Firehose quantities

Generative AI has the potential to lock us into an ecosystem of empty stories

Rail City Fan Fest!

Quick thing before I get into the meat of today's post: tomorrow's my birthday, and I'll be appearing at Rail City Fan Fest to talk about the history of cosplay, and to hang out for a bit with the 501st. I'm not sure if I'll be in costume or not (I'll likely be dragging my child along with me), but it should be a fun time. I'll be talking at 11AM, and I imagine that it'll be a somewhat informal thing. If you're in the area, stop by!

An endless stream of noise

There's been a lot that's roiled the ranks of SF/F's Extremely Online personalities in recent weeks. The launch of generative artificial intelligence platforms like Dall-E, Midjourney, and ChatGPT have sparked a considerable amount of controversy: Clarkesworld Magazine getting inundated with AI-generated stories, while the Self Published Fantasy Blog-Off shutting down its category for cover art after this year's winner was found to have used AI. This is something that the SF/F publishing world is going to have to contend with, and SFWA has put out a statement regarding its use.

On the other side of the Twitter streams is a loud, ongoing argument about the nature of "cozy horror", in part because of a piece that went up on The Mary Sue. It's not a great take: I find some of its conclusions about genre tendencies and horror to be a bit overly broad, but some segments of the SFF world have been using it to gin up some petulant arguments and online discourse that only reaffirms my belief that social media (and especially Twitter) was a bad idea.

On the face of it, these two topics feel like they're pretty far apart, but there's a kernel of an idea that underpins both is the relationship between art and content, and the uneasy boundary that the genre's had to navigate since its inception.

A number of years ago, I wrote up a short piece for io9 about an anthology from Microsoft called Future Visions: Original Science Fiction Inspired by Microsoft, featuring the likes of Elizabeth Bear, Greg Bear, David Brin, Nancy Kress, Ann Leckie, Jack McDevitt, Seanan McGuire and Robert J. Sawyer. There was a comment that's stuck with me ever since: "I actually like a lot of these writers, and so am dismayed to see them attach their names to this kind of advertising. Everything about this damages their own brands."

The underlying assumption – that these stories are in some way diminished because they were commissioned by a large tech firm – bothered me, because that's been a factor in the history of science fiction and popular culture since the early 20th century. In some ways, vast parts of the body of work that is the science fiction genre is a loss-leader for advertisements: science fiction magazines (then and now) feature plenty of advertisements, while the very structure of some of the genre's best-known and beloved works are shaped around commercial constraints. In one way, the genre was created as a stream of content, supported by the need to connect advertisers to readers. For much of its existence, science fiction has been a commercial enterprise, aided by a small and devoted population of fans who consume it. When I helped pitch and edit a the Better Worlds anthology for The Verge, corporate sponsorship was a key element of that: the partnership folks at Vox secured Boeing as a sponsor.

We don't look at science fiction of this era as mere content; a lure to ensnare hapless readers into spending their money. Magazines facilitated an environment for artists to carve out stories that expanded the boundaries of imagination. Successful and would-be authors and artists worked within those commercial constraints to generate new worlds, new characters, and new ideas for their readers.

Bill Watterson, the creator of Calvin & Hobbes is an excellent example of how you can carve out an artistic space within a commercial environment. (KaptainKristian has an excellent video about this.) Indeed, there are plenty of notable stories that have come forth from the minds of writers in the pages of magazines and on television screens, brought there by the needs for an advertising network. We're reminded of this every time a beloved series is given the ax by the cruel network executives who point to low viewership rates to justify the cost for keeping it going for one wished for season.

We remember the stories those writers produced; we don't remember the ad copy.

At some point in the last couple of decades, it feels as though there's been some shift in how we consume our entertainment. Nerd things aren't the backwater things they once were: Marvel Comics and Dungeons & Dragons are cool now, and there's an audience that's ready to snap up anything with one of the big brand names on it. There's always been a level of commodification with storytelling: look at the history of the comic artists who created the worlds that Marvel and Disney are spinning out in theaters and on computer screens, but it feels like it's becoming overwhelming, at least at the widest public levels. We're fed stories that exist only because they were an entry once placed on a franchise roadmap a decade ago, created to appease the unending hunger for CGI action scenes and to catch the weak gusts of social media trends. Capitalism is a hungry beast, and it doesn't have the patience to be fed what we need to grow.

There are rare instances where those artists are able to carve out an unexpected morsel. Tony Gilroy's Andor was an abrupt shock to the system; a story that had legs and a powerful message wrapped up in the trappings of a Star Wars adventure, while Across the Spider-verse bursts with energy and vibrancy as it spins out a story about belonging.

