AI Created images (generated by Midjourney) lose registered copyrights

[YellowJersey]

14 hours ago, Arika S said:

well ignoring the prompt part i train my own models, create hypernetworks and embeddings. all in all i would say i've spent over 100 hours fine tuning my models to get them how i want them to be. is that not sufficient effort?

 

  Reveal hidden contents

i don't actually super care about copyrighting AI art, i'm mainly playing devil's advocate.

 

Fair point. There's also the fact that most of the people creating/interpreting these laws (politicians and judges) are so old that they probably don't understand the technology; not exactly an uncommon problem.

Edit: @Arika S  TBH, the work you describe sounds like it would fall more under patents than copyright. So perhaps the system you create to generate the artwork could be subject to IP protection via patents, but the actual artwork itself would not, at least under current copyright law.

[LAwLz]

On 2/25/2023 at 3:32 PM, Kinda Bottlenecked said:

Kisai's post describe it pretty well. These models aren't context aware yet. Lighting, values, colors, framing, shapes, guiding lines aren't learned in the model. 

The AI isn't aware, but the human at the computer who can control those things are aware.

You are inadvertently agreeing with me.

 

 

 

On 2/25/2023 at 3:32 PM, Kinda Bottlenecked said:

And yet will you ask your 8 year old niece to be a a wedding photographer?

I have seen plenty of people have their children take pictures at their weddings. Sometimes, they turn out good. Most of the time, they turn out bad. Can you guess what that reminds me of? AI Art.

They are not the only ones taking pictures, but they still retain all the copyright of the images they take. So why should AI art be any different?

 

 

I feel like you are grasping at straws right now. What does being a wedding photographer have to do with retaining copyright for images taken with a camera? Taking pictures used to be a really difficult process because you needed to understand the chemical process to develop the film. All of that has been taken care of now by computers. You needed to change the film based on the amount of light in the scene. That is done automatically now. You used to need to set the focus. That is taken care of now. You used to need to set the shutter speed. That is done automatically now.

You can still control those things if you want, but 99% of pictures taken are 99% controlled by computers nowadays. It's just point and shoot. What used to be a process that involved making 20+ decisions is now boiled down to maybe 2 decisions. When to take the picture, and how to point the camera, and even those are somewhat automated these days with things like OIS and burst mode.

 

 

Me not entrusting my wedding to a single 8 year old with a smartphone camera is not a good argument for why photography is somehow still deeply a human skill that requires a lot of effort, especially not when compared to what photography used to be. An 8 year old did not use to be able to take a picture at all. Nowadays they can. The pictures they take might not be good, but how good something is has nothing to do with whether or not someone should get copyright over it.

The fact of the matter is that with the help of technology, more and more of the human has been taken out of the equation when it comes to photography. The whole "a human still needs to compose" is equal to AI art users saying a human is still needed to input a prompt. In both cases a human needs to give inputs which a computer processes. 

 

 

 

On 2/25/2023 at 3:32 PM, Kinda Bottlenecked said:

Yes you can add prost processing to adjust the image after the capture but it also does not manipulate the overall composition of the scene for you. 

I don't think you understand how much "post processing" as well as preprocessing and just general processing goes on in a modern camera. Post processing is not what I am taking about when I say a camera changes a lot about the image. The raw data that a camera captures, before any processing is done, is simply not something humans can even interpret. Without a metric ton of processing that happens in a camera, you would not even be able to see the picture taken with a modern camera. You wouldn't even realize it's a picture you're looking at.

 

 

On 2/25/2023 at 3:32 PM, Kinda Bottlenecked said:

Can the camera give instructions to your subjects in your scene? It does not.

Does an AI art program give you instructions on how to change the subject in your scene? Nope, it doesn't.

 

 

 

 

On 2/25/2023 at 3:32 PM, Kinda Bottlenecked said:

I think the overall issue in this case is that we have a human assisted machine creation rather than a machine assisted human creation. None of which are alike the examples you've given. 

I think the overall issue is that you have already decided that AI art generation programs are "human assisted machine creation" and cameras are "machine assented human creation", yet you have no logical or rational explanation for how they are different. You have just decided that one is the former and another one is the latter.

 

If you disagree that you have drawn arbitrary lines then I'd like for you to describe exactly what the differences are. To me, taking a picture with a camera is just as much "human assisted machine creation" as using an AI art program is.

