My new short story was commissioned by Google and produced using a new AI-powered editor called Wordcraft. Titled Author’s Note, it is presented alongside new stories from several other very impressive writers.
I’ll say more about this project below, but first, I want to present the complete list of stories I’ve published in the past year:
Elyse Flayme and the Final Flood, for MIT Technology Review
Harriet Amber in the Conan Arcade, for Commercial Type
Author’s Note, for Google
There’s also the new editions of Mr. Penumbra’s 24-Hour Bookstore and Sourdough, both with their prequel stories. All in all, a pretty solid set of offerings!
While I’m mentioning those new editions, the ones with bonus material, I want to apologize for being so parochial in their promotion. “Hooray, Sourdough’s prequel story is now available!” announces Sloan —
The U.S. editions, and my relationship with my U.S. publisher, are so front-and-center for me that it’s easy to forget about the global picture; but I really shouldn’t, and, in the future, I won’t.
Onward, to my …
Notes on writing with Google’s AI
I’m going to dive into some detail here; it’s probably not for everyone, so if this subject doesn’t interest you, feel free to skip merrily ahead.
There is a ton of context about the overall Wordcraft project on the website; I really encourage anyone interested in the intersection of AI and art to spend some time exploring. These technologies are now mature enough that we can begin to move from speculating about their uses to … actually using them! And not only to make cool demos, but to do real work.
That last part, the “real work”, is important to me. I am done, personally, with the genre of “I see what you did there”—of making things that are noteworthy primarily for their application of a frothy new technology. That’s really saying something, because I have made a LOT of things like that! I spent several years engaged in AI writing explorations of my own, almost entirely in this spirit.
Approaching this project, I felt committed to writing a story that could stand on its own; a story that achieved the same things I want ANY of my stories to achieve; a story to which the response might be not, “I see what you did there”, but: “I loved this!”
Whether I succeeded, you can judge for yourself. For my part, I’d happily have published this story on my website, or distributed it as a zine, with no mention of AI tools at all.
Okay, but what about those AI tools?
Over on the website, you can see some screenshots of the Wordcraft editor and read more about its affordances. It operates basically like a super-upgraded version of my AI text completion plugin from way back in 2016. Everything about it is much more capable, but the workflow is the same. You can, at any point, ask the AI to generate new text, and when it does, its suggestions reflect the surrounding draft.
The Wordcraft team didn’t require or even suggest that my story be written in any particular way, or that it contain any particular fraction of AI-generated text. My brief was: write what you want to write, the way you want to write it.
Okay, then! After noodling for a while, I decided I’d try to write some kind of locked-room mystery, because (a) it’s a genre I enjoy, that (b) I’ve never tried before, and (c) I liked the idea of giving the AI the role of the detective —
It didn’t work at all. Absolute crash and burn. I had overestimated the model’s powers. Paragraphs into my nascent mystery, it had no idea what was happening; or at least, not enough of an idea to “participate” in any meaningful way. It knew I was writing a mystery; it reached for the appropriate terms and tropes; but that wasn’t anywhere close to being sufficient.
So I retreated, a bit dejected. I kept a log of my experiences, for the benefit of the Wordcraft team. A few notes from that log made their way into the team’s paper, which I think is a quietly pathbreaking contribution: an almost ethnographic study of real writers doing real work with this real AI tool. Hype, this ain’t! The Wordcraft team has demonstrated clear-eyed courage in synthesizing and presenting our experiences, many of which contain some version of my own retreat and dejection.
For a couple of weeks, I thought about other genres. On a medium-long drive, an opening line occurred to me. Beside a pool in Fresno, a story took shape. Then, I cracked it, by establishing a fenced-in space for the AI to contribute what it could.
This AI, like many of its predecessors, is very good at “riffing”: given an example, it can generate lots more, often with impressive fidelity to genre and vibe. This new story called for a litany of grisly deaths, so, I asked the AI to dream up grisly deaths. I asked it, also, to tell me the names of the dead: characters from a fantasy saga, cliched in the best possible way. Red meat! This assignment, it could handle.
Now, when you read the story, you’ll know that the litany of deaths is mostly the work of the AI. I prompted and prodded; I curated and edited; I smoothed and re-arranged. But “Captain Cirrus, trampled by a hippopotamus”? That’s 21st-century technology in its fullest bloom bringing you that!
Three thoughts, to close.
First, I’m impressed as hell by the Wordcraft team. Daphne Ippolito, Ann Yuan, Andy Coenen, Sehmon Burnam, and their colleagues engineered an impressive, provocative writing tool, but/and, more importantly, they investigated its use with sensitivity and courage. This puts them leagues ahead of nearly everyone else out there peddling AI tools. (Indeed, I wish I’d been this circumspect back when I was yakking about my own AI writing explorations, several years ago.)
Second, I’m proud of my story! I love its George R. R. Martin-alike narrator; I love its villain and his scheme. And, yes, I love the deaths, conjured gleefully by a computer that’s read it all.
But, third: I have to report that the AI did not make a useful or pleasant writing partner. Even a state-of-the-art language model cannot presently “understand” what a fiction writer is trying to accomplish in an evolving draft. That’s not unreasonable; often, the writer doesn’t know exactly what they’re trying to accomplish! Often, they are writing to find out.
Will the AI get there, someday? Will I be able to recruit my detective at last? I understand very well the current pace of improvement in this field, and even so … I don’t see it coming anytime soon.