There are certainly plenty of other films that find large audiences outside of the franchise names: The Daniels's Everything Everywhere All At Once was a zany, brilliant take on multiverses, Syfy's (and later Amazon's) The Expanse looked at inequality in a frightening interplanetary future for humanity, Christopher Nolan's Interstellar introduced a whole bunch of people to the idea of relativity, but these sometimes feel like they're just the exception, rather than the rule. Of course, the literary scene of short stories and novels is bursting with imagination, but those stories rarely break into the cultural zeitgeist and mainstream world.

There are tendencies in this modern fandom mindset that trouble me: there's a hollow emptiness that comes with stories that exist only to fill a schedule or fill some committee-generated set of criteria: what if character x meets/fights/teams up/encounters character y (who you might remember from Story B, with a cameo in Story F.)

Studios plan much of this, assuming that fans will want more and more and more of the same: more crossovers, more cameos, more tie-ins that turn into homework. There are plenty of people who'll happily take it in: there are entire ecosystems of outlets and personalities that breathlessly report on every morsel of an update for the next films and television shows on the horizon. We argue about what we should expect from films that might never see the light of day, of casting choices that were plucked from thin air from a disreputable channel, and we scream at authors who haven't delivered the next installment of a series yet. We want more and more content but it's never, ever enough.

With the rise of AI generative platforms, there's been a trend going around: imagine %InsertBrandNameHere but in the style of %InsertMovieDirector. Harry Potter in the style of Wes Anderson, Star Wars in the style of Stanley Kubrick or Andrei Tarkovsky.

Tech companies and personalities in Silicon Valley have pointed to generative artificial intelligence as a suite of tools that can unlock anyone's artistic potential to bring their imagination to life. I can't say that I disagree with the rationale on paper: creators have always figured out ways to tease out stories as new technologies present themselves. Typewriters, personal computers, mimeographs, the internet, Photoshop; in the right hands, they're tools to help create, distribute, and transform. Yudhanjaya Wijeratne utilized a homebrewed AI system to generate some poetry and planetary environments for his novel The Salvage Crew, while Ken Liu used AI trained on his body of work to write "50 Things Every AI Working with Humans Should Know" for Uncanny Magazine.

There's ethical concerns about how these platforms have been trained: artists have found their watermarks embedded in the art, and without good receipts from the companies behind these platforms, it's impossible to tell what the mix of ingredients are that end up the works they generate. I don't know that the wider public really cares about these sorts of concerns, and a post from a writer called Bloonface (via Daring Fireball) has some useful thoughts that parallel my own when it comes to how people use these tools: people will use the tools that are most useful or convenient.

There's a potent mix of conditions at play here: a fandom that has an insatiable hunger for content and the ready access to technologies designed to produce just that. I worry even more about how those early adoptions will be perceived as the future by the companies that produce our entertainment.

Already, we're seeing people making these arguments: an author making "thousands" by producing dozens of short stories generated with AI, while another went viral for generating and releasing a children's book on Amazon. And while writing this, Marvel released its latest Disney+ series, Secret Invasion, complete with opening credits made with AI. (I actually dig the look, given the shape-shifting nature of the series.)

Technology gives us the tools to ease the path of production, and in a heavily industrialized creative field like the film industry, I can't help but see a future where the bar is raised just out of reach for creators. I do believe that there'll always be demand for human creativity, but if AI is so ingrained in the creative process, those opportunities afforded to creators will become scarce. We don't even need generative systems to do this: independent authors like Craig Alanson have neatly adapted to the demands that platforms like Amazon's Kindle Direct Publishing program puts on its readers: a marathon of turning out content as quickly as possible to appease the demands of readers who just want the next installment as quickly as possible. No shade to Craig – I enjoyed his book and it's a process that works for him. But it certainly doesn't work for everyone, and a future of just that type of storytelling feels bleak to me.

I worry that something the environment that the KDP program foreshadows the type of creative environments that we'll have in the future: that a technology like ChatGPT will supercharge these tendencies, spitting out a stream of endless shadows of stories that don't do much beyond taking up space.

There's a contradiction in all of this: studios and platforms are arguably giving people what they want, and there's a long and distinguished tradition of authors working to formula to put down as much ink to paper in order to make a living. The pulp era is loaded with these examples: authors toiling away as they produce the next installment of the Hardy Boys, Nancy Drew, Tom Swift, or Doc Savage. The books that sold the best when I worked at Walden Books? Piles of romance mass market paperbacks, as well as books from big names like James Patterson and others.

I don't think that there's anything wrong with books that are comfortable to read; long-running stories and characters that you can sink into for volume after volume, episode after episode, because you're there for the journey or because of the familiar surroundings. Genre literature is loaded with works that play to these tropes: urban fantasies featuring an supernatural anti-hero protagonist saving their city from evil, the commander of a starship engaged in battle after battle, to say nothing of the flood of comics that feature superheroes who've been fending off villains for book after book.

