AI is stealing writers’ words and jobs…

[trappedslider]

Its OK for you to have it? Sure. You can also draw a 1 to 1 rip of any image/art/text, and as long as you dont try and put it off as your own, or otherwise infringe on the license, nobody is going to show up at your home to take it.

Is it OK for a billion dollar company to train its software and profit off of the theft which lets you generate that image for your personal use?

In no world of ethics or logic.

you didn't answer my question: What is the difference between being influenced by an IP and outright copy violation? Is it simply being made by a big cooperation that's the line or not?

If it's not then there's thousands of freelance artist who take commissions who's artwork you can tell was influenced by jim lee's art style and others who would be in violation.

If you're against copyright violation that includes the little guy and not just the copros.

 

[Scribe]

you didn't answer my question: What is the difference between being influenced by an IP and outright copy violation? Is it simply being made by a big cooperation that's the line or not?

I'm not a copyright lawyer. I know that said lines do exist, and there is litigation around this clearly.

Copyright violation would likely be down to profiting from said theft.

If it's not then there's thousands of freelance artist who take commissions who's artwork you can tell was influenced by jim lee's art style and others who would be in violation.

Style is a lot harder to prove, and likely would be borderline unenforcable. There is a huge difference between a style, and me typing in 'Terminator robot' and getting a...well Terminator Robot. Dont you think?

If you're against copyright violation that includes the little guy and not just the copros.

Yes. If I tried to sell a book around my killer robots, and it was full of images generated that looked like the Terminator robot, I would be 10000% at fault, in the wrong.

That these models are literally trained on outright theft, is still not up for debate.

Agreed? Cool.

 

[trappedslider]

That these models are literally trained on outright theft, is still not up for debate.

Not all models are trained equally

 

[Scribe]

Not all models are trained equally

Clearly not, a few pages ago it was stated they cannot handle red eyes and blue dresses...patently false.

I'm not interested in your 'not all men' side plot.

Whats the title of this thread? Is it happening? Did major companies engage in clearly open and obvious theft?

Yes, yes, yes.

 

[trappedslider]

Clearly not, a few pages ago it was stated they cannot handle red eyes and blue dresses...patently false.

I'm not interested in your 'not all men' side plot.

It wasn't meant to be a side plot at all, I've encountered at least one that was trained on CC content and the only way I knew that was due to the creator saying so. The other models, I use locally I have no clue on what they were trained on other than a number of images.

FYI: This thread could have also just been titled "corporations being scrum bags" and it still would have made the point, but the word "stealing" is more likely to get clicks and reactions.

 

[Scribe]

FYI: This thread could have also just been titled "corporations being scrum bags" and it still would have made the point, but the word "stealing" is more likely to get clicks and reactions.

AI models can demonstrably be shown to have stolen content to train their programs.

Fact. Accept it and move on.

 

[Jfdlsjfd]

Clearly not, a few pages ago it was stated they cannot handle red eyes and blue dresses...patently false.

SDXL Juggernaut current model, prompted for "3 girls, one with blue dress and red shirt, one with green dress and yellow shirt, one with green dress and white shirt". 4 random images, no pick, from nightcafe website (easy to replicate for free if you think I lied about the methodology used to generate the images as well).

Either it's "patently false", as you said, and you're seeing four images depicting three girls in the same outfit in each. In any other case, I would respectfully suggest you retract your accusion of me lying about concept bleeding being still being something being imperfectly (to say the least...) managed, barred some prompt manipulation (up to total rewriting, like Dall-E does) and further refinement in data captioning (which will undoubtedly happen in the future as models continue to progress).

 

[Scribe]

Goal Post Shifting - 101.

Edit: rofl

 

[FrogReaver]

AI models can demonstrably be shown to have stolen content to train their programs.

Fact. Accept it and move on.

I believe they had fair use of such content to develop generative AI. I don’t believe they have use to profit off that. One is theft, the other is not.

 

[Scribe]

I believe they had fair use of such content to develop generative AI. I don’t believe they have use to profit off that. One is theft, the other is not.

If these companies actually paid for every single bit of the content they consumed, compensated every author and artist in some way, sure, I'll retract that.