“Communicative AI: A Critical Introduction to Large Language Models” By Mark Coeckelbergh / David J. Gunkel
What Can We Generate in Return?
Relying on writing makes us forgetful. That was Socrates’ warning about the written word in Plato’s Phaedrus. Written words, he argued, can only repeat the same answer no matter what questions we put to them. The two authors of this book offer a bold reinterpretation of this ancient concern: writing has always been an artificial technology requiring tools. In that sense, “writing” itself was already a kind of AI.
The authors argue that generative AI is shaking the very “operating system” underlying modern Western thought. Within this system, speech is regarded as directly connected to thought, while writing is treated as merely a secondary system of signs—a structure the philosopher Jacques Derrida called logocentrism. Nor is the “author” a universal concept. As the critic Roland Barthes observed, modern Western culture has often imagined the relationship between author and work in terms of parent and child. It is perhaps no coincidence that the English word plagiarism ultimately derives from a word meaning the kidnapping of a child.
Where, then, does meaning arise? Following Barthes, the authors answer: “The unity of a text lies not in its origin but in its destination.” Meaning emerges not from the writer’s intention, but in the act of interpretation by the reader. The criticism that large language models—the technology behind generative AI—do not understand the meaning of words is correct. But it would be premature to conclude that their outputs are therefore meaningless. As with any text, meaning arises on the reader’s side. AI has simply made this process visible.
I have long believed that “to read is to write,” so I find this reversal compelling. It also points toward a productive way of using generative AI. I can give AI something I have written, read the response it returns, and use that response as a foothold from which to write again. What matters is not how good the AI’s output is, but what I myself can generate in return. This loop can become a means of understanding oneself anew.
The book concludes that the future of “writing” lies in reading: in choosing what and how to read, and in curating what we encounter. Nor is this necessarily a solitary activity. Like a reading group, it can be a collaborative practice that unfolds through discussion with others. The book embraces the disruption brought about by generative AI as an opportunity to move beyond the assumptions of modernity. Provocative as it is, it is also a book that generates an image of hope. Translated by Akio Tabata. (¥2,640)