In both cases, the human provides varying degree of guidance to a computer which then generates an image. I

 

To me they are the same thing, especially when talking about modern cameras where a human is 100% dependent on the camera doing a ton of processing to get anything usable out of it. I don't think it's fair to say something is "assisted" if a human just has to make one decision (how to position the camera) and then press a button, while the camera itself has to execute millions of lines of code to automatically determine tens if not hundreds of settings and do several stages of processing.

 

Why is AI art generation "human assisted machine creation" while a camera is "machine assisted human creation"?

[WildDagwood]

Mentioned this in another thread:

 

"Art in it's most basic form, is a method of expression. What are you expressing? Your "unique", internal thoughts and emotions as they manifest in your imagination. If everything your providing is just the product of others peoples deconstructed work, filtered down to your parameters, you can't really claim any credit, imo. Won't even go into the matter of the time, effort, and skill it requires to actually hone those skills that are being used for training these AI, since that's another matter, and a big F U to some of these artists."

 

I agree there's a lot of subjective lines that can be drawn, but for me, it's somewhere around unique, artistic input, and how much of the project was generated with direct vs. passive influence. Problem is you'll have a hard time getting people to agree on what constitutes unique, artistic input, but I think it might be a bit "easier" (not entirely) to draw a line on what's direct vs. passive.

[Kinda Bottlenecked]

1 hour ago, LAwLz said:

Why is AI art generation "human assisted machine creation" while a camera is "machine assisted human creation"?

 

It is the name "AI art generation". 

 

Thanks for the lengthy reply but I do feel that you are also agreeing with me slightly. 

 

1 hour ago, LAwLz said:

Does an AI art program give you instructions on how to change the subject in your scene? Nope, it doesn't.

 

This very phrase would agree with me because both you and I know that the AI does not give you any instructions, it does it for you. The current law would say that one has to be the mastermind of the scene. Which unfortunately for this specific case the author of the comic did not achieve. 

 

1 hour ago, LAwLz said:

I think the overall issue is that you have already decided that AI art generation programs are "human assisted machine creation" and cameras are "machine assented human creation", yet you have no logical or rational explanation for how they are different. You have just decided that one is the former and another one is the latter.

 

If you disagree that you have drawn arbitrary lines then I'd like for you to describe exactly what the differences are. To me, taking a picture with a camera is just as much "human assisted machine creation" as using an AI art program is.

In both cases, the human provides varying degree of guidance to a computer which then generates an image. I

 

To me they are the same thing, especially when talking about modern cameras where a human is 100% dependent on the camera doing a ton of processing to get anything usable out of it. I don't think it's fair to say something is "assisted" if a human just has to make one decision (how to position the camera) and then press a button, while the camera itself has to execute millions of lines of code to automatically determine tens if not hundreds of settings and do several stages of processing.

 

Why is AI art generation "human assisted machine creation" while a camera is "machine assisted human creation"?

 

There is a slight misunderstanding here, what I meant by "machine assisted human creation" can also involve an AI, so if one uses SD or something else in the pipeline to achieve a final piece that is a derivative of the SD generation this would be counted as a "machine assisted human creation" because I am not bound by set parameters and I am working "upwards" to build a final composition.

 

When purely using AI art generation you're working "down" from the millions of images that the model has been trained on to then choose the one single image from the seed that you fancy the most. This an entirely difference process altogether.  

 

When you point and shoot, you're the one deciding your subject in frame and you're the one in charge of the composition.

 

1 hour ago, LAwLz said:

You have just decided that one is the former and another one is the latter.

 

No I have not. Which is why I specify that in this case where the author has done next to nothing to transform or alter the AI artwork is more of a "human assisted machine creation". If you're looking carefully at my posts you would realize that the way humans are interfacing with AI art generating model currently is too limiting and does not ensure that the user is the one masterminding the actual composition is what I'm driving at.

 

What you have done here is decide that for me, which is exactly what all the AI art generation models do now. 

[Kisai]

42 minutes ago, WildDagwood said:

 

 

I agree there's a lot of subjective lines that can be drawn, but for me, it's somewhere around unique, artistic input, and how much of the project was generated with direct vs. passive influence. Problem is you'll have a hard time getting people to agree on what constitutes unique, artistic input, but I think it might be a bit "easier" (not entirely) to draw a line on what's direct vs. passive.

 

The problem in the Art AI generators is that there's no "artistic input". The original artist when a piece was originally made, when it was actually "art", had an artistic input. Even photographic works. The AI however isn't picking the most "artistic" or even equal works to draw from. It's picking the most weighted reference point for a pixel, or subsample or whatever the AI decided to associate.

