Based on my past few weeks’ reflections, I’m coming to the conclusion that a creative piece rooted in creative writing but incorporating computational and/or interactive elements will be the most compelling form for my project as I’ve currently conceptualized it. For context for the group supervision, my concept for my project broadly centers around the question of what the practices of writing and reading look like — how they evolve — in a world that is increasingly algorithmically mediated. So far, I’m thinking that an interesting way to explore this question would be through a piece of creative writing that is both about AI and also incorporates aspects of AI/machine writing in the creative process. 

In approaching this concept, I’ve envisioned a story made up of two layers. The first would be a short story-within-the-story, something symbolically resonant, that evolves and is retold multiple times as the frame story progresses. The second layer would be paratext surrounding the evolving story (marginalia, commentary, etc.), produced by a fictional character who is studying the story-within-the-story. 

At first, I considered creating a larger cast of characters who interact with the story, but now I’m thinking that it could be more interesting to explore through the point-of-view of one very long-lived character, who grows and evolves alongside the story itself. Perhaps this could be a human scholar whose body and mind are augmented, and who begins interacting with and using AI for reading and writing. Following the evolution of this single character could open up interesting questions around identity, memory, and the nature of consciousness. I wonder, for example: Could their consciousness get partly taken over by AI at one point, as they give over the very human processes of writing and reading to an AI? But then perhaps there’s a point at which the scholar must reconstruct a piece of the story. They might not know what’s real anymore, because of the way that algorithms and AI can make something that looks real (realer than real) but isn’t actually. Maybe they don’t even care what’s real, if they’re overwhelmed and awed by the sublimity of the AI and its processes. But faced with this task, the nature of memory becomes relevant: AI memory is time-bounded, a matter of storage space (e.g., 32,768 tokens for GPT-4), whereas human memory is much more fluid, constantly evolving, taking on new meanings given new contexts, and being spurred to recollection by experiences in the present moment. So it’s less factual than AI memory, but more contextualized. And so on, as we see the overarching story played out through both the evolutions of the story-within-the-story and the scholar’s paratextual contributions.

My sense of how I’ll incorporate computational and programmatic elements into my own creative process (and/or the reader experience of the piece) is still far vaguer. I would love to do so in a way that destabilizes common assumptions about machine creativity, human-AI relationships, etc., while also maintaining the emotional connection and arc that I believe is one of the most important and compelling elements of fiction. (I.e., I want to avoid doing a cool thing with AI that is solely a novelty, and that doesn’t add emotional/thematic resonance to the story.) 

I have seen some interesting reflections on using AI or programmatic tools to aid creativity in the ideation stage or even while writing, but personally I find myself less drawn to these approaches because it seems that they often take the tack of using AI to mimic or augment a process that the human writer is already doing or capable of doing. In contrast, I’d be interested in bringing in an entirely new or inhuman facet to the work, something that a human writer is incapable of accomplishing alone.

The question for me then becomes, “What can machines do well that human writers can’t?” This has yielded a few possible directions, including creating incredibly fast iterations on a text, introducing randomness, analyzing quantifiable differences or evolutions between texts (maybe the idea of literal/figurative palimpsests becomes relevant here), and moving toward a defined outcome in ways that break conventions that humans wouldn’t think to break. Additionally, as I hinted at earlier, AI and technologies like the algorithms of social media platforms are very good at creating things that resemble truth or resemble reality but that are not true or real. However, I still don’t feel that I’ve hit upon a compelling way of incorporating any of these ideas into my process.

I do think that the final output could be enriched if I present it in a digital, or even interactive, form — e.g., via a website that might include dynamically updating text (speaking to the themes of story evolution and instability), or where readers could navigate around the text in a nonlinear fashion (reflecting the often nonlinear process of engaging with and reconstructing stories with many versions). It might even be interesting to include some sort of insight into my own creative process, like different versions of the text or key places where a computational process created a turning point. The specifics here would differ depending on how exactly I end up incorporating AI or programming elements into the work.

Thank you for reading, and I look forward to your thoughts on any and all of this!

Group supervision #1 (week 5) / Anja Hendrikse Liu by is licensed under a