AI Created images (generated by Midjourney) lose registered copyrights

[mr moose]

On 3/18/2023 at 10:52 AM, LAwLz said:

Or does this rule only apply to AI programs where natural language can be used as the input? It seems like a lot of people make that distinction, for some reason.

 

 

I really don't see it as being that complicated, human created work with a  camera or pencil or even digital workbench versus AI created work derived from text commands seem to be two easily distinguishable conditions. Did the AI simply fix or modify an existing piece of human art? or did the AI create the entire piece based on instructions given it?

 

 

[05032-Mendicant-Bias]

40 minutes ago, mr moose said:

 

I really don't see it as being that complicated, human created work with a  camera or pencil or even digital workbench versus AI created work derived from text commands seem to be two easily distinguishable conditions.

I make a photo of my StableDiffusion image, or open photoshop and apply a 0.01% sharpen filter, and voila, I have copyright now!

[LAwLz]

1 hour ago, mr moose said:

I really don't see it as being that complicated, human created work with a  camera or pencil or even digital workbench versus AI created work derived from text commands seem to be two easily distinguishable conditions. Did the AI simply fix or modify an existing piece of human art? or did the AI create the entire piece based on instructions given it?

I think the problem with that view is, how do we quantify and qualify what is a "human created work" vs an "AI created work"?

What measurable thing makes a picture taken with a camera a "human created work" and what measurable thing makes a picture generated with a program like Stable Diffusion a "AI created work"?

 

Samsung were recently "caught" using AI to modify pictures of the moon on their camera. Is a picture taken with said camera a "human created work" or an "AI created work", and where is the line?

 

I feel like that line of thinking, which many people seem to have, is based on a notion that they don't have to quantify these criteria because they "know it when they see it", but then in the next breath rejects anyone who doesn't see it the same way. For example I would argue that modern cameras are basically "generating pictures" on the behalf of the human holding the camera.

[Kinda Bottlenecked]

The new guidelines that explicitly mentioned AI and its uses certainly do make it clearer for everyone. I do mostly agree with them too. They've done a much better job at explaining than I could have for my past couple of replies. Usage of AI beyond "de minimis" should rightfully be disclosed and excluded from the copyright.

 

I do also hope this sets a good precedent for AI users to disclose the models they've used and so on. Like a digital artist would for the applications they've used. 

 

In other news, a new tool has been released for artists.

 

For as long as we as a collective do not set ethical guidelines for ourselves, I fear this constant back and forth will continue. 

 

[Quackers101]

"it analyzes your art", "become one of us!"

[HenrySalayne]

8 hours ago, LAwLz said:

Samsung were recently "caught" using AI to modify pictures of the moon on their camera. Is a picture taken with said camera a "human created work" or an "AI created work", and where is the line?

[...]]For example I would argue that modern cameras are basically "generating pictures" on the behalf of the human holding the camera.

I honestly think Samsung is a bad example here, because they did not "create" the picture, they "guessed" details and sharpness back into the picture.

 

9 hours ago, LAwLz said:

I think the problem with that view is, how do we quantify and qualify what is a "human created work" vs an "AI created work"?

What measurable thing makes a picture taken with a camera a "human created work" and what measurable thing makes a picture generated with a program like Stable Diffusion a "AI created work"?

[...]

I feel like that line of thinking, which many people seem to have, is based on a notion that they don't have to quantify these criteria because they "know it when they see it", but then in the next breath rejects anyone who doesn't see it the same way.

There is not one point where a picture suddenly becomes AI generated. In the near future you might be able to replace a pole with a tree or change the season or weather of a picture. But the picture composition is still based of a real picture and this underlying picture is a human created work. You cannot draw a clear line here.
When you start with only a text prompt and t the rest is done by the automatom, you are missing a vital layer.

[mr moose]

19 hours ago, 05032-Mendicant-Bias said:

I make a photo of my StableDiffusion image, or open photoshop and apply a 0.01% sharpen filter, and voila, I have copyright now!