 

When the AI training sees 10,000 photos, artworks and random scribbles of "dog", it doesn't know which of these are by the same artist, it doesn't know an artistic style. It doesn't even know what a "dog" is. It can only differentiate these because of other keywords found in the AI. That's why we keep seeing totally messed up hands and fingers in AI art generators. That information is not tagged, and thus the AI is learning "this is (keyword)" and every time (keyword) appears these symbols also appear, of which it seems to understand "digits" but not how many there should be. Yet some how understands that a dog and a human have two eyes... usually.

 

Every time I see people go "wow, this AI is soo good at producing [Sakimi-chan] level artwork", I see the same glaring errors. Sakimi-chan is an extremely popular artist that has a very distinct style that many other good artists are accused of ripping off. 

 

It always comes back to people looking at these AI generated works and some how being amazed at the illusion. These are the same people fooled by stage "magicians". If you've never seen the trick before, you might be fooled, once. But see the same trick over and over, and you see through the illusion.

 

When it comes to copyright, at best, a copyright can not, and should not be held on a derivative or a transformative work if it can be traced back to the original work. Since a work can be "duplicated" by re-entering the same prompt to the same AI, there is no reason to permit copyright, since two different people might enter the same keywords, by chance, and come up with the same image. 

 

AI generated artwork, should always be tagged "AI Generated Artwork", and if a regulatory approach is used, the model, keywords and any other "input" given to the AI should be used, much in the same way we go about patents. "Here's the prior art"

 

 

[WildDagwood]

9 minutes ago, Kisai said:

 

The problem in the Art AI generators is that there's no "artistic input". The original artist when a piece was originally made, when it was actually "art", had an artistic input. Even photographic works. The AI however isn't picking the most "artistic" or even equal works to draw from. It's picking the most weighted reference point for a pixel, or subsample or whatever the AI decided to associate.

Pretty much agree with everything you said.

 

I just know there's people that will argue that spitting out ideas/parameters etc. and having an A.I. take it from there is still unique input, and utilizing a tool to produce something new. I don't personally agree with it, but they're out there.

[05032-Mendicant-Bias]

2 hours ago, Kisai said:

AI generated artwork, should always be tagged "AI Generated Artwork",

So also outputs of photoshops where ML brushes were used should use the label "AI Generated Work"?

What about smartphone camera with ML filters?

58 minutes ago, Lightwreather JfromN said:

I sincerely doubt this is the case. I myself have inputted the exact same prompt into the same AI based on the same model with wildly different results. Note: The prompt was quite detailed, I forgot what exactly it was but I do know it was at least 30-50 words.

With DallE eand Midjourney you never get the same image twice because it always uses a new seed.

With more advanced tools like StabilityAI (that you can run locally on your GPU), you can keep the previous seed and get the same result.

Stability AI has learned to allucinate images from noise and a prompt, if the noise is the same, the allucination persists.

 

It's useful to keep the seed static when you want to control variables, so see exactly the effect of a changed word in the prompt results with the same seed.

[Holmes108]

45 minutes ago, WildDagwood said:

Pretty much agree with everything you said.

 

I just know there's people that will argue that spitting out ideas/parameters etc. and having an A.I. take it from there is still unique input, and utilizing a tool to produce something new. I don't personally agree with it, but they're out there.

 

Yep, the definition of "art" for me, broadly anyways, is much more about the output than the input. Which is where a lot of these disagreements may be stemming from. This is more about results, products, and business to me. For other's it seems to be more about artistic integrity.

[Lightwreather]

1 hour ago, Kisai said:

Since a work can be "duplicated" by re-entering the same prompt to the same AI

I sincerely doubt this is the case. I myself have inputted the exact same prompt into the same AI based on the same model with wildly different results. Note: The prompt was quite detailed, I forgot what exactly it was but I do know it was at least 30-50 words.

[Kisai]

1 hour ago, Lightwreather JfromN said:

I sincerely doubt this is the case. I myself have inputted the exact same prompt into the same AI based on the same model with wildly different results. Note: The prompt was quite detailed, I forgot what exactly it was but I do know it was at least 30-50 words.

The part people ignore is the random seed. If I enter the exact same prompt with the same model and seed, it will ONLY produce the exact same image, because it makes those connections identically.  If you have no control over the seed, then getting the AI to induce the exact same image by chance is harder, but not non-zero.

 

That said, the output matters. If you present an image as being "real" without disclosing alterations to it, you set yourself up for failure later when you can't replicate it.

https://www.usatoday.com/story/college/2014/07/02/students-blast-photoshop-for-altering-beauty-standards-body-image-norms/37392597/

 

Photo: Anna Hill

Quote

Bottom line, as she writes, "it is not the Photoshopped female who is our enemy; we do a good enough job destroying ourselves on our own."

