Battling boredom, Dave Jilk wrote an epic poem that almost defies description
SunLit: "Epoch" is not the kind of book you see every day. How do you describe it in terms of storyline and genre?
Dave Jilk: The storyline is essentially science fiction, though it is slightly unusual in being told from a first-person point of view by a character who is not a human. It's "hard" science fiction, perhaps more so than is typical, in that its scientific elements aim to elaborate the likely rather than imagine the possible.
Beyond that core plot, though, I've sometimes described it as "trans-genre": it is an epic poem (an extended narrative poem with certain traditional plot elements); it explores philosophical themes both ancient and modern; it includes considerable scientific and technological detail; and it has some characteristics of a postmodern novel, with a variety of formal modes and styles and the extensive references. Some might see it as a mess, a yard sale, but there is intention behind all of the choices.
SunLit: This isn't your first venture into poetry. Tell us about your relationship with that literary form, and particularly your decision to combine an epic poem with science fiction. What influences informed the project?
Jilk: After semi-retiring from the business world, I wanted to find a way to express myself creatively. I've always been a solid writer, and I had written a few poems here and there. But the clincher came when I told a friend that I was a little bored with things, that it felt like new experiences weren't doing much for me.
Knowing me well, he suggested that I was too focused on abstract ideas and should try to experience things more concretely. Well, like most art forms, poetry is about the concrete. It is about the actual words: how they sound, how they are spelled, what they look like on the page. It is about exempla: yellow pollen dust on blue petals, a gas-gargling prop plane flying toward the foothills, the poke and pinpoint chill of an errant raindrop on your face. I started exploring it, and I was hooked – it was a new way to be present, to experience and enjoy life.
This was true of reading poetry but even more when writing it. After writing some lyric poems and self-publishing a couple of collections, I had a desire to try my hand at a longer narrative poem. I didn't have to think too long or hard to realize that artificial intelligence was the obvious subject matter.
As to influences in general, there are too many to name, and the extensive references illustrate this. For epic poem influences, I had previously read only "Beowulf," Shakespeare's "The Rape of Lucrece," and some Alexander Pope. But I read several epic poems while I was writing the book, including Derek Walcott's "Omeros," Homer's "The Iliad," and Milton's "Paradise Lost," and the influences from those are fairly apparent.
Unfortunately, I did not discover Frederick Turner until fairly late in the project. He has written three science fiction epic poems, and literally "wrote the book" on epics ("Epic: Form, Content, and History"). I have just finished reading his epic "Apocalypse" (2017) and it has a fascinating overlap with several of the themes I cover.
One of the thematic points "Epoch" attempts to make is that our intellectual and cultural influences are frequently implicit and unconscious (at one point I call these "cultural canon-balls", get it?), traveling to us through multiple layers of adaptation and interpretation by others and unintentional integration by our own minds. The point here is that, while Turner's epics did not influence me directly, it's likely that either they did so indirectly or that there were other aspects of the modern milieu in which we both shared.
SunLit: You've written several academic papers on cognitive neuroscience and concerns about AI. What inspired you to address those topics in a completely different way?
Jilk: As I mentioned, the book started out as a poetry project, not an AI project. But as I thought about my plan for the work, I realized that the epic poem format had some advantages over more typical modes of discussing these topics. My views on the big questions about artificial intelligence, such as how it is likely to be built, what is likely to happen after it is built, what we can do to improve our probable outcomes, and how we might assess those outcomes, are not at all mainstream.
So writing something that looks completely different from an academic paper or a series of posts on LessWrong made sense. But further, among those who think about AI options and outcomes, there is a tendency to try to seek certainty, to try to "game it all out." Nick Bostrom's "Superintelligence" has that feel, as does a subset of other papers and books in the field. By expressing my views as poetry, I am making a statement about the uncertain nature of such predictions and the folly of attempting to find the one true plan.
SunLit: Artificial intelligence obviously has arrived front and center as one of the more compelling issues that we're dealing with worldwide. What did you want to say about the controversy – and do you consider an AI "takeover" not only possible, but likely? Or is that a mechanism for addressing some of the other themes in the book?
