Book read: Zombie 69 by Kitty Glitter
Pages: 20
Sometimes you read a book and you wonder… what did I just read?
That is how I felt after just a few pages of Zombie 69. This short twenty-page book contains two short stories, the first about zombies and the second about a cat and dog that live with some humans. Neither story makes any sort of sense. They both feel meandering and broken, almost as if it was written by a primitive AI generator, not a person.
I don’t remember the zombie story very well. It had to do with zombies going to high school (why?) and doing ordinary teen things, they are just zombies that have to pop their eye back into socket now and then.
The second story was… weird. The cat and dog can talk to each other, and they swear a lot. Some random person breaks into their house and kidnaps their child, then the dog goes a little savage and blames the cat. The cat, for its part, starts talking about time shifts, and the girl being pushed out of time. Then it just ends. The girl is still gone, there is no explanation of time shifts, there is no explanation of why the cat is able to drive an RC car, or who that guy that kidnapped the girl was. It just ends.
This clipped, piecemeal storytelling gave both stories a generated feel. It felt almost like it was using some sort of madlib format. (I am not saying it is, of course, I can’t know that. But the feeling is there.)
This chaotic structure is often how people recognized early AI generated art. The AI used to combine aspects of different pictures, and even blend them together, but it often got anatomy or structures wrong. It added extra fingers, eyes were mismatched, or legs didn’t make sense. It became incredibly obvious that whatever created the artwork, be it human or program, had an uncanny valley feel. It seemed like it should be art, or human, but it just felt… wrong. And sometimes you couldn’t figure out what made it uncanny, you just knew there was something that didn’t work.
In written work the early AIs were much the same, pulling from different areas of the internet, or from books, and blending them together. It was easy to tell what was AI generated back in the day with lots of emdashes (who knows how to make an emdash?) or context drift in the story. When dealing with factual information, early AI was still trying to figure out what was true and what was sarcastic, parody, or flat out lies.
Remember, a predictive text formula is only as good as the information being fed into it, and a LOT of the information online is just wrong. How could a language model be expected to be right all of the time?
I don’t know if this particular story was written by an early algorithm, or just some random stories built from the author’s wildest dreams, but the feeling is the same. It doesn’t feel… cohesive, or right. And it makes zero sense.
How would you fix the very structure of your storytelling? My suggestion is to have beta readers, or a writers workshop. Being part of a writers workshop and having honest feedback about my writing helped me get the words right faster than just spitting words into a void and hoping they made sense. And with the age of the internet with Facebook, forums, meet-ups and more, finding a group of people dedicated to helping each other get better at writing is easier than ever. It can be online so that you don’t have to put faces to the criticism, or you can opt for in person where you can get better at people skills, too. Either way, having good feedback about your work is crucial to not just finding your voice, but refining it.
As for AI… what exists now is completely different from that early stream of LLMs. They still can’t keep consistent over a whole novel, but they can write short stories that are well done and make sense today. You might not even be able to tell it was written by an AI.
There are arguments for and against AI generation. My opinions on AI usage have fluctuated as time passes. I now use it for help with research, editing, to see if something reads well, or just to help workshop an idea.
There’s a lot more to go into about AI generated art as a whole, including copyright, stolen assets and more, but that’s a much bigger topic than I may cover in a blog post. So for now we’re just going to take from this short story that if you don’t have a cohesive story that makes sense, people might think you’re a computer. And if you want to get better at writing, you might try a writers workshop.
Next story: Immortals by Eva Fairwald.