 

With photoshopping, it creates these false beauty standards that people mistakenly believe are real, and it fuels everything from plastic surgery to gender identity politics. "You're not a real (woman/man) unless you look like this"

 

AI photoshopping can take this to the logical extreme, identifying things we deem imperfect in an input photo, and do all this automatically. But the initial photo is still taken by a human, somewhere in the process, and if the subject's eyes aren't open, it can't magically open them without knowing what the eyes look like. 

https://www.nvidia.com/en-us/geforce/news/jan-2023-nvidia-broadcast-update/

Quote

Notice how the guy's eyelashes pretty much disappear.

 

I'm a bit less concerned about people "stealing other artists work" via laundering it through an AI, be that AI generated artwork, music, speech, singing, etc. It will never produce something that is 100% foolproof, it can not produce something with the fidelity, impact, thought, composition, etc. It's at best a poor imitation of the source material, and the sooner we move on from trying to clone an existing living or dead artist the better. It has it's purposes, where it can bring back someone from the dead, or allow artists to to continue to use a style/voice they no longer are able to (eg accessibility reasons) due to medical reasons, physical reasons, or even just time constraints. 

 

Like if you're a really in high demand artist, you can't just make 100 versions of yourself, but you can have an AI train off your work, do about 60-80% of it, and let humans "clean up" the rest and still call it your work, and be paid for this use of your work.

 

But again, there are people out there, arguing in bad faith that they should be able to use these "general" models to reproduce specific artists styles, but are putting way, waaaay, too much faith in the AI being able to eventually "learn" it's way out of plagiarism. It won't. It won't even try. 

 

What I feel is going to happen is that better models will come along that purge some of the input sources that are mislabeled, and add more diversity to the actual art, to reduce certain styles from being overfit. As you may have noticed, many of the "art" generated by midjourney, stable diffusion, and dalle2 have a similar digital painted style.

 

[LAwLz]

5 hours ago, Kinda Bottlenecked said:

It is the name "AI art generation". 

 

Thanks for the lengthy reply but I do feel that you are also agreeing with me slightly. 

So if we called photos "camera generated images" then that would also be classified as "human assisted machine created"?

Do I understand you correctly. Your argument for why AI art generator is "human assisted machine created" is because of the name? Do you not have any other logic for why you make the distinction other than the name?

 

 

  

5 hours ago, Kinda Bottlenecked said:

This very phrase would agree with me because both you and I know that the AI does not give you any instructions, it does it for you. The current law would say that one has to be the mastermind of the scene. Which unfortunately for this specific case the author of the comic did not achieve. 

What exactly does the AI do for me that a camera doesn't do for a photographer?

Also, where in the law does it say "one has to be the mastermind of the scene"? Can you give me a citation for that particular part of the law?

 

 

  

5 hours ago, Kinda Bottlenecked said:

There is a slight misunderstanding here, what I meant by "machine assisted human creation" can also involve an AI, so if one uses SD or something else in the pipeline to achieve a final piece that is a derivative of the SD generation this would be counted as a "machine assisted human creation" because I am not bound by set parameters and I am working "upwards" to build a final composition.

 

When purely using AI art generation you're working "down" from the millions of images that the model has been trained on to then choose the one single image from the seed that you fancy the most. This an entirely difference process altogether.  

 

When you point and shoot, you're the one deciding your subject in frame and you're the one in charge of the composition.

You are avoiding the question.

What makes a picture taken by a camera "human created, machine assisted" and what makes a picture created with an AI program "machine created, human assisted"?

 

Can you quantify it in any logical and objective way?

 

 

Yes, when you point and shoot you're the one deciding the subject in frame and you're the one in charge of the composition.

But that is true for AI art generation as well because you are the one deciding the subject (with the prompt you input and settings you change, just like with a camera you change the input by moving it around).

 

Let me ask you this, have you ever actually used Stable Diffusion? Because it sounds like you don't even know what it is or how it works. It sounds like you are making a bunch of assumptions that are all wrong, and that's why you don't understand why it is very similar to taking a picture with a camera.

 

In what way, according to you, is inputting a prompt into a program like Stable Diffusion different from pointing a camera at a subject and clicking the shutter button? 

