if the AI “stole” content, then so did every artist who ever watched a Disney movie, and every author that read a book, and every programmer that looked at a repository.
This is a terrible argument. Training a model on someone’s work is not inspiration. You’re anthropomorphizing a machine learning model. It’s not inspired. It doesn’t have an internal experience that informs its creations.
It’s generating output directly based on the statistical likelihood of its training set. That’s it. This is proven by the fact that many models can recreate styles and likenesses that would warrant litigation if a human did it. A machine learning model will always be definitionally derivative.
Clearly you’re trying to justify your AI use, which I understand. Inferencing locally does help, but unless you’re training your own models, that’s what actually takes the majority of resources.
Believe it or not, I’ve experimented with a lot of AI. I’ve learned the math of how they work. I’ve run Ollama on various infrastructure. I’ve trained my own small models. I given it an honest shot. At the end of the day, their output (especially “creative” stuff) is unreliable at best and complete garbage at worst.
Many people don’t want to buy AI art, and in my opinion, they shouldn’t. Instead of blaming consumers, artists, developers, etc should do the work to learn how to actually make shit themselves and stop complaining that people don’t want slop.
This is a terrible argument. Training a model on someone’s work is not inspiration. You’re anthropomorphizing a machine learning model. It’s not inspired. It doesn’t have an internal experience that informs its creations.
It’s generating output directly based on the statistical likelihood of its training set. That’s it. This is proven by the fact that many models can recreate styles and likenesses that would warrant litigation if a human did it. A machine learning model will always be definitionally derivative.
Clearly you’re trying to justify your AI use, which I understand. Inferencing locally does help, but unless you’re training your own models, that’s what actually takes the majority of resources.
Believe it or not, I’ve experimented with a lot of AI. I’ve learned the math of how they work. I’ve run Ollama on various infrastructure. I’ve trained my own small models. I given it an honest shot. At the end of the day, their output (especially “creative” stuff) is unreliable at best and complete garbage at worst.
Many people don’t want to buy AI art, and in my opinion, they shouldn’t. Instead of blaming consumers, artists, developers, etc should do the work to learn how to actually make shit themselves and stop complaining that people don’t want slop.