None of this will ever get anywhere. And if it ever does, then way more people are going to be affected than you think.
For a start, no-one has ever managed to prove that an AI physically samples the original artworks people use in their prompts. Not on a commercially-available model, at least. The Afghan Girl "gotcha" was done using close lookalikes without any further forensic analysis (because the pixels would still be completely different, since that's not the way AI works).
The Andy Warhol case went nowhere and at best dictated that AI work can't be copyrighted.
The chimp selfie case dictated that an animal has ownership rights if it manages take a picture with photography equipment you otherwise own. An AI is not an animal.
If an artstyle can be copyrighted, then that's immediately millions of pieces of hand-drawn fan art that would now be illegal, and if anyone made money from them, they would owe the original rights holders.
There is absolutely nothing Ghibli can do unless it uses a commercially-available, non-tampered with AI model and can prove it's using original Ghibli art in the art it produces. And people have tried for the last 3 years.
The Professor test had someone request a picture of a specific professor and got a picture that looked like her About Me picture. Likely the only image of her to access.
Did it look like an existing picture? Or was it a duplicate of an existing picture?
Because the first one has every individual pixel being different from the source. The second one is a reproduction, which is clearly covered by copyright law.
I will bet a significant sum it was the first one.
Next you will request a picture of Ronald McDonald and then be surprised that the Neural Network draws a trademarked clown?
The spirit behind Nightshade is malicious, but it isn't causing much harm. It's like if someone poured gasoline into the water supply of Flint, Michigan. The water ALREADY can be lit on fire, right out of the tap, their actions have done no further harm to the Flint inhabitants, but the hatred for the innocent and sickly behind the act should still be admonished.
The ACTUAL words of the Japanese government: "its not copyright infringement if its just copying the style its only copyright infringment if its copying Ghibli"
What the article title and article author says: "Banning Ghibli style AI art is being discussed by the Jap government . Yes i know the Jap government actually said something else but actually banning it still needs to be discussed. Why? Because i said so"
Leftists already saw themselves in their communist utopia where they don't have to work and can "do art" all day. And now AI is taking that away from them :D
After what the voice actors union and their members did to video game voice acting I say: good, fuck 'em, put the fries in the bag, squeaky voice!
I think AI art should at least attribute the sources they use. Sometimes i see souless ai art and wonder who or what was the source material because im curious if the source material is good. Sometimes its a pretty anime girl and id want to know who draw those faces or figures.
Most open source generators embed the prompt into the metadata. Try opening the image in notepad and scanning through the first dozen lines or so. On the other hand, many image hosts strip out the metadata.
Why is this a left/right issue? "AI Art" is retarded anyways. Why are there so many midwits who are eager for this? Do you think it's going to make your uncreative ass profitable? Because you can copy the Ghibli style? For what.. memes?
I seriously don't get it.
Plus.. look at the people who own this technology and want to sell it to you.. do they seem like they're interested in the same social outcomes you are? Do you think these left wing companies are going to let you make the "art" you want?
Fucking baffling. It's either gay trolling or people too retarded to think this through.
Maybe because it has the potential to revolutionize the workflow for creating media, allowing teams of just a few people to go head to head with the multimillion dollar companies currently dominating the American media landscape?
You know, one guy with a vision creating something unique, instead of corporate slop that's been focus tested over and over again until it's inoffensive as possible.
Individual images aren't the endgoal here. you just haven't conceived of how generative AI might be used in a larger workflow to bolster the limited capabilities of an individual or small team.
Guy, you know that no one really owns the technology, right? Lots and lots of Neural Network projects are open source, OR rely heavily on open source implementations inside a larger framework.
If you wanted you could build a server using cheap (second hand) commercial off the shelf (CotS) parts, compile your own Neural Network model (or three) and then run them on your own hardware.
I had lunch today with someone who has three servers they use for running experimental versions of very recent Neural Network architectures.
I admit that I don't fucking trust Google or Microsoft who are rolling out services accessible through phones etc, but I have never trusted Google or Microsoft. Why is this different?
So, here is a question, since it is literally trivial to own the hardware AND the source code running on the hardware, how can "these left wing companies" stop anyone from making the "art" they want?
Every single time you open your mouth, you just demonstrate that you know almost literally nothing about the subject. You just vomit irrational hatred and spew insults. It is kind of tiring.