The latest grumblings over genres like "cozy horror" might have gained some steam because of a dashed-off article, but at its core, I see complaints from the same circles of writers and readers who believe that every sentence must be beautifully crafted, that each book should break new ground and confront a genre's worth of tropes in a new way in order to have some of legitimacy with their audience. These are the same arguments that have been leveled against genre authors for decades: that their fantastical subject strip them of any artistic value. It's condescension that I've seen in plenty of indie bookstores, literary festivals, and folks in non-genre spaces.

Where do you draw that line between empty storytelling and writing a story crafted to audience demands and expectations? Is it the existence of a biological heartbeat, the result of electrons capturing one's experiences trapped in the wet mush contained in a skull? Is it the human touches that a creator brings to an outline and world that was created by committee, or the spark of random inspiration?

I recently finished a couple of books that neatly fall into the "cozy" categories: John Scalzi's Starter Villain, and Alex Shvartsman's The Middling Affliction. They're what I'd describe as aggressively fine. They were fun: I wasn't bored by their stories or characters, but I'm not sure that they're books that I'd return to in any meaningful way, where I might glean additional insights into their worlds, characters or prose like I recently did with Ann Leckie's Ancillary Justice and Iain M. Banks' Consider Phlebas. There were things that I'd have wanted to see turn out differently, and things that I can see were put in there because it was plain to see that they were having fun with the process.

Are they the types of stories that could be generated with AI? Probably not: both authors have decades of experience writing and editing, and in both cases, they've put together a story that pulls in their readers for the journey. But central to those stories is intention: they had a fun idea that they've put pen to paper (key to word processor?) that they were able to carry out, making up the story beat by beat.

This, I think, is key to figuring out where that line lies: the craft that goes into the process of creating the story, whether it's spending the time to craft elegant prose and plots or if someone uses their experience and knowledge to construct a compelling, readable story.

I don't disagree that this sort of technology can unlock someone's creative potential, but that explanation just reads to me that it's aimed at someone who doesn't care to take the time to understand what goes into a good story, world, or characters. They don't care about the craft, the thought, the struggle, or the effort to translate their imagination into words on a page. When we pick up a book, we understand that there's more beneath the surface: time that someone has spent learning how to tell stories, that their story draws from their human experiences.

AI in the hands of a skilled storyteller could be a useful tool that can make writing easier or efficient. I don't think it's the best way to go about writing or coming up with a story, but I do believe that there could be ways in which these types of systems can be used, much like we use Photoshop or a word processor. But people being people, I think it's far more likely that it'll be used as a means to churn out an firehose of content from people only interested in quantifying their stories by the number of installments and volume and efficiency.

I hope that readers and fans will be able to tell the difference, and let the market run its course.

Currently reading

Since my last update, I've finished only one of the books on my to-read list: Ancillary Justice, which I hadn't read in a decade. It holds up really well, and it was in preparation for another book on my to-read list, Translation State, which I've just started.

I've since finished three others that were on my longer TBR pile:

The Bright Ages: A New History of Medieval Europe by Matthew Gabriele and David M. Perry. I've had this on my to-read list for a while, and downloaded the audiobook for a recent trip down to Pennsylvania. It's an interesting look at the state of Medieval Europe, positing the thesis that it wasn't a dark age, but one where there were complicated politics, art, cross-cultural connection, and innovation. I enjoyed this one quite a bit.

The Rise and Reign of the Mammals: A New History, from the Shadow of the Dinosaurs to Us by Steve Brusatte. I loved Brusatte's The Rise and Fall of the Dinosaurs: A New History of Their Lost World when I read it a couple of years ago, and wasn't disappointed by this one. It's a sweeping, epic history that helps explain how (and why) mammals thrived after the demise of the dinosaurs, and specifically, how we can track evolutionary changes based on the fossil record.

The Middling Affliction by Alex Shvartsman. I recently picked up the audiobook for this when it was on sale, and have been listening on the drive to and from work. It's a fun read, one that falls into the sort of comfortable content bucket that I was talking about earlier. As noted, it's fine: a fun world with some entertaining characters, with a plot that I'd have like to have seen tightened up a bit more.

Other books on the to-read list:

The Icarus Plot by Timothy Zahn. Still plugging away on this one, but am enjoying it.
White Sun War by (General, retired) Mick Ryan. I need to get back to this one.

I had started Cory Doctorow's Red Team Blues, but I've set that aside: too smarmy, "as you know, Bob", and too much crypto, and I couldn't get through the first couple of chapters.