There is still an easily distinguishable difference between using camera to make a copy of something and taking a picture that includes said art.  There is a reason you cannot take a photo of the monalisa and copyright it.  Exactly the same concept applies here, taking a photo of something that can;t be copyrighted doesn't make the photo copyrightable.  Maybe if you take a photo of it and change it quite substantially without using text only AI to do the work, then you could.

 

18 hours ago, LAwLz said:

I think the problem with that view is, how do we quantify and qualify what is a "human created work" vs an "AI created work"?

I think that has been quantified already.  If you create the work yourself and only use AI to do some of the work then it is human, if you tell an AI to make it then it is not. 

18 hours ago, LAwLz said:

What measurable thing makes a picture taken with a camera a "human created work" and what measurable thing makes a picture generated with a program like Stable Diffusion a "AI created work"?

Because the picture had to be framed and the person had to physically do the work of picking the target, choosing he settings and pressing the shutter button.  Text input AI does none of this.

 

18 hours ago, LAwLz said:

 

Samsung were recently "caught" using AI to modify pictures of the moon on their camera. Is a picture taken with said camera a "human created work" or an "AI created work", and where is the line?

The photo of the moon is copyrightable because A human took an actual picture of the moon,  The AI just modified it.  I don't see why this needs to mean anything.  I know the difference, you know the difference.  Claiming an image that is purely the result of text commands is not human created anymore than asking a painter to do your portrait then trying to claim you did the work.

18 hours ago, LAwLz said:

I feel like that line of thinking, which many people seem to have, is based on a notion that they don't have to quantify these criteria because they "know it when they see it", but then in the next breath rejects anyone who doesn't see it the same way. For example I would argue that modern cameras are basically "generating pictures" on the behalf of the human holding the camera.

I feel that criteria has already been quantified.   Your argument about camera's is wrong, the only picture they are taking is the one the human points it at and adjusts the settings for.  If the human typed "take a picture of the beach" then the camera got on a bus and went to the beach and took a picture, then you might have an argument.  But they don't.  The human element is in the fact that a human actually had to do some work to get the camera to the place they want a picture of, then choose the right settings (and AI features), then frame the shot.  Telling a computer to create an image of X is not even remotely the same, even if you spend hours feeding it samples.

 

I'm all in favor of AI art being accepted as art, but I don;t buy for a second that it is no different to human created art.  

[LAwLz]

On 3/21/2023 at 8:43 AM, mr moose said:

I think that has been quantified already.  If you create the work yourself and only use AI to do some of the work then it is human, if you tell an AI to make it then it is not. 

I don't think it has been quantified.

The whole debate is a spectrum so it is a bit silly to just pretend like it's binary. What defines "some work"? What defines "tell an AI"? Is me pointing the content aware fill at a pixel in Photoshop me "telling an AI" to do something? Is me typing a string of commands to a program me "telling an AI to make something"? What if the AI portion was replaced with more traditional, non-neutral network based functions? Is it suddenly me doing the creation even though I am "just" typing in a bunch of commands into a computer?

 

 

On 3/21/2023 at 8:43 AM, mr moose said:

Because the picture had to be framed and the person had to physically do the work of picking the target, choosing he settings and pressing the shutter button.  Text input AI does none of this.

I can kind of see the logic when it comes to text input, but what about img2img AIs where the input is an image? In those scenarios it is unarguably me, who wrote the original image, that decided things like how the image should be framed/composition and what should be included in the image. Yet I think a lot of people would argue that it is still the "AI who created it".

This is what im2img looks like if case you haven't seen it. Someone drew the thing on the left, inputted it into Stable Diffusion, tweaked some settings and added some prompts, and the right picture is the output. Would you be okay with this being copyrighted? I feel like a lot of people wouldn't, because this is a spectrum and people draw lines in different places.

Spoiler

The denoising scale makes it even more clear that this is a spectrum and not binary. I think different people would have different opinions for when it stopped being "AI enhanced" and just became "AI created" in this picture (warning, picture of a woman in a bikini):

Spoiler

 

Also, even when using text input it is very possible for a human to pick the target, choosing the settings and pressing the "capture" button for an AI. I don't see why putting "sunset" into a text prompt is less "human created" than pointing a camera at a sunset. 