Well, now it is your chance to call me a gay troll faggot or whatever. Go to.
You have to have training data. Since you're so in the know you do realize this? The guy with "three neural nets" (holy fuck bro!) either has large annotated data lakes already or is using open source training data.
You know those aren't free to build or use? You know they're rarely high quality sources because of this? I mean if one works why use three? Must be a lot of success there or something.
This "technology" is a clarion call to midwits and failing managers everywhere. I am, understandably, hostile to this set. Like people who try to turn this basic and entirely objective technological observation into retarded political garbage.
Open source software (OSS) is source code developed and maintained through open collaboration. Anyone can use, examine, alter and redistribute OSS as they see fit, typically at no cost.
An example of a high profile Open Source Neural Network project is Llama 2.
Llama 2 is the next generation of Meta's open source large language model. It is free for research and commercial use, making it a great project to contribute to. It has a large community support from both Meta and Microsoft and is ideal for those interested in large language models and natural language processing
So to repeat, Llama 2 (by Microsoft / Meta partnership) has freely available source code AND is free for research and commercial use.
There are versions of the Neural Network you can get that are already trained, or you can, as you pointed out, use Open Source training data OR you can compile your own training data set to supplement (or even replace) an open source set.
For what it is worth, I think that most of the ways that Large Language Models will be used will be a net negative. I am frankly shocked at the low standard of teaching and the low skills and "care factor" of teachers. It seems counterproductive to give teachers and their managers, who don't seem to have standards or accountability, access to LLMs to plan their lessons for them and then mark the work of their students.
But this is changing the subject. You said that people would not be able to produce the art they wanted, due to commercial control of the technology and the derivative services.
None of this will ever get anywhere. And if it ever does, then way more people are going to be affected than you think.
For a start, no-one has ever managed to prove that an AI physically samples the original artworks people use in their prompts. Not on a commercially-available model, at least. The Afghan Girl "gotcha" was done using close lookalikes without any further forensic analysis (because the pixels would still be completely different, since that's not the way AI works).
The Andy Warhol case went nowhere and at best dictated that AI work can't be copyrighted.
The chimp selfie case dictated that an animal has ownership rights if it manages take a picture with photography equipment you otherwise own. An AI is not an animal.
If an artstyle can be copyrighted, then that's immediately millions of pieces of hand-drawn fan art that would now be illegal, and if anyone made money from them, they would owe the original rights holders.
There is absolutely nothing Ghibli can do unless it uses a commercially-available, non-tampered with AI model and can prove it's using original Ghibli art in the art it produces. And people have tried for the last 3 years.
Why is that even a problem? Every single artist uses preexisiting art for inspiration or to straight up copy things from. Every last one of them.
It should only be a problem for whoever trained the model, not the consumer who used the software.
Is there really a difference? Models get trained with existing art. So do artists.
Are you then supposed to make a licensing deal with the chimp...?
What about the college professor test?
How does fucking your student and then being crowned Canadian PM by the WEF relate to this case?
That's an NPC test.
The Professor test had someone request a picture of a specific professor and got a picture that looked like her About Me picture. Likely the only image of her to access.
Did it look like an existing picture? Or was it a duplicate of an existing picture?
Because the first one has every individual pixel being different from the source. The second one is a reproduction, which is clearly covered by copyright law.
I will bet a significant sum it was the first one.
Next you will request a picture of Ronald McDonald and then be surprised that the Neural Network draws a trademarked clown?
Look up Anne Graham Lotz AI. It's been a few years.
Here's an article that shows the comparison
https://archive.ph/6IjBU
Here's the paper that discovered it
https://archive.ph/fbfHd
Here's an article on the paper
https://archive.ph/DjrWZ
if AI doesn't physically sample original works then why AI users freak out over Nightshade?
No ML user or researcher is worried about Nightshade because Nightshade was just grifters scamming retartists.
Most don't.
The spirit behind Nightshade is malicious, but it isn't causing much harm. It's like if someone poured gasoline into the water supply of Flint, Michigan. The water ALREADY can be lit on fire, right out of the tap, their actions have done no further harm to the Flint inhabitants, but the hatred for the innocent and sickly behind the act should still be admonished.
https://archive.md/jzGDW
The ACTUAL words of the Japanese government: "its not copyright infringement if its just copying the style its only copyright infringment if its copying Ghibli"
What the article title and article author says: "Banning Ghibli style AI art is being discussed by the Jap government . Yes i know the Jap government actually said something else but actually banning it still needs to be discussed. Why? Because i said so"
Leftists already saw themselves in their communist utopia where they don't have to work and can "do art" all day. And now AI is taking that away from them :D
After what the voice actors union and their members did to video game voice acting I say: good, fuck 'em, put the fries in the bag, squeaky voice!
