I am a failed writer in the sense that I have never had my fiction professionally published. These posts, which will run on most Fridays, are an attempt to keep myself creatively motivated and just generally discuss the creative process from someone trying to figure it out. I genuinely love the process of making things — any things, from writing to drawing to music to woodworking to baking. Maybe my own failures can be a source of amusement or interest to others.
A new imitative AI product is available for “authors”. It promises to take a simple prompt and by chaining together several imitative AI systems to generate an entire novel. Putting aside the quality issues (which, given how poorly imitative AI writes past a certain size in certain generic domains, you probably shouldn’t.), this program fundamentally misunderstands writing in two ways. First, it seems to think that ideas are the hard part. Second, it does not understand that writing, especially fiction writing, is a person-to-person communication.
I am probably taking this too seriously. I suspect the point of his tool is the swamp Amazon with self-published Fancy Clippy novels in the hopes that sheer volume will make you some real money before the whole system craters under the weight of AI garbage. But I see enough apparently sincere people who think that writing is the product not the process that it’s probably worth addressing why I think they are fundamentally wrong.
First, ideas are not hard to come by. I have a hundred or more ideas in a OneNote file — ideas happen all the time. I have done this before, but I am happy to give away a few:
-
Zombies taste of chicken — a zombie story where agriculture is based on raising an eating zombies. Because we are not descendants of hunters, we are descendants of the clever little buggers that figured out of you put a boy cow and a girl cow in a pen, pretty soon you had all the steak you wanted. Why kill zombies when you can feast on them?
-
Fsck English magic — a historical fantasy where the workers ruined by loom technologies in the early 1800s have magic to fight back with. Think if it as a Marxist Johnathon Strange and Mr. Norell.
-
Luddites vs Necromancers: The looms are powered by the captured spirits of dead craftsmen.
-
Would you kill a billion people to make the rest nicer: Someone creates a means of infecting people with empathy, but it kills about 1 out of every 8 people. Could be the story of the people who do it, could be the story of the people trying to stop them.
-
Exasperated Space Watson: Sherlock Holmes an upper-class twit in spaaaaaaace and his Watson has to go behind him and really solve the murders his class bias doesn’t let Sherlock understand.
I could go on, but you get the point. Ideas are easy — execution is hard. Feel free to write a story based on these ideas, by the way (though I am actively working on the last one, so maybe let me see if I can finish it first?). Whatever you write will not be the same thing I would have written.
That last is the critical point. Even if you argue that the people behind these systems agree that ideas are easy, execution is hard, they miss the point of the execution. Imitative AI systems are just that — imitative. It is why they have such a hard time with copywriter issues. All they can do is imitate. Even assuming you can control for plagiarism perfectly — which no system has yet proven able to do — what they produce is going to tend toward the middle of quality line. These systems calculate what comes next based on their training data, so it is inevitable that they will converge on the mean output. And the mean outcome is pretty bland and pointless.
Every writer, even the most derivative and hacky (looks around guiltily), brings something of themselves to the story. That is why Eorl Flynn starred in a classic Robin hood move and Kevin Costner … did not. the story was largely the same, but the execution was widely different. And that is good (personally, I loved Alan Rickman as the villain in the Costner Robin Hood much more than the bland villain they had in the Eorl Flynn version). The communication of human meaning, spirit, heart — choose whatever corny descriptor you like — is what makes fiction worth reading. Even the worst piece of prose every written by a person has something of them in it, and that something provides at least the hope of enjoyment and/or meaning for others. These autocompletes on steroids, these Fancy Clippys, cannot by definition provide that spark. They are simply regurgitation machines, destined to collapse to the mean.
I get it. Writing is hard. I am a failure at it too, in the sense that no one has published me. But the hard is what makes it worth doing. The hard is where the good comes from, where the connection, the reason something is worth experiencing, are created. No imitative AI system can, by its very nature, produce that meaning. I am no tech hater. I am in IT and am well aware of how much better our lives are because of technology. But that does not mean I want to substitute the pleasure of real human writing and creativity for a wan imitation produced faster.
When and if we get real general Artificial Intelligence intelligence that can feel and reason for itself, not just copy what others have done, I will be happy to read what it writes. Until then, I want actual, honest, human connection and meaning in what I read.
Weekly Word Count
three pages on the Clockmaker’s wife script. Work has swallowed me whole the last week.
I have a functional murder in the Sherlock Holmes in spaaaaaaaaaaaaaaaace potting, but it’s pretty lame as murder mysteries go. May need to play down the mystery for other aspects.