 

Is it just because people aren't used to text being the portion the human puts in? Because I would argue that things like mouse movements and keyboard shortcuts are "text" as well.

 

 

 

On 3/21/2023 at 8:43 AM, mr moose said:

The photo of the moon is copyrightable because A human took an actual picture of the moon,  The AI just modified it.  I don't see why this needs to mean anything.  I know the difference, you know the difference.  Claiming an image that is purely the result of text commands is not human created anymore than asking a painter to do your portrait then trying to claim you did the work.

I don't actually know the difference, which is why I am asking.

I think everyone has a different interpretation of this and it is a spectrum, which is why I am against people just brushing things off and going "I know it when I see it", which is what you are doing now.

Do I understand you correctly. Your objection is purely against images where the only human input was text and settings. That is how you distinguish between if something is "human created but AI enhanced" vs strictly "AI created". Correct?

 

 

On 3/21/2023 at 8:43 AM, mr moose said:

Your argument about camera's is wrong, the only picture they are taking is the one the human points it at and adjusts the settings for.

Yes, but what if I said "the only picture being created is the one the human came up with and created using a program"? 

At the end of the day, a program like Stable Diffusion is just acting according to the instructions given by a human, just like a camera is acting according to the instructions given by a human. I don't see the difference, except one is text input (in some cases) and the other one is physical movement inputs.

 

 

On 3/21/2023 at 8:43 AM, mr moose said:

If the human typed "take a picture of the beach" then the camera got on a bus and went to the beach and took a picture, then you might have an argument.  But they don't.

Isn't this basically what drones do? A human tells it to go somewhere, it does, and then it can take a picture for you. We literally have cameras that can go to places for you and take pictures, and I don't see anyone disputing the idea that the drone operator is the one "creating" the image, even though they 100% rely on very advanced software to not only steer the drone but also capture the image to begin with.

 

 

On 3/21/2023 at 8:43 AM, mr moose said:

The human element is in the fact that a human actually had to do some work to get the camera to the place they want a picture of, then choose the right settings (and AI features), then frame the shot.  Telling a computer to create an image of X is not even remotely the same, even if you spend hours feeding it samples.

But isn't inputting the creativity and ideas into the prompt the "human element"?

Isn't creating and inputting the actual pictures the AI might be trained on also an "human element"?

 

If I drew 100 pictures myself, and created an AI model based on that, and then spent hours tweaking settings and prompts until I got a picture that matched the idea I had in my head when I set out to create an image, is that not my creation?

[Arika]

On 3/21/2023 at 6:43 PM, mr moose said:

I think that has been quantified already.  If you create the work yourself and only use AI to do some of the work then it is human, if you tell an AI to make it then it is not. 

which is going to be an extremely hard thing to prove.

 

If i draw a picture, colour it, then run it through my SD model to create the "final" picture. how would one determine that there was effort of the full drawing beforehand vs just creating an image based on a prompt? Using the same model, the output would be of a very similar style and still have the telltale signs of an AI picture.

 

For example, here is 2 pictures. that that was purely prompted. and 1 that was ran through img2img with one of my unreleased drawings with the prompt chunk text removed.

 

Spoiler

 

 

Spoiler

 

 

Could i release the original drawing? yes...but lets say i have now deleted the original drawing as i no longer have use for it. what now? i can't -prove- to the world there was effort put in beforehand...but there was.

[mr moose]

1 hour ago, LAwLz said:

I don't think it has been quantified.

The whole debate is a spectrum so it is a bit silly to just pretend like it's binary. What defines "some work"? What defines "tell an AI"? Is me pointing the content aware fill at a pixel in Photoshop me "telling an AI" to do something? Is me typing a string of commands to a program me "telling an AI to make something"? What if the AI portion was replaced with more traditional, non-neutral network based functions? Is it suddenly me doing the creation even though I am "just" typing in a bunch of commands into a computer?

 

 

I can kind of see the logic when it comes to text input, but what about img2img AIs where the input is an image? In those scenarios it is unarguably me, who wrote the original image, that decided things like how the image should be framed/composition and what should be included in the image. Yet I think a lot of people would argue that it is still the "AI who created it".

