Post

Another thing to look out for, besides AI prompters claiming to be human artists, are users who claim things like, "I shouldn't be labeled for AI because I trained my own model." That statement isn't accurate for a couple of reasons. 🧵 1/4

This post is unavailable.

311 reposts707 likes

In our case for @aimod.social we don’t put much stock in a bio. If someone describes themself as using AI, we might consider it a clue to look closer and confirm if they actually post AI images, but we don’t take anything at face value, either way.

Another thing is that generating new AI models that produce decent quality results requires a ton of energy. Not something a random person can afford to do by themselves. They would need to rely on an infrastructure that's sapping our limited water and creating a fuckton of emissions.

They'll literally do anything except paint their own pictures huh. Like, if AI is so great why aren't they in their OWN community(these exist)? Why pretend they make art? Almost as if fooling others is the point, not enjoyment of the… 🤢, "medium"… 🤮.

I'm honestly really thankful for people who are able to sus these accounts out because I'm going thru some mental health stuff rn & I rely on the block lists. I just need to pay attention to which lists are legit... there's so much deception going on, it's hard for me to keep up 😰

I'm sorry 🫂 Yeah, there are definitely some people out there who will use block lists irresponsibly. We have a team that cares a lot about getting it right and earning everyone's trust. If we make a mistake, we want to know so we can fix it!

1) AI is AI, regardless of the model's origin. The @aimod.social labeler is not interested in labeling the methodology or intention behind an image. Those things are much harder, or impossible, to determine and could be less factual. We are simply answering the question, "Is this AI-generated?" 2/4

What’s the proper way to report? Over the past few days I’ve sent some reports in through the on-post menu by checking the labeler, but I haven’t seen any labels on the posts or users flagged.

2. Training a model on only one person's art is not feasible. Generative models require massive amounts of diverse data to learn patterns. Even the most prolific artist couldn't produce enough art to meet that data threshold. 3/4

If someone claims they created their own model from their own art, it would likely mean they used a general model already trained on other datasets and fine-tuned it with their art—meaning it still relies heavily on other people's work. 4/4