For an elective pottery class at a large university, the instructor told his students at the beginning of the semester that they could choose to be graded on the quality of their final piece, or the quantity of pieces they produced. But they had to decide right then, in the first week of the class, and couldn’t change their decision after.
About half of the students chose to be graded on the quality of their final piece. They were confident that they could produce a high quality piece if they worked on it hard enough, and could probably save some time too. The other half chose quantity over quality: they figured that was the safest course, since even if they never learned how to excel at pottery, at least they would still get an A for putting in their hours.
Those who were graded on quality focused all of their time and energy on producing their final project. Those who were graded on quantity churned out pieces as quickly as they could, without focusing too much on any one piece.
At the end of the semester, a funny thing happened: the final projects of those students who chose quality over quantity were actually inferior to the last few pieces of the students who had chosen quantity over quality. By practicing on multiple pieces and putting in the tedious hours that the other students had hoped to avoid, they had paradoxically learned how to produce higher quality work—even though that wasn’t what they had been focused on.
I’ve been experimenting a lot with AI-assisted writing recently, mostly with ChatGPT, but I just started using Sudowrite as well. Contrary to popular belief, these are difficult tools to master, with a learning curve that starts off fairly shallow but ramps up to be dauntingly steep fairly quickly. This summer, I plan to write a novel with these tools, but because I have much to learn—and literally no one to teach me, since the technology is so new that we are all pioneers—I have decided to start with short stories instead.
My goal for this week is to produce ten AI-assisted short stories, at least up to the rough draft stage. So far, I’ve produced four, which is actually more than I expected to have at this point. Yes, they’re rough—each of them will have to pass through what I call the “human filter” before they’re ready to put out into the world: the choppier the AI version, the harder the work in the human filter stage—but they are still recognizeable as stories.
But just like the kids in the pottery class who chose quantity over quality, I’m not going to try to make any of them perfect—at least, not right now. My goal right now is just to get them done. Later, I’ll do my best to perfect them.
Along the way, I’m keeping notes on things that work and things that don’t. It’s getting to be a pretty big list, and I’m sure it will get much bigger as I go. At some point, I may share it as a blog post.
I’ll also be experimenting with different story generation methods, such as:
- Starting with ChatGPT from scratch and moving to Sudowrite for refinement,
- Starting from scratch with Sudowrite,
- Starting with the Mythulu cards and using those as inspiration for AI guidance,
- Writing a “seed” of a couple hundred words or so and seeing where the AI models take it, or
- “Seeding” both the beginning and ending, and using the AI models to fill in the middle.
I’ll probably come up with some other combinations, but those are pretty good to start with.
When the whole thing is over, and all of these stories have been passed through the “human filter” and polished enough that I feel that they’re ready to send out into the world, I’ll probably put them all together into a short story collection and release that directly, rather than releasing them individually as short story singles. I may do that later, but with how quickly I can put out these stories (and how few of the traditional markets are willing to publish them), there’s really no reason to wait.
Anyhow, that’s my thinking at the moment. But I may be getting ahead of myself, since I’ve only written four stories so far. Ask me where things are on Saturday, after I’ve written all ten.