There’s more than a little meta-textuality in Author’s Note; it is surely no coincidence that the assistant, in the end, gets dunked in the pool.
I don’t think I’ve ever written about my theory of short stories, even though I talk about it a lot. It’s very simple: I think all good short stories are, in one way or another, about death. That’s not true of novels; good novels can be about lots of things. But a good short story is always basically a memento mori.
This is one of the reasons Bullet in the Brain is the best short story.
I won’t defend my theory —
Remember, I come by my interest in AI honestly; I’ve been thinking about this stuff since 1996. Here’s the title card for my “documentary” on AI, produced at Troy Athens High School using a VHS editing rig:
View the whole thing on YouTube, if you dare!
Disintegration to discoverability
A few notes and links —
Where Twitter’s disintegration is concerned, I stand by my previous assessment.
“It was terribly creative and demanding, because you’re making design decisions at every moment. So it was high design.” Who wouldn’t want to crawl underneath a table and work on this? Fabulous.
Here is some very nice banjo-playing by Rob Stenson, who I know mainly as the creator of Coldtype, a library that powers cool text animations.
Here is the moon, “melting” behind the pressure waves blasting out the end of a rocket. Here, the effect is winningly explained by Scott Manley.
Here’s a beautiful paean to a Studio Ghibli-adjacent movie I’d never heard about, from the great Animation Obsessive newsletter. They are really doing something unique and valuable over there.
The movie, Princess Arete, is available to watch for free in many regions until December 2. Here’s the clip that Animation Obsessive shared:
Guillermo del Toro, as ever an evangelist for imperfection:
“I wanted to return the controls of animation to the animators, and treat [the animators] as actors,” del Toro said. He encouraged his team to make “failed acts”—the tiny stumbles and imperfections that make a performance by a live actor feel real —
and resist what he called the “codification of animation into a ‘cool’ language that is almost like emojis.”
I thought I knew things about distribution
I am not really involved in the video game industry, but I’m nevertheless an avid reader of the GameDiscoverCo newsletter by Simon Carless, because I find his lens so inspiring.
Simon doesn’t write about programming games; he doesn’t write about playing them. He writes about the space in between. How does a person FIND a video game? What makes them decide to purchase and play it? What buttons, levers, and warp zones influence that process?
“Discoverability” is a good 21st-century word. Most of the digital dreams of my cohort, the partisans of the open web, ran aground on this reality. It’s not enough to make something and post it online; you must also inject it into some channel that will carry it to people. The web itself doesn’t do that; you need an extra layer, some reservoir of attention and/or curiosity, whether it’s Google, the blogosphere (RIP), StumbleUpon (RIP), Twitter (RIP) … hmm, there seem to be a lot of dead channels out here.
Back in the 2000s, I thought I knew things about distribution, about attention and networks —
It was, honestly, the experience of publishing a book with FSG that showed me what distribution really looks like, and taught me that you just cannot be starting from scratch every time. You need supply chains —
There is discoverability in books, though of course it’s totally different from discoverability in games. Different storefronts, different algorithms, different tastemakers. I’d like to think this newsletter contributes to book discoverability, in a small way, from time to time.
I’d say most people in both industries, even the professionals, muddle through discoverability, hoping for the best, sometimes getting it. There are new kinds of professionals elsewhere who won’t take any chances. The crass video producers who avidly game YouTube’s algorithm, surfing its subtlest contours to millions of views —
What I admire about Simon Carless is that he knows it, too, but he is never crass, not for a second. His raison d’être is the support of interesting, independent video games. He crunches the numbers and monitors the algorithms on behalf of, no kidding, art.
There really ought to be a discoverability newsletter of this quality for every medium, in every marketplace.
This is my season for driving, as I haul tons of olives from grove to mill for Fat Gold, the small olive oil company I help operate. I cover a lot of ground —
This year, I have observed
the great fulfillment centers alongside I-5 near Tracy, lit up at night like enormous alien depots, each of them a quarter-mile on each side, easy. This is the cold-pulsing heart of commerce for the whole Bay Area; it doesn’t look like much.
a gravel quarry with a wind turbine spinning overhead and silvery lights winking along the various chutes and conveyor belts, and the gravel itself piled in shadow below. An appealing industrial silhouette.
an orchard uprooted, the trees all laid down on their sides to dry for a year, now being fed into an enormous chipper. The process continued for several days, and each day, when I passed, there were fewer trees, until finally they were gone, replaced by an empty field dotted with a few towering heaps of bright wood chips. (I am 95% certain this field will now be planted with almond trees.)
the California Aqueduct, the artificial river that carries meltwater from the Sierra Nevada to Southern California. Driving around, I found myself crossing and re-crossing it all along its course. Each and every time, it impressed: easily 100 feet wide, formidable, inexhaustible (or so it appears), the water black-blue under the winter sun.
In the couple of days I waited to send this newsletter, the rain arrived. Clouds came scudding in from the west, wonderful scratchy tufts. Everywhere I drove, I could see a sheet of gray out one eye and a patch of blue out another; the rain never fell where there wasn’t sun shining at the same time. That’s rainbow weather, of course, and they floated all over, not full arcs but little scraps of refracted color pinned to the air. On Monday, I experienced one of my nicest California days ever in my eighteen years living here, as I trundled along, water above and water below, goggling at the show.
From the San Joaquin Valley,
P.S. You’ll receive my next newsletter on December 7. Before then, you’ll receive my annual gift guide, sometime in late November.