Jilk: It's worth noting that I started working on this book at the end of 2019 – prior to the revolution in large language models like ChatGPT. At that time, only a relatively small number of people were really thinking about the possibility of "superintelligence" or "superhuman AI" and what that meant, although the notion had been pointed out decades ago.
Even now, there is considerable confusion between worries about current AI, like privacy, scams, deepfakes, and labor displacement, in contrast to superintelligence and its potential to displace homo sapiens as the dominant form of intelligent agency on Earth. The former can and will be managed just like most technologies have been managed. The latter can't be managed at all, which is why people talk about "existential threats" and "p(doom)" (the probability of humanity's complete destruction by AI).
Assuming that human cognition and agency are fundamentally a kind of computation, and that we don't destroy ourselves first, it seems to me rather likely that eventually humans will build a system that performs all the same cognitive and agentic computations as humans. Once that happens, the "AI takeover" is probably not far off. Even if such systems are no "smarter" than us, they will have capacities that we cannot compete with, like backup and restore, or traveling over wires and radio waves. You'll have to read the book to see my full answer to that question, including how it might play out.
What I found fascinating as I wrote the book was that this novel concern – the possibility that humans will create a fully human-level artificial intelligence and the consequent existential risks – really puts a spotlight on some of the deepest questions of philosophy. So I go there, and explore issues such as identity, purpose, future selves, transcendence, the importance of beauty, the nature and ethics of technological progress, and humanity's tragic flaws.
SunLit: The book has 80 pages of references related to everything from allusions to literature, music and art to scientific papers. Why did you want to include this background?
Jilk: It started as merely keeping track of my sources, so that I could give credit where appropriate. As I dug into the book's subject matter, I read a lot of material to support the ideas and the writing, and the number of sources started growing rapidly. I realized that they also added some depth – if you were to read one of the poems or watch one of the movies, you could appreciate the book at another level.
Then the whole thing took on a life of its own after a friend who was an early reader said she interpreted it as being like a machine learning system that had been trained on many different books (her comment pre-dated Large Language Models). That hadn't occurred to me before. The very last paragraph of my Acknowledgements points out that inclusion of the references also is a kind of tribute to the shoulders on which I stand, our human culture and intellectual history. Alternatively, David Foster Wallace has been criticized by some as "showing off" with his extensive footnotes, so maybe I'm doing some of that, too.
Several early readers have told me that they found these "osculations" (as I call the references) really enjoyable. I should mention that on the book website (epoch.fyi), there is a PDF file that is formatted to be easily readable on a phone, so that readers who want to follow the references as they read don't have to flip back and forth in the book. The site also has a playlist of some of the songs.
SunLit: You've said before that this book "is not recommended for casual consumption." Who is your target audience and what do you hope for them to take away from the experience?
Jilk: I wrote the book for myself, without a strong sense of a target audience. Consequently, by publishing it I am hoping to learn just who its audience is. I think a reader will have to be willing to put a little more effort than usual into the reading – to look up unfamiliar words, to read the poetry at a slower pace, to access the references or Wikipedia to understand more of the technical and philosophical discussion.
But it is not at all impenetrable, and for the most part the poetry is not cryptic, as lyric poetry frequently is. Among readers who are open to that effort, any who enjoy science fiction generally and are interested in stories about artificial intelligence in particular should enjoy the book. Readers who like big ideas to infiltrate the plot will appreciate what I've done. Someone who regularly reads poetry but has limited familiarity with the existential issues surrounding AI should find the book eye-opening while enjoying the poetic elements.
That probably doesn't represent a large audience, but it's a worthy one. If anyone reading this belongs to a book club that enjoys more challenging books, I'll offer right here to join a discussion session about the book via video call. They can contact me through the book website.
SunLit: What were the biggest challenges to taking on a project like this that has, with the multiple writing forms, the science and the narrative, so many moving parts?