This is what im2img looks like if case you haven't seen it. Someone drew the thing on the left, inputted it into Stable Diffusion, tweaked some settings and added some prompts, and the right picture is the output. Would you be okay with this being copyrighted? I feel like a lot of people wouldn't, because this is a spectrum and people draw lines in different places.

  Reveal hidden contents

The denoising scale makes it even more clear that this is a spectrum and not binary. I think different people would have different opinions for when it stopped being "AI enhanced" and just became "AI created" in this picture (warning, picture of a woman in a bikini):

  Reveal hidden contents

 

Also, even when using text input it is very possible for a human to pick the target, choosing the settings and pressing the "capture" button for an AI. I don't see why putting "sunset" into a text prompt is less "human created" than pointing a camera at a sunset. 

 

Is it just because people aren't used to text being the portion the human puts in? Because I would argue that things like mouse movements and keyboard shortcuts are "text" as well.

 

 

 

I don't actually know the difference, which is why I am asking.

I think everyone has a different interpretation of this and it is a spectrum, which is why I am against people just brushing things off and going "I know it when I see it", which is what you are doing now.

Do I understand you correctly. Your objection is purely against images where the only human input was text and settings. That is how you distinguish between if something is "human created but AI enhanced" vs strictly "AI created". Correct?

 

 

Yes, but what if I said "the only picture being created is the one the human came up with and created using a program"? 

At the end of the day, a program like Stable Diffusion is just acting according to the instructions given by a human, just like a camera is acting according to the instructions given by a human. I don't see the difference, except one is text input (in some cases) and the other one is physical movement inputs.

 

 

Isn't this basically what drones do? A human tells it to go somewhere, it does, and then it can take a picture for you. We literally have cameras that can go to places for you and take pictures, and I don't see anyone disputing the idea that the drone operator is the one "creating" the image, even though they 100% rely on very advanced software to not only steer the drone but also capture the image to begin with.

 

 

But isn't inputting the creativity and ideas into the prompt the "human element"?

Isn't creating and inputting the actual pictures the AI might be trained on also an "human element"?

 

If I drew 100 pictures myself, and created an AI model based on that, and then spent hours tweaking settings and prompts until I got a picture that matched the idea I had in my head when I set out to create an image, is that not my creation?

I don't understand why you cannot see it.  In one situation the AI is simply modifying an existing artwork and in the other the AI is creating the artwork literally from the ground up.   It literally is that simple.

 

Telling a computer to frame a specific way is still not  the same as going out to a place, pointing the camera and physically through your interaction with the camera making the foundation for your art.   If you draw a picture then you have started with actual art that you have created,  that is fine, I don't see anyone or laws arguing you can't draw a picture then use AI to adjust it until you re happy with the end result.

 

I think the issue is you are putting way to much thought and importance on defining what art is and not enough thought into the basic self evidence of human creativity.  We can all describe a picture, but only some people can create that picture without AI.  AI created (when the human doesn't do anything except tell the computer to generate an image) is still art and many people will still want to look at it and have it be a thing,  but there is a difference (and I believe it is a self evident one) between and human artist and a bunch of code that is simply given instructions.  As I said earlier, when you commission an artist to paint you a picture, you don;t claim to be the one that created the picture, you just gave someone the instruction on what you wanted.  It is the same in building, you wouldn't claim to have built a house just because you described it to a builder and they made it happen.

 

 

 

 

 

 

 

 

[mr moose]

2 minutes ago, Arika S said:

which is going to be an extremely hard thing to prove.

 

If i draw a picture, colour it, then run it through my SD model to create the "final" picture. how would one determine that there was effort of the full drawing beforehand vs just creating an image based on a prompt? Using the same model, the output would be of a very similar style and still have the telltale signs of an AI picture.

I would argue that it is copyrightable art, simply because you did the foundation work, framing the image etc.  Obviously it will become very hard to differentiate between human derived art and wholly AI created art, but I also don't think that changes what I am saying,  I don't think you have to be able to prove how something was created to be able to draw a line between AI art and human created art.

