Having heard this complaint about my AI Lore books for about the thousandth time (not an exaggeration), I think I might be finally ready to concede that – in some way – my books are indeed “not real books.”
What I mean by that is that the format of an ebook (or print book) merely serves as a vehicle to deliver what amount to complex narrative networks. To quote Wikipedia on the matter:
A networked narrative, also known as a network narrative or distributed narrative, is a language partitioned across a network of interconnected authors, access points, and/or discrete threads. It is not driven by the specificity of details; rather, details emerge through a co-construction of the ultimate story by the various participants or elements. […]
Networked narratives can be seen as being defined by their rejection of narrative unity.[1] As a consequence, such narratives escape the constraints of centralized authorship, distribution, and storytelling.
Let’s put it another way, perhaps even more simply…
My books consist of sets of reference points, some of them textual, some of them image-based. The reference points are arranged in a certain order within each book, and also include hyperlinks out (physically encoded into the ebooks, as well as non-coded conceptual or thematic ones) to reference points contained in other books.
Let’s have a quick refresher on network topologies:
Instead of nodes in a network, think of them as nodes in a narrative, which consists of nodes and their relationships (arrangement) with other nodes. What’s a “node” in this context? Non-exhaustively, we could say it is something like entities (persons, places, things), events, etc. It’s a thing with some substance in a story.
Most conventional fiction could probably be represented as a pretty simple linear (line) topology. That is, you deliver one “reference point” or node, one after another, and the reader passes through them in the path laid out linearly by the author. Perhaps a choose your own adventure book might be mapped out to resemble something like a tree or a mesh, where the user chooses from among multiple pre-defined paths and branches to arrive at their own experience. And maybe a dictionary or encyclopedia might look like a “fully connected” network topology.
My books consist of kind of all of these smooshed together into a hybrid narrative network topology. Each book is a narrative node in itself, composed of many other sub-nodes and relationships. And then the reader traverses the nodes in basically any order, composing their own experience as they go along. This is not the way that I think of most other fiction books working usually. And above and beyond anything I’ve done using AI, I think this model, this structure, is what sets my books apart in the end.
If this is hard to parse, let’s pull in someone else’s diagram to help illustrate. This comes from a paper on ResearchGate, which has a set of illustrations, of which is this one, called “Narrative Network Graphs: examples of two far-right narratives in 2016.” Here’s the picture, which seems to represent narrative elements mapped as a visualization of relationships and proximity:
This is kind of a “latent space” approach to narratology, I think. And I suspect it might be somewhat aligned with how AIs “think” about narratives (I don’t think they actually think, however). When you invoke a narratively-flavored output from a generative AI service, it takes all your tokens that you input, finds the others laying around in the neighborhood that are likely to be related, and spits them back out. It outputs them in a linear order (A –> B –> C), but my hunch is that this linear order is not actually intrinsic to how AIs approach fulfillment of these tasks. It doesn’t care much about what the order is.
Humans, however, do seem to care a great deal about linear progressions through time within narrative, which Kurt Vonnegut’s bit on the 8 types of stories (of which he only depicts a few – see the rest here) ably illustrates:
I suspect the reason AI often crafts “shitty” narrative progressions is that 1) it is not intrinsically concerned with the order of presentation, only that nodes and their relationships are represented, and 2) it has no lived emotional experience, so has to make guesses as to what outputs ought to trigger which emotional states in humans.
The thing is, though, I like that weird quality, the Uncanny Valleyness of it all. The fact that it struggles and sputters with narrative unity. I like that AI currently does NOT actually fundamentally understand what makes a good, rich, and interesting story to humans. That failure, if interrogated well and empathetically, can actually be terrible fascinating all on its own. But it doesn’t make good “regular” books – yet. That day will come though.
So for me ultimately, what I want to say is that the outward form of an ebook or printed book is “fine” for me for now, because it is a common, well-understood, and more or less efficient means to distribute chains of reference points, or networked narrative nodes and their relationships. The same underlying nodes could be presented in countless other ways (lists, image sets, videos, immersive VR experiences, endless others), and over time I hope I have the opportunity to explore those other directions of AI-assisted storytelling, and where they intersect with “The Book” and where they can transcend it.
While I’m on this topic, here is an – I think – previously unreleased PDF document I made some six years ago (2018!), back when generative AI was barely a twinkle in Bill Gates’ eye. It predates any of the Quatria books, and it absolutely predates the AI Lore books, focusing more on Early Clues LLC, and its many exalted offshoots.
Even though it predates all of those things, it gives a fairly accurate (as these things go) “skeleton key” to understanding the rest of my extremely messy and convoluted networked narratives. Skimming around in this diagram cloud, I think, also gives a good visceral experience of what it’s like to try to navigate the stories that pass through all my other books – where the reader/viewer is largely left to their own devices to make sense of it all.
Leave a Reply
You must be logged in to post a comment.