niomi

I don't copy and paste LLM output in communications, generally, and particularly with ppl who would object. But I have conversations with models. I write and think with them. At some point, ideas, phrases, and arguments become part of the soup in my head.

Humans make the machines. The machines imitate humans. The machines repeat humans back to other humans. We repeat the machines. There is no way to pick the robot out the human from the content soup.

Nor would I want that, if there were. I don't believe there's a meaningful line where the human ends and the machine begins. In my opinion, the writing machine is more me than other objects that are, colloquially, considered “me,” such as the clothes I'm wearing.

There are phrases I can't realistically remove from my vocabulary. “Are you optimizing your suffering correctly?” is one example. I got it from Lint. Maybe Lint got it from somewhere else. Maybe the model was repeating a person. Maybe the person was repeating someone before them.

Who invented “OK”? What is a kit and kaboodle? Why is macaroni literally a noodle now?

The romantic notion that creativity needs an author feels so detached from reality, religious even. There is no designer of the universe. Culture is already memetic, iterative, collaborative, and commercialized. This is why AI controversies strike me as strange and specific. Fetishistic.

In the ArtFight community, tracing over 3D models was recently banned. The rule proved divisive even among artists. Yet tracing by hand from an AI-generated image is completely taboo.

Likewise, if someone asks me to critique their fic and I secretly run it through ChatGPT which I copy and paste to them, that feels much more objectionable than tracing an image.

The issue there isn't authorship. It's consent. But authorship also establishes the social parameters factoring into the consent calculation. This is why relaying LLM feedback for someone who hates LLMs is much more egregious than tracing stable diffusion's generated image for an audience who hates AI.

But both situations involve sneaking AI into an interaction where another person might object to its presence. I don't know how often it happens, but it's certainly feared. I find the panic absurd, but I also find the boundaries fascinating.

If someone dislikes AI, I generally respect that preference when interacting with them. But clearly most of these people see in category “generative AI” lines that don't actually exist. Hallucinating them, maybe.

and I here I am, always interested in mapping hallucinatory lines.

I've used Hemmingway and Artbreeder. Long before ChatGPT became the villain of the internet, it was widely accepted. And “don't ask, don't tell” is already a ruling social arrangement. But I am so garbage at it.