Towards Generative Procedural Writing

 

Writing is not a one-way path from author to passive reader: it’s a dynamic and interactive process. In the age of AI, text can behave like a musical instrument or a program: a system that unfolds new ideas, styles, and even ethical questions as it’s played.

Reading becomes conversational and non-linear. More and more people consume information in fragments, jumping from one part to another, weaving multiple narratives into unique textures that make sense for them. They might not read the whole book but instead ask ChatGPT to provide its essence. They might be interested not in content but in the style. They might not look for a summary but the parts that are missing.

Let us set aside the criticisms regarding short attention spans and the impact of AI on our reading habits. While this observation may be accurate, it also reflects a shift that has already occurred in the past. Every new technology changes our habits, and somehow we find a way to evolve with it. What is the transformation that’s happening now?

Writing and reading are evolving into generative and procedural practices. A single text is a multidimensional system of vectors where certain concepts and topics are aligned in a certain way. When it is unfolded, these relationships come into play, and probabilistic engines deliver their version of reality that will be as unique as it is different from the rest. A text can be seen as a system of rules that are activated every time a person engages in an interaction with it. A system of principles, stylistic choices, axioms, beliefs, ideologies, and imperatives comes alive as it’s generated in real time.

Writing a text for this medium then becomes an exercise in setting an improvisational score. We don’t know exactly how it will be played, but we know that it’s going to have a certain bias, style, and aesthetic appeal. We know the relations we put into space will pull the meaning into certain directions that will be different from the generic, common denominator version of this reality. When the reader engages, they will pull the threads that resonate with them, and the text will respond accordingly.

That’s why we think that it can be interesting for authors and publishers to embrace the AI as a medium while also, of course, addressing the criticisms and concerns. AI should not replace the books, but it can provide an additional conversational medium to interact with the book’s content. For example, every book that we published on Circadian has a visual, auditory, and text chat representation. It opens up additional modalities to interact with its content.

For instance, we can visualize a book’s content as a knowledge graph, emphasizing its main topics and ideas. The reader can click on ideas they find relevant and get to the part of the book that resonates. They can also explore the relations between the ideas proposed by the book, reading it as a network of relations rather than following a linear narrative.

Alternatively, readers can talk to all of our books using a text chat interface. It allows us to embed agentic AI workflows that communicate Circadian’s goal to put words into action: allowing the books to speak to a particular situation our readers are interested to discuss. It also can be used to make books speak to each other and see how ideas from one book connect or contradict ideas from the others.

Voice interface is another version of this approach. It adds more intimacy to the conversation and lets the reader focus on the aesthetics of the book. It also helps engage the readers who would not normally like to read or write. Its auto-translation capabilities open the content up to the people who don’t speak the language the book was originally written in. NotebookLM can be used to turn books into conversations where the listener can intervene and steer the direction of the narrative.

This technology is still in its nascent form, and there are a lot of issues that have to be addressed: from content ownership and its impact on the environment to making it less synthetic, more human-like, and ensuring that content does not take over the style. However, as a new medium, it has a lot of potential and we at Circadian are happy to work with writers and publishers who are willing to try it out.

You can learn more about some practical ways you can do that at www.circadian.co/ai