 

 

  

5 hours ago, Kinda Bottlenecked said:

No I have not. Which is why I specify that in this case where the author has done next to nothing to transform or alter the AI artwork is more of a "human assisted machine creation". If you're looking carefully at my posts you would realize that the way humans are interfacing with AI art generating model currently is too limiting and does not ensure that the user is the one masterminding the actual composition is what I'm driving at.

If you looked carefully at how cameras work you would realize that the way humans are interfacing with cameras currently is too limited and does not ensure that the user is the one masterminding the actual composition.

 

See, I can make wild claims too that are not based in any objective truth. What I am asking you is to define what you actually mean in actual, measurable terms.

How do you measure and define where something goes from being "machine created, human assisted" into "human created, machine assisted"? Because you seem very set on defining what goes into which category, so I want you to explain how to measure that.

[LAwLz]

5 hours ago, Holmes108 said:

Yep, the definition of "art" for me, broadly anyways, is much more about the output than the input. Which is where a lot of these disagreements may be stemming from. This is more about results, products, and business to me. For other's it seems to be more about artistic integrity.

Same here. Art to me is about expressing ideas and/or feelings. It's to me more about the result rather than the method. That's why so many different things can be called art even though the process are widely different. A photograph can be art just like a painting can be art, or even acting and dancing. All of them have very different methods but all of them can express ideas or emotions, and that's why they are art.

 

If someone thinks of something, and then have that idea or feeling transferred to someone else who cares about the nitty gritty details of how it was done?

I am sure the people who used to be the only ones who had access to a particular way of expression care that said type of expression is now more widely available, because it makes them feel less special, but other than those assholes who cares?

[Arika]

3 hours ago, LAwLz said:

Same here. Art to me is about expressing ideas and/or feelings. It's to me more about the result rather than the method. That's why so many different things can be called art even though the process are widely different. A photograph can be art just like a painting can be art, or even acting and dancing. All of them have very different methods but all of them can express ideas or emotions, and that's why they are art.

 

If someone thinks of something, and then have that idea or feeling transferred to someone else who cares about the nitty gritty details of how it was done?

I am sure the people who used to be the only ones who had access to a particular way of expression care that said type of expression is now more widely available, because it makes them feel less special, but other than those assholes who cares?

what i find most amusing is when AI art is presented as actual art to a group of people (or to a competition) and people start analyzing it and saying how good it and and praise the composition and overall aesthetics....but then they find out it's AI generated and suddenly it's no longer art and find a way to back pedal or not acknowledge that they did actually like it at first.

 

it is a dick move to present AI art as """"real"""" art but it is amusing to see the before and after reactions.

[Kisai]

47 minutes ago, Arika S said:

what i find most amusing is when AI art is presented as actual art to a group of people (or to a competition) and people start analyzing it and saying how good it and and praise the composition and overall aesthetics....but then they find out it's AI generated and suddenly it's no longer art and find a way to back pedal or not acknowledge that they did actually like it at first.

 

it is a dick move to present AI art as """"real"""" art but it is amusing to see the before and after reactions.

Hypocrisy isn't exclusive to rich billionaires and politicians.

 

My personal opinion is that if someone fails to label the tools used to make a thing, like you would with a patent, then you are not entitled to call it that thing. eg, Almond/Soy Milk is not Dairy Milk, but somehow we allow both to be called milk, despite the two things are extremely far apart, nutritionally, ethically, and in sustainability (really, almond milk is even less sustainable than dairy milk.) Thus we have to make distinguishing labels for the thing that originally was the original product.

 

So with AI Art generators, if people keep mislabeling it as "Artwork" in the same space as human artists, we're going to have to start demanding human artists show WIP's and their tools just to avoid accusations of being an AI (which is ALSO something that Sakimi-chan has been accused of being.)

 

[Kinda Bottlenecked]

6 hours ago, LAwLz said:

So if we called photos "camera generated images" then that would also be classified as "human assisted machine created"?

Do I understand you correctly. Your argument for why AI art generator is "human assisted machine created" is because of the name? Do you not have any other logic for why you make the distinction other than the name?

 

If there is a camera that generates images then yes but we haven't had that yet and I don't think we would need to have. My argument for "human assisted machine creation" isn't solely in the name "ai art generation" as evident from the further explanation I gave to you in my previous reply. I was referring to the name in itself as a good descriptor of the overall process of ai art that makes it so different from "machine assisted human generation". 

 

6 hours ago, LAwLz said:

What exactly does the AI do for me that a camera doesn't do for a photographer?

Also, where in the law does it say "one has to be the mastermind of the scene"? Can you give me a citation for that particular part of the law?

 

Simple, it can't compose the scene for you. Although, my use of the word "law" might have been misleading here since as I understand that in this case this is the current  interpretation of the copyright laws that the courts have.

