AI-Generated Sequels
This week’s Let’s Know Things (a news analysis podcast I produce each week) was on creative assets, and in particular how some such assets are being virtualized, replaced by AI knock-offs, or expanded upon with AI-generated media.
Many stories focused on “digitization meets creative field” have been published recently, but one that I find to be especially interesting (for a variety of reasons) is that of an author named Manjari Sharma who published a book on Wattpad and Inkitt—two indie publishing platforms that are especially popular in fanfic and romantasy circles—before being asked by the Inkitt folks if she would be interested in moving her work on their premium platform, Galatea, where she went on to earn a fair bit of money from it.
This is where it gets interesting: the Galatea people offered to help her publish a series, but only if she could have the second book done in just a few weeks, which wasn’t something she could make work with her schedule. They offered the alternative of having a ghost writer jot something for her—she would collect royalties as usual, but they’d handle the writing part—and that’s what she ended up doing.
They then further expounded upon her book’s footprint by turning it into a 49-video series (each a minute or two long), the production of which they also handled.
She didn’t learn until later that rather than having another human write that second book, they used AI tools to produce it. The same was true of the videos.
Readers didn’t care for the second book; it was apparently pretty bad compared to the first one. And it’s pretty obvious why that might be when you consider that the company has a skeleton crew of humans working with AI tools to handle the assets of something like 400 authors.
I find this fascinating in part because the business model is one we’re seeing everywhere right now: replicate an existing business, but replace all the expense people and assets with AI tools and AI-generated versions of the same. You can then theoretically earn as much as those other businesses, but your scale could conceivably be far grander, and your expenses will probably be a lot lower.
It’s not a bad idea, if you think of it from that angle. Or from the perspective of folks who just want more content, even if it’s not all going to be of an award-winning quality. The tools and the skill of the people using them will likely improve over time, so the outputs will probably get better. This could maybe someday lead to content of all sorts, all media types, that’s too cheap to meter, but which is good enough to replace some of today’s lower-tier (human-made) content.
The downside is that these tools, today at least, are being trained on actual authors’ (and other creatives’) work—which is potentially a copyright violation, but even failing that has some serious moral and practical issues attached to it (like how will all of today’s creative people feed themselves if their work is stolen and value-diminished in this way, and where will the next generation of AI get their training data if there are fewer working creatives?).
It does feel like this is the tip of the spearhead of something that’s going to be big, though. The optimist in me wants to believe there’s a way to leverage these tools and this rough business model to better empower writers and other creative people, rather than all the gains going to the tech folk who are building and profiting from these tools (which is predominantly the case, today).
What if, as an author, you could write worlds and characters instead of books? What if you could flesh them out and then click a button, and an AI tool, working from that body of creative worldbuilding you own and control, would spit out endless stories based on that universe and those characters?
What if you could do something like Galatea is doing and train an AI model (that you own) on your corpus of work and then have it produce ideas for followup stories, spin-off videos, and maybe even video game concepts for you? What if another creative, a visual artist, for instance, could then take an AI they own and control (trained on their work) and your respective AI systems could collaborate on building a game that you would collectively own—using your characters and worlds, and their visual style?
These aren’t fully fleshed-out ideas, but I feel like there’s got to be a way to use these things without them representing the end of paid creative output, and without stomping on the rights and intellectual property of others, in the process.
It may be that these tools never improve to the point that their output is worth much, and AI content may thus forever be a matter of quantity over quality: a lot of stuff churned out cheaply, but none of it terribly interesting or valuable.
It may also be, though, that someone combines the best of the existing creative industries (which are laden with their own contemporary flaws) with the best of these new, burgeoning technologies to come up with something even better, and which serves everyone involved.