2 minutes ago, Arika S said:

For example, here is 2 pictures. that that was purely prompted. and 1 that was ran through img2img with one of my unreleased drawings with the prompt chunk text removed.

 

  Reveal hidden contents

 

  Reveal hidden contents

 

Could i release the original drawing? yes...but lets say i have now deleted the original drawing as i no longer have use for it. what now? i can't -prove- to the world there was effort put in beforehand...but there was.

Maybe not, but my arguments aren't actually for or against what should be copyrighted and what shouldn't, I'm just saying the process of understanding what is wholly AI and what human art is not as complicated as some seem to think it is.  The ultimate truth always exist regardless if we have proof for it or not.  I could speed to work tomorrow, but without passing a speed camera there would be no proof, that does not mean I didn't speed, didn't break the law and didn't take a risk. It just means no one can prove I did. It's the same with your art, one one hand you can create what we would define as copyrightable art but on the other hand have no proof it wasn't all AI, therefore you can't copyright it. (just like I should get a ticket but won't)

 

 

[HenrySalayne]

1 hour ago, Arika S said:

For example, here is 2 pictures. that that was purely prompted. and 1 that was ran through img2img with one of my unreleased drawings with the prompt chunk text removed.

Can you describe the process in more detail? You used the same image generator - once with a text input and once with an image input?

[Kisai]

4 hours ago, Arika S said:

which is going to be an extremely hard thing to prove.

 

If i draw a picture, colour it, then run it through my SD model to create the "final" picture. how would one determine that there was effort of the full drawing beforehand vs just creating an image based on a prompt? Using the same model, the output would be of a very similar style and still have the telltale signs of an AI picture.

 

Just on the basis of what's going on. If you drew an image in CSP and then pushed it through SD to get a colored/shaded result. That's likely fine, and isn't particularly the argument being made in the copyright case. In the copyright case, it was generated without a drawing, and only trivial edits were made to the image generated by midjourney.

 

4 hours ago, Arika S said:

For example, here is 2 pictures. that that was purely prompted. and 1 that was ran through img2img with one of my unreleased drawings with the prompt chunk text removed.

 

  Hide contents

 

  Hide contents

 

Could i release the original drawing? yes...but lets say i have now deleted the original drawing as i no longer have use for it. what now? i can't -prove- to the world there was effort put in beforehand...but there was.

Well that's on you then if you were called out on it. If you're creating only personal works, then it's likely not even a problem for YOU. However if you were creating a work that you wished to publish and profit from, it would be a problem because non-disclosure of the AI used would lead to cancelation of the copyright registration.

 

 

4 hours ago, mr moose said:

I would argue that it is copyrightable art, simply because you did the foundation work, framing the image etc.  Obviously it will become very hard to differentiate between human derived art and wholly AI created art, but I also don't think that changes what I am saying,  I don't think you have to be able to prove how something was created to be able to draw a line between AI art and human created art.

I think, pedantic arguments aside, that most people on "tech" forums are actually correct. It doesn't matter who or what "generates" the art, as long as there it's evident where the human portion began BEFORE being pushed through the AI. If it starts with the AI, then it can't be copyrighted. Simple as that. Anyone with the same seed/words/mode/software combo could get the exact same image, where as someone with a camera, will get a different image one second to the next, and two people standing side by side with the exact same model of camera, will still get a different image. With a camera, it is impossible to "duplicate" an image without all the conditions being static, and most don't try for that.

 

 

4 hours ago, mr moose said:

Maybe not, but my arguments aren't actually for or against what should be copyrighted and what shouldn't, I'm just saying the process of understanding what is wholly AI and what human art is not as complicated as some seem to think it is.  The ultimate truth always exist regardless if we have proof for it or not.  I could speed to work tomorrow, but without passing a speed camera there would be no proof, that does not mean I didn't speed, didn't break the law and didn't take a risk. It just means no one can prove I did. It's the same with your art, one one hand you can create what we would define as copyrightable art but on the other hand have no proof it wasn't all AI, therefore you can't copyright it. (just like I should get a ticket but won't)

 

 

 

There will of course be people who will try to pass off things as human created works when they are not, and doing so creates a risk to them, just like speeding does. You can speed all the time and not get caught, but that one time you do, it's an expensive lesson.