I think AI art should at least attribute the sources they use. Sometimes i see souless ai art and wonder who or what was the source material because im curious if the source material is good. Sometimes its a pretty anime girl and id want to know who draw those faces or figures.
Most open source generators embed the prompt into the metadata. Try opening the image in notepad and scanning through the first dozen lines or so. On the other hand, many image hosts strip out the metadata.
oh i didnt know that hmmm
Why is this a left/right issue? "AI Art" is retarded anyways. Why are there so many midwits who are eager for this? Do you think it's going to make your uncreative ass profitable? Because you can copy the Ghibli style? For what.. memes?
I seriously don't get it.
Plus.. look at the people who own this technology and want to sell it to you.. do they seem like they're interested in the same social outcomes you are? Do you think these left wing companies are going to let you make the "art" you want?
Fucking baffling. It's either gay trolling or people too retarded to think this through.
Maybe because it has the potential to revolutionize the workflow for creating media, allowing teams of just a few people to go head to head with the multimillion dollar companies currently dominating the American media landscape?
You know, one guy with a vision creating something unique, instead of corporate slop that's been focus tested over and over again until it's inoffensive as possible.
Individual images aren't the endgoal here. you just haven't conceived of how generative AI might be used in a larger workflow to bolster the limited capabilities of an individual or small team.
Guy, you know that no one really owns the technology, right? Lots and lots of Neural Network projects are open source, OR rely heavily on open source implementations inside a larger framework.
If you wanted you could build a server using cheap (second hand) commercial off the shelf (CotS) parts, compile your own Neural Network model (or three) and then run them on your own hardware.
I had lunch today with someone who has three servers they use for running experimental versions of very recent Neural Network architectures.
I admit that I don't fucking trust Google or Microsoft who are rolling out services accessible through phones etc, but I have never trusted Google or Microsoft. Why is this different?
So, here is a question, since it is literally trivial to own the hardware AND the source code running on the hardware, how can "these left wing companies" stop anyone from making the "art" they want?
Every single time you open your mouth, you just demonstrate that you know almost literally nothing about the subject. You just vomit irrational hatred and spew insults. It is kind of tiring.
Well, now it is your chance to call me a gay troll faggot or whatever. Go to.
You have to have training data. Since you're so in the know you do realize this? The guy with "three neural nets" (holy fuck bro!) either has large annotated data lakes already or is using open source training data.
You know those aren't free to build or use? You know they're rarely high quality sources because of this? I mean if one works why use three? Must be a lot of success there or something.
This "technology" is a clarion call to midwits and failing managers everywhere. I am, understandably, hostile to this set. Like people who try to turn this basic and entirely objective technological observation into retarded political garbage.
Go whine to a chatbot or something.
No, I said he had three servers for Neural Networks. Want to take a guess on how many servers he has for other uses? I mean, it is his literal job.
Do you really, truly, honestly not know what Open Source means? Fuck! you don't! Okay. I can help.
From IBM.com
An example of a high profile Open Source Neural Network project is Llama 2.
So to repeat, Llama 2 (by Microsoft / Meta partnership) has freely available source code AND is free for research and commercial use.
There are versions of the Neural Network you can get that are already trained, or you can, as you pointed out, use Open Source training data OR you can compile your own training data set to supplement (or even replace) an open source set.
For what it is worth, I think that most of the ways that Large Language Models will be used will be a net negative. I am frankly shocked at the low standard of teaching and the low skills and "care factor" of teachers. It seems counterproductive to give teachers and their managers, who don't seem to have standards or accountability, access to LLMs to plan their lessons for them and then mark the work of their students.
But this is changing the subject. You said that people would not be able to produce the art they wanted, due to commercial control of the technology and the derivative services.
So the question is: Did you know that was a lie?
Now we're championing AI-slop to own the fags?
Now we have a moral obligation to make more Ghibli art because that is a BS ruling.
Now we're championing AI-slop to own the fags?
Yes, and? Fuck them.