Zarya images "not protected by copyright." Words and arrangement remain protected.
See full article...
See full article...
As a thought experiment, how much AI do you thing Disney would rely on if it puts a cloud over their content?
Does anyone know what, if anything, there is for case law on generative works in general? Say I write a program to draw a flower with turtle graphics, do I own the copyright for that image?
I used midjourney to generate what is (to me) a stunning and compelling image, in a few minutes, and it's now the wallpaper on my phone and I love it. Do I think it should be copyrightable? Hell no! The prompt was only about eight words, I tried a few variations and then got something that clicked, and there's no way I should have any sort of control over its use (or that someone else who generated a near identical image should be able to control my use of it).
If AI generated images of this sort could be copyrighted, in my opinion it would be a field day for copyright trolls to claim ownership of vast amounts of imagery by using common text prompt combinations in an entirely mechanical way.
I want to see a significant amount of human creativity involved before copyright applies (and I realise "significant amount" is very hard to define, but hey, that's what the courts are for).
But you do if you take a picture of that seashell in the beach. You don't even have to take the seashell home.
This is easy. If you publish it, you. If someone else publishes it, them.If the humans' inputs aren't significative enough to be granted copyright, are the humans' inputs also not significative enough if I want to create something that resembles a previously copyrighted material?
If I ask stable diffusion to draw Mickey Mouse, who is violating Disney copyright?
That makes sense to me, too. I saw Meshuggah's video for "They Move Below", done by Wizardhead, and it seems like a lot of work went into that, even though not every frame was drawn in the normal way. I would expect one or both of those entities to be able to claim copyright on it.
View: https://www.youtube.com/watch?v=c6Q1v4zRt1w
Hard disagree. "Throwing paint at a canvas" is only one small aspect of the whole process. What colour to throw, where to throw the paint at what part of the canvas, how hard to throw the paint, how much to throw at any given time, etc.The Copyright Office's letter repeatedly emphasizes the degree of creative control that the artist exercises over the artwork's composition. As I understand their analysis, throwing paint at a canvas