 

Ultimately, if an AI work is not eligible for copyright protection that does open the flood gates to people "stealing" copyrighted works and claiming them as AI that isn't protected. So it's likely some kind of AI copyright may be created at some point that doesn't so much protect the AI, or AI-generated works, but rather requires disclosing the entire input to protect "that work" from having no copyright. eg, "X words, Y model, Z hyperparameters, creates this image" , therefor this use of THIS image in this project is copyrighted.

 

But use of that the image outside the project is still fair game as long as it's generated the same way.

[05032-Mendicant-Bias]

On 2/24/2023 at 8:08 PM, Lightwreather JfromN said:

Adobe has been including a lot of ML

That was fast.

 

[mr moose]

17 hours ago, Kisai said:

 

Ultimately, if an AI work is not eligible for copyright protection that does open the flood gates to people "stealing" copyrighted works and claiming them as AI that isn't protected. So it's likely some kind of AI copyright may be created at some point that doesn't so much protect the AI, or AI-generated works, but rather requires disclosing the entire input to protect "that work" from having no copyright. eg, "X words, Y model, Z hyperparameters, creates this image" , therefor this use of THIS image in this project is copyrighted.

 

But use of that the image outside the project is still fair game as long as it's generated the same way.

The current copyright process requires you to show your works, therefore if you want an official copyright and your artwork is not AI (because you showed your human input before anything else) then no one can argue the copyright isn't valid.

 

If we are talking about art work that is not officially copyrighted but relies on that part of law which automatically gives you copyright, then as a civil law matter it is up to the plaintiff to provide the proof not the defendant, the only thing the defendant has to do is defend against any evidence not the plaintiff's claims.  And as Arika has already pointed out, it's going to be hard to prove which one it is just by looking at the finished product.  It would be easier for someone who has used a copyright image to incite fair use policy as their defense rather than try to make claims they likely won't be able to sufficiently evidence before a judge.

[Kisai]

10 minutes ago, mr moose said:

The current copyright process requires you to show your works, therefore if you want an official copyright and your artwork is not AI (because you showed your human input before anything else) then no one can argue the copyright isn't valid.

 

If we are talking about art work that is not officially copyrighted but relies on that part of law which automatically gives you copyright, then as a civil law matter it is up to the plaintiff to provide the proof not the defendant, the only thing the defendant has to do is defend against any evidence not the plaintiff's claims.  And as Arika has already pointed out, it's going to be hard to prove which one it is just by looking at the finished product.  It would be easier for someone who has used a copyright image to incite fair use policy as their defense rather than try to make claims they likely won't be able to sufficiently evidence before a judge.

More concerned of this kind of scenario:

1) I produce something, music, video, etc

2) I hire someone to create some non-trivial artwork, and it's not disclosed to me it's AI artwork and thus "not copyrightable"

3) The combined work (eg the visual work is incorporated into the full work in a non-trivial way, eg cover artwork, backgrounds, etc)

 

And then someone (perhaps who ever actually initiated the AI art generator) goes "I want that" and rips off the entire work, knowing that it contains an undisclosed piece of AI artwork, and thus challenges the copyright on the entire work.

 

Like it's kind of like the submarine patent problem. Where an undisclosed AI part of the artwork will result in termination of the copyright on the entire work, and you need that copyright on the entire work in order to impose punitive damages on an infringer. 

 

What it feels like, is that using AI art generators in any capacity is poison to retaining the copyright on the entire work, so it must be fully disclosed, even if the "AI" part is not protected, omitting that is committing fraud.