 

 

6 hours ago, LAwLz said:

You are avoiding the question.

What makes a picture taken by a camera "human created, machine assisted" and what makes a picture created with an AI program "machine created, human assisted"?

 

Can you quantify it in any logical and objective way?

 

 

Yes, when you point and shoot you're the one deciding the subject in frame and you're the one in charge of the composition.

But that is true for AI art generation as well because you are the one deciding the subject (with the prompt you input and settings you change, just like with a camera you change the input by moving it around).

 

Let me ask you this, have you ever actually used Stable Diffusion? Because it sounds like you don't even know what it is or how it works. It sounds like you are making a bunch of assumptions that are all wrong, and that's why you don't understand why it is very similar to taking a picture with a camera.

 

In what way, according to you, is inputting a prompt into a program like Stable Diffusion different from pointing a camera at a subject and clicking the shutter button? 

 

And you are avoiding my answer. As I've said, the whole building up to a composition is different from working "down" to select the best image from a seed. 

 

Unfortunately for you I have dabbled ai image generation including other derivatives of SD and custom models. I am also familiar with img2img and openpose. If this is your attempt at driving a line between us then it is futile because I am also a creative that also primarily works in 3d. Now I do know that working on image generation can be as lengthy to get the image that you want and I am not going to deny that but at the end of the day this is my understanding of both these workflows. Your questions have been answered in the previous reply and if you found them dissatisfactory then it is what it is. 

 

 

 

6 hours ago, LAwLz said:

If you looked carefully at how cameras work you would realize that the way humans are interfacing with cameras currently is too limited and does not ensure that the user is the one masterminding the actual composition.

 

See, I can make wild claims too that are not based in any objective truth. What I am asking you is to define what you actually mean in actual, measurable terms.

How do you measure and define where something goes from being "machine created, human assisted" into "human created, machine assisted"? Because you seem very set on defining what goes into which category, so I want you to explain how to measure that.

 

You seem very set on on misconstruing my replies because they are not wild at all. The way that we can interface with most of the tech right now is limiting. 

[Arika]

1 hour ago, Kinda Bottlenecked said:

My argument for "human assisted machine creation" isn't solely in the name "ai art generation" as evident from the further explanation I gave to you in my previous reply. I was referring to the name in itself as a good descriptor of the overall process of ai art that makes it so different from "machine assisted human generation". 

well since you bought up Kisai's response, i would like to know what your answer to a question they posed:

 

Quote

what is the minimum threshold of human input required?

because you also went on to say:

 

1 hour ago, Kinda Bottlenecked said:

Now I do know that working on image generation can be as lengthy to get the image that you want and I am not going to deny that

 

Where exactly is your line? because it's super unclear.

 

lets go with 2 extremes to start off with:

 

Would a child picking up a camera and clicking the button on accident be copyrightable if they just pressed a button and there was exactly 0 thought into composition and framing because they are a child?

 

and vice versa

 

Would someone who spends hours testing and curating prompts and settings, training their own models, creating personal embeddings and hypernetworks to get the perfect outputs be copyrightable?

[LAwLz]

4 hours ago, Kinda Bottlenecked said:

If there is a camera that generates images then yes but we haven't had that yet and I don't think we would need to have.

Can you please define "generates"? Because I would argue that a camera generates an image for me.

I would have no idea how to take the sensor reading and then converting that into a JPEG if the camera didn't do that for me. Is that not the camera "generating an image"? It generates it based on the input, but so does a program like Stable Diffusion.

 

 

4 hours ago, Kinda Bottlenecked said:

My argument for "human assisted machine creation" isn't solely in the name "ai art generation" as evident from the further explanation I gave to you in my previous reply. I was referring to the name in itself as a good descriptor of the overall process of ai art that makes it so different from "machine assisted human generation". 

I think you were being very vague when you "further explained" your point and you didn't answer my questions.

Where do you draw the line and how do you measure the difference between "human assisted, computer created" and "human created, computer assisted"?

 

 

4 hours ago, Kinda Bottlenecked said:

Simple, it can't compose the scene for you.

Can you elaborate on what you mean?

Because I would argue that you can compose a scene using an AI art program as well. I can create prompts that results in images that are pretty close to what I pictured in my head. If the argument is that it isn't identical to what I pictured in my head then I would raise the issue of cameras not producing outputs that match the viewfinder either. When you take a picture with a camera, it will not look like you think it would look. It will be fairly close, but if we were to start examining the color values of the pixels very few if any would be what the photographer actually intended. But it will be close enough that we have oversight with it.