Jilk: The biggest challenge was what in software development (and perhaps other fields) is called "scope creep." Originally, I thought this story might be 80 or 100 pages. It ended up at three times that, because as I started writing each chapter I'd realize there were all sorts of things beyond the bare plot outline that ought to be said. I had to fight these impulses to some extent, but it was also self-correcting. I'd reach a point of exhaustion with a given chapter, and find myself thinking more about the next one. That was a sign that I needed to stop adding new content.
Before this project I had not written any long narrative fiction. I found filling in the details of the plot challenging: You have to make it up, but also it has to be plausible, entertaining, and add something to the broader themes.
The overarching plot was already coarsely outlined in my mind when I got started, and much of the science and technology – software engineering, artificial intelligence, neuroscience – is familiar terrain for me. As should be apparent from the book, I see form as just another literary/poetic/rhetorical device that helps to convey the specific content. So I'm comfortable writing prose, poetic forms, and free verse.
SunLit: How did you approach such a huge undertaking in terms of the writing? Did you outline the project or let the writing take you where it wanted to go?
Jilk: I did write down a rough outline of the plot, and revised it a few times. In particular, at some point I started taking the "epic" element more seriously, and researched the elements and characteristics of historical epics. This necessitated some changes.
I mostly wrote the book in order. There are a few "flash-forward" chapters in the first half of the book, and I wrote those all at once and then inserted them into the flow of chapters. After I was about half done with the book, I decided I needed to go back and add a little more character tension, and that turned into the dream sequence in the chapter "Anxiety."
Within each chapter, I had to let the writing take me where it wanted to go, subject to the overall storyline. I used the philosophical themes, the references (which I had in many cases attached to specific chapters before writing them), and the poetry to drive that creative process.
SunLit: How did you choose the cover art?
Jilk: I knew that I wanted something different, something that brought out the emotional content of the book and pointed to its larger themes, something that said that this is a literary work first and foremost. In other words, not a scary robot face.
Then, during the writing process I discovered Bob Thompson's work, and on a whim I contacted Michael Rosenfeld Gallery, which represents those works. "Black Monster" in particular appealed to me, as it had both obvious and more subtle connections to the book, and it's a beautiful and colorful painting. The gallery generously granted me permission to use the painting on the cover as long as the image was not modified. It's worth noting that the Denver Art Museum has at least one Bob Thompson in its collection. Then, in the process of cover design I realized we needed something on the back. A robot was the obvious thing, as a contrast to the front cover and a signal that this is, in fact, science fiction. But it's not necessarily scary, and its evocation of da Vinci's "Vitruvian Man" indicates that the story is not just another attack of the killer robots.
SunLit: Tell us about your next project.
Jilk: Right now, interleaved with marketing this book and creating some peripheral content (such as a hyperlinked version of the references), I'm trying to just enjoy reading and writing lyric poetry again. This book took more than four years to write and about six months to get it into production, so I probably need a break from big projects.
That said, I have a few ideas I'm pondering. These are all over the map: fiction and nonfiction, prose and poetry. Or maybe I'll just write lyric poems and collect them again after a while.
A few more quick questions
SunLit: Which do you enjoy more as you work on a book – writing or editing?
Jilk: I see it all as writing. Editing is just re-reading longer sections to identify what I don't like, at different scales, and then re-writing it. And I enjoy it all.
SunLit: What three writers, from any era, would you invite over for a great discussion about literature and writing?
Jilk: I mostly read dead people, so I'd say Alexander Pope, Edith Wharton, and Friedrich Nietzsche. There would be many laughs along with the insights!
SunLit: Do you have a favorite quote about writing?
Jilk: Mary Ruefle: "I suppose, as a poet, among my fears can be counted the deep-seated uneasiness that one day it will be revealed that I consecrated my life to an imbecility."
SunLit: Soundtrack or silence? What's the audio background that helps you write?
Jilk: I almost always need quiet when I write. Infrequently I will play some music, but it has to be instrumental-only, since lyrics disrupt my phonological loop.
SunLit: What music do you listen to for sheer enjoyment?
Jilk: I always feel good when I listen to Bebop jazz, late 1950s and early '60s: Miles Davis, Art Pepper, Bill Evans, Gerry Mulligan, Dave Brubeck.