 

[mr moose]

2 minutes ago, Kisai said:

More concerned of this kind of scenario:

1) I produce something, music, video, etc

2) I hire someone to create some non-trivial artwork, and it's not disclosed to me it's AI artwork and thus "not copyrightable"

3) The combined work (eg the visual work is incorporated into the full work in a non-trivial way, eg cover artwork, backgrounds, etc)

 

And then someone (perhaps who ever actually initiated the AI art generator) goes "I want that" and rips off the entire work, knowing that it contains an undisclosed piece of AI artwork, and thus challenges the copyright on the entire work.

 

Like it's kind of like the submarine patent problem. Where an undisclosed AI part of the artwork will result in termination of the copyright on the entire work, and you need that copyright on the entire work in order to impose punitive damages on an infringer. 

 

What it feels like, is that using AI art generators in any capacity is poison to retaining the copyright on the entire work, so it must be fully disclosed, even if the "AI" part is not protected, omitting that is committing fraud.

 

 

Then you have a problem with people misrepresenting their work to you,  this already happens and is not a fault of copyright law (the part being that you can't CR certain AI work).    IP laws didn't stop many big name manufacturers getting sold dodgy capacitors in the 90's.

 

Also if you are about to use an image for any commercial/business endeavor then you really should copyright it or you may have trouble suing someone who uses it.  This process will make the artist liable for any fraud on their part.

 

[HenrySalayne]

1 hour ago, Kisai said:

More concerned of this kind of scenario:

1) I produce something, music, video, etc

2) I hire someone to create some non-trivial artwork, and it's not disclosed to me it's AI artwork and thus "not copyrightable"

3) The combined work (eg the visual work is incorporated into the full work in a non-trivial way, eg cover artwork, backgrounds, etc)

 

And then someone (perhaps who ever actually initiated the AI art generator) goes "I want that" and rips off the entire work, knowing that it contains an undisclosed piece of AI artwork, and thus challenges the copyright on the entire work.

 

Like it's kind of like the submarine patent problem. Where an undisclosed AI part of the artwork will result in termination of the copyright on the entire work, and you need that copyright on the entire work in order to impose punitive damages on an infringer. 

 

What it feels like, is that using AI art generators in any capacity is poison to retaining the copyright on the entire work, so it must be fully disclosed, even if the "AI" part is not protected, omitting that is committing fraud.

I might not know the laws concerning IP at your location, but generally this is not how it works.

A patent is an explicit protection, copyright is an implicit protection. You have to apply for a patent and you have to "prove" the ingenuity. Creating a work is the only step required to have it copyrighted.

European copyright laws explicitly states that "translations, among other alternations, are protected as original works without prejudice to the copyright in the original work". If you upload a video with snippets from another video (and we ignore fair-use), this video is an independent original work. Nobody can claim it and use it, it is your work and you have the copyright.

The copyright holder for the snippets has to permit the use of their work. If they do not permit the use, they can take down your video (and you can no longer "use" your work). But they cannot monetize* or own or use your work. If you use non-copyrighted material (works with expired copyright or AI generated material), nobody can do anything. There is nothing from the non-copyrighted material bleeding into you work. Somebody could extract the non-copyrighted material from your work and use them for their own purposes. but that's it.

 

 

*) Youtube has this cheesy tactic to just redirect the monetization for the use of copyrighted material and still get their share. It is not really in accordance with copyright law.

[Kisai]

3 hours ago, mr moose said:

 

Then you have a problem with people misrepresenting their work to you,  this already happens and is not a fault of copyright law (the part being that you can't CR certain AI work).    IP laws didn't stop many big name manufacturers getting sold dodgy capacitors in the 90's.

That was a different situation where a dodgy formula was stolen and then used by a bunch of different companies.

 

3 hours ago, mr moose said:

Also if you are about to use an image for any commercial/business endeavor then you really should copyright it or you may have trouble suing someone who uses it.  This process will make the artist liable for any fraud on their part.

 

Yes, but that puts it back into this catch-22 situation of "I can't own the copyright on something AI generated, and if it was not disclosed to me, then I can lose the copyright on everything containing it as well."

 

Like, fortunately, I'm not interested in AI generated artwork for producing something commercially, but I might want to produce parody or satire material, and that is already something that artists don't want their name attached to due to the crazyballs twitter cancel culture. Like a lot of foreign artists basically have in their terms and conditions "no politics, no disturbing the peace". So AI generated disposable images fits this, and no artist can whine about not being paid, because they would have never accepted the job.