 

I would also argue that you can take pictures with a camera without composing a scene, or at the very least that you can take a picture without putting any thought into the scene's composition. For example taking a picture without looking at the camera or viewfinder, so you have no idea what the picture will look like.

 

Do you not agree that I can compose a scene with a program like Stable Diffusion?

Do you not agree that a photo can be taken without composing a scene?

 

If you agree with both of them, then "composing a scene" can not be the differentiator between "human created, machine assisted" and "machine assisted, human created".

 

 

Is a blind person taking a picture with a camera, without even knowing where they are or what they are taking a picture of, still a "human composing a scene"?

 

 

 

4 hours ago, Kinda Bottlenecked said:

If this is your attempt at driving a line between us then it is futile because I am also a creative that also primarily works in 3d.

I think you meant to say that you are letting a computer generate 3D models for you. I mean, you are not doing the processing yourself right? Nor are you creating the output. You are commissioning a computer program to generate that for you. 

 

 

I am somewhat kidding, but also partially serious. What you are doing can be described as commissioning a computer to generate models for you. If we look up each of the words in a dictionary they will accurately describe what you are doing. It's just that we typically don't use those words to describe 3D modeling and for some reason we happened to use those words when talking about 3D art. That's why I am trying to get you to define and elaborate on certain words and to use other words as well, because right now it feels like the conversation is basically:

 

"I am not cooking food, I am making dinner"

and when asked what the difference is you go "one is cooking and the other is making, and one is food and the other is dinner".

I feel like you're just regurgitating certain words and phrases without elaborating or making distinctions between them. It just becomes circular reasoning.

[mr moose]

On 2/26/2023 at 3:55 AM, wanderingfool2 said:

 

Don't conflate what YouTube does with copyright law.  YouTube does the over-sensitive approach where they assume things sounding similar (not even exact) will be copyright infringement due to not wanting to get sued/tied up in lawsuits.

 

Ain't that the truth, I uploaded a video with Bach in it a few years back,  youtube stopped the upload and actually asked me which (of the 4 symphonies it thought produced it) it came from.  It would not let me upload unless I picked one and let them monetize it and send royalties to some US orchestra..  Forget the fact it was none of them playing it. Load of bollocks their system is. 

 

On 3/1/2023 at 12:26 PM, Arika S said:

what i find most amusing is when AI art is presented as actual art to a group of people (or to a competition) and people start analyzing it and saying how good it and and praise the composition and overall aesthetics....but then they find out it's AI generated and suddenly it's no longer art and find a way to back pedal or not acknowledge that they did actually like it at first.

 

it is a dick move to present AI art as """"real"""" art but it is amusing to see the before and after reactions.

This happens all the time in audio,  self proclaimed "audiophiles" have now created an argument as to why double blind tests mean nothing..  I guess they were tired of being proven the fools.

[Sauron]

On 2/23/2023 at 2:41 PM, LAwLz said:

Why is it good that AI generated art can't be copyrighted? 

Because copyright is bad. Shouldn't be limited to just AI though

[Isuck Assimov]

All AI generated art that comes from models trained on datasets which are not 100% unique and created for the sole purpose of training AI, should be public domain.

Anyone trying anything else has zero interest in advancing the technology and just wants to stash cash in their pockets with other peoples hard work



 

[Sauron]

2 hours ago, Assimov said:

All AI generated art that comes from models trained on datasets which are not 100% unique and created for the sole purpose of training AI, should be public domain.

I think everything should be public domain but are you going to argue that, say, painters don't draw direct inspiration from existing works?

[Kinda Bottlenecked]

On 3/1/2023 at 4:07 PM, LAwLz said:

Can you please define "generates"? Because I would argue that a camera generates an image for me.

I would have no idea how to take the sensor reading and then converting that into a JPEG if the camera didn't do that for me. Is that not the camera "generating an image"? It generates it based on the input, but so does a program like Stable Diffusion.

 

Its interesting that you'd think that way when you've explained how a camera works by capturing the light entering its sensors instead of generating the scene. By that logic, the sun should hold all copyright in the world because its light rays render the world around us 

 

On 3/1/2023 at 4:07 PM, LAwLz said:

I think you were being very vague when you "further explained" your point and you didn't answer my questions.

Where do you draw the line and how do you measure the difference between "human assisted, computer created" and "human created, computer assisted"?

 

There was no further explanation of my point, it was just a regurgitation of the point itself. You can refer to the the earliest post again for the explanation again if you wish to, beyond that I don't see how else I can entertain you. 

