’s theory helps us understand: these texts are born of the same discursive structures, just without subjectivity.
They are the ghost of discourse—structure without soul. Or to use Foucault’s words, “today's writing has freed itself from the theme of expression.” He continues,
“Writing unfolds like a game that invariably goes beyond its own rules and transgresses its limits. In writing, the point is not to manifest or exalt the act of writing, nor is it to pin a subject within language; it is, rather, a question of creating a space into which the writing subject constantly disappears. The uncanny feeling emerges because the form is familiar, but the hand behind it is absent.”
The Machines become the message
LLMs don't mark the beginning of machine creativity. They replicate what Foucault articulated decades ago. And this fact poses for use questions: Does language precede intention? Is it discourse speaks through us?
I’m not sure many of us resonated with the chilling question that closes “What is an Author?” “What difference does it make who is speaking?”Because, in everyday discourse, we want to—and need to—fix ourselves on answering that question.
But we might begin to distrust what the Romantics taught us about authorial creativity—and the degree to which Romanticism still flows through us culturally. They told us to think that the author is the genius behind a text. Foucault flips this idea: the text exists because of discourse, not because of the author. And that concept is easier today, well, to conceptualize.
Obviously, Foucault didn't live to see LLMs—but his work anticipated our experience. In systems that generate coherent, meaningful text without human authorship, we see not artificial intelligence, but a perfect mirror of what writing has always been: A product of the episteme, of systems, structures, and histories, but not of individuals.