 

But that's just a thought experiment for the moment.

 

2 hours ago, HenrySalayne said:

 

European copyright laws explicitly states that "translations, among other alternations, are protected as original works without prejudice to the copyright in the original work". If you upload a video with snippets from another video (and we ignore fair-use), this video is an independent original work. Nobody can claim it and use it, it is your work and you have the copyright.

 

Europe has it's own problems regarding copyright, but generally the world follows the US on it, because it's the US strong-arming foreign countries into adopting aggressive copyright laws to protect US content.

 

Stuff like this

https://www.streemtunes.com/post/how-did-pewdiepie-get-a-copyright-strike-for-using-his-own-music

Is why AI generated content will be a problem. If it's de-facto uncopyrightable, everyone will just rip off AI generated stuff. Which probably wouldn't be that bad of a thing in the grand scheme of things, but you will have people who used some AI-generated material, and in turn made some commercially viable content with it.

 

For example:

Also this one

 

These are songs made from the results of "google translate" and google's text to speech. The results of both of those are AI, but the creator also created the tunes and auto-tuned the text to speech voice. But under this copyright rule, the voice track would not be copyrightable because it's clearly an AI. And yes, that is also clearly a text prompt into google translate.

 

Then you have Vocaloid. Vocaloid is defined as an instrument. But how is it really any different from Google Translate and auto-tune? Over-simplified, it's not significantly different in any way other than how it's controlled. GT will not always produce the same output given the same input, but Vocaloid will because Vocaloid is a VST/Midi instrument. 

 

I feel what's going to happen is people will zero in on the fact that something contains material that is not generated by a human despite the actual creation being created by a human. Like the songs above start as text in google translate, as shown in the video, not as musical compositions that an AI voice was added to.

[HenrySalayne]

1 hour ago, Kisai said:

Stuff like this

https://www.streemtunes.com/post/how-did-pewdiepie-get-a-copyright-strike-for-using-his-own-music

Is why AI generated content will be a problem.

You do know that is an ad?

 

1 hour ago, Kisai said:

Europe has it's own problems regarding copyright, but generally the world follows the US on it, because it's the US strong-arming foreign countries into adopting aggressive copyright laws to protect US content.

I just read into it and the US says exactly the same thing for derivative works as the European copyright laws. ¯\_(ツ)_/¯

 

There is absolutely no issue with the current copyright laws about incorporating AI art into your work. It is still your work and nobody can take it away.

You are basically panicking about a potential abuse of the copyright strike system that is already happening today without any AI content being a part of it. So the only conclusion is to make the current systems more resilient against abuse and not to worry about pieces of AI created works ending up in your works.

[mr moose]

19 hours ago, Kisai said:

That was a different situation where a dodgy formula was stolen and then used by a bunch of different companies.

Nope, the situation is exactly the same, a company misrepresented a product which landed multinational manufacturers in hot water which they had to fix.    It's no different to someone telling you the artwork is copyrightable then finding out after you've gone to market that it wasn't.

 

19 hours ago, Kisai said:

Yes, but that puts it back into this catch-22 situation of "I can't own the copyright on something AI generated, and if it was not disclosed to me, then I can lose the copyright on everything containing it as well."

 

Like, fortunately, I'm not interested in AI generated artwork for producing something commercially, but I might want to produce parody or satire material, and that is already something that artists don't want their name attached to due to the crazyballs twitter cancel culture. Like a lot of foreign artists basically have in their terms and conditions "no politics, no disturbing the peace". So AI generated disposable images fits this, and no artist can whine about not being paid, because they would have never accepted the job.

This is no more a catch 22 than it already is, anyone could sell you artwork that is A, not copyrightable because it is already in the public domain or B, already copyrighted to someone else.  In either situation you are screwed because you used it.  The only thing that has happened now is they have added AI generated art to the list of not copyrightable content.  It is literally just the addition of one more thing to one side of a potential problem, not a new problem.