 

On 3/1/2023 at 4:07 PM, LAwLz said:

Can you elaborate on what you mean?

Because I would argue that you can compose a scene using an AI art program as well. I can create prompts that results in images that are pretty close to what I pictured in my head. If the argument is that it isn't identical to what I pictured in my head then I would raise the issue of cameras not producing outputs that match the viewfinder either. When you take a picture with a camera, it will not look like you think it would look. It will be fairly close, but if we were to start examining the color values of the pixels very few if any would be what the photographer actually intended. But it will be close enough that we have oversight with it.

 

I would also argue that you can take pictures with a camera without composing a scene, or at the very least that you can take a picture without putting any thought into the scene's composition. For example taking a picture without looking at the camera or viewfinder, so you have no idea what the picture will look like.

 

Do you not agree that I can compose a scene with a program like Stable Diffusion?

Do you not agree that a photo can be taken without composing a scene?

 

If you agree with both of them, then "composing a scene" can not be the differentiator between "human created, machine assisted" and "machine assisted, human created".

 

 

Is a blind person taking a picture with a camera, without even knowing where they are or what they are taking a picture of, still a "human composing a scene"?

 

Although this section is lengthy, I've explained both the process of  building up and working down towards something in my earlier replies to you. Although yes you can definitely snap pictures of subject quickly there is always some basic composition needed to take one. Its funny you bring up the blind person because I feel just as blind working with the AI whenever it shows me a collage of variants of the image even though I know some of the technicalities behind it 

 

On 3/1/2023 at 4:07 PM, LAwLz said:

Do you not agree that I can compose a scene with a program like Stable Diffusion?

Do you not agree that a photo can be taken without composing a scene?

 

Can you compose a scene with the help of SD? You certainly can. However the issues in this case arise when you let the AI take the wheel isn't it? You don't say you drove when you set a waypoint on uber. You say you were driven. 

 

On 3/1/2023 at 4:07 PM, LAwLz said:

I think you meant to say that you are letting a computer generate 3D models for you. I mean, you are not doing the processing yourself right? Nor are you creating the output. You are commissioning a computer program to generate that for you. 

 

 

I am somewhat kidding, but also partially serious. What you are doing can be described as commissioning a computer to generate models for you. If we look up each of the words in a dictionary they will accurately describe what you are doing. It's just that we typically don't use those words to describe 3D modeling and for some reason we happened to use those words when talking about 3D art. That's why I am trying to get you to define and elaborate on certain words and to use other words as well, because right now it feels like the conversation is basically:

 

"I am not cooking food, I am making dinner"

and when asked what the difference is you go "one is cooking and the other is making, and one is food and the other is dinner".

I feel like you're just regurgitating certain words and phrases without elaborating or making distinctions between them. It just becomes circular reasoning.

 

And by that logic I am not writing my replies to you the poster, I am writing to your computer because you did not construct the post bit by bit. You commissioned the computer to do it for you  Although I'm curious which dictionary are you looking up because I can't seem to find those that support your claim in its description of the process. 

 

I also do feel similarly about this conversation because it has essentially become this "You have answered it but can you answer it again?". 

 

On 3/1/2023 at 12:41 PM, Arika S said:

Where exactly is your line? because it's super unclear.

 

My personal line? Or the line the courts drew for this case? 

 

My personal line would be that AI generated images used in its entirety without significant modifications cannot not be copyrighted. Which make sense, I can't buy a license for an image and copyright that. Even if the license allows it for monetary purposes. 

[Isuck Assimov]

58 minutes ago, Sauron said:

I think everything should be public domain but are you going to argue that, say, painters don't draw direct inspiration from existing works?

"Drawing inspiration" is a human trait. A computer program doesn't "get inspired", even with deep learning. This would require a level of self awareness which machines will not reach in our lifetimes and if they did would have a whole plethora of ethical implications. I don't think we're even equipped to actually design tests that could suggest or prove human level self awareness in AI.

You seem to imply that human painters and AI are pretty much the same thing with just minimal differences. Either you are overevaluating AI or severly underevaluating human artists.

 

[Mihle]

On 3/2/2023 at 12:18 PM, Sauron said:

Because copyright is bad. Shouldn't be limited to just AI though

If I lived in a world without copyright, I wouldn't bother making things on my own and a lot of other people wouldn't either.

[Mihle]

I know it's not exactly what this is about but:

If the AI itself had the copyright, (it shouldn't) where do you draw the line where it's rights stop compared to humans?