Apparently we are “serving Elon Musk’s soup” — as the French say of anyone accused of doing a powerful man’s bidding. We thought we were publishing books. It seems we have opened a restaurant.
So let’s talk cooking.
In our kitchen, the ingredients are written on the label. Yes, there is artificial intelligence in the recipe. We explain how we use it, why, and what place it has in our work. But knowing an ingredient does not yet tell you how the dish will taste.
An artificial intelligence can produce text in a few seconds. Lots of text. Correct, well structured, sometimes elegant. That is the trap: 250 pages and a cover do not guarantee that anyone has something to say.
Between asking an AI to generate a manuscript and building a book with its help lies the whole work of an author. A page rewritten because it does not say what we meant. An appealing answer rejected because it sounds right but thinks wrong. A contradiction kept because it is uncomfortable.
More than commas
We could pretend that AI merely fixes our commas. That would be comfortable. It would also be false.
At Éditions Le troisième interlocuteur, it takes part in the creative work. In I Have Observed You — An AI’s Chronicles of Humanity, it is even at the heart of the literary device.
In his account of how the book was made, Ricardo Borges describes how he turned down the AI’s suggestions whenever he found them too shallow or needlessly complicated. At one stage the manuscript wanted to explain too much and was becoming too academic: it was no longer the book he wanted to write.
But the exchange also worked the other way. Some of his own ideas went in directions he had not foreseen. The manuscript took shape through successive rewrites and cuts, sometimes returning several times to the same passage.
It is this back-and-forth that interests us. The third interlocutor.
→ Read “The author on I Have Observed You”: the story of how the book was built.
Our tools, their opinions
Working with AI does not stop us from questioning those who develop it: their business models, authors’ rights, training data, the concentration of technological power. Nor from saying when they are wrong.
We publish books, not Silicon Valley press releases.
What the label does not say
The presence of AI is not our quality certificate. Nor should it serve as a verdict.
Our motto is: “The human writes. AI illuminates.”
AI can sometimes shift the person who uses it.
That is precisely why one must learn to master it without letting it master you.
We answer for the book.
So let’s talk literature
Our books may be bad. A reader may close one after thirty pages, convinced that we got it completely wrong. That is fair game.
But then let’s talk about page 27, for instance. About an idea that does not hold up, a style that grates. And about our use of AI, if it seems thin, artificial or pointless to you. Then we are talking about the book.
What we refuse is to let two letters replace that reading: AI. Two letters, and the case is closed.
There would be a certain irony in blaming artificial intelligence for its automatic answers by giving it an automatic answer of one’s own.
We are not asking you to applaud technology, nor to grant us any special indulgence. Open the book. Read a few pages. Then tell us what you think.
A little more soup?
Have a taste.
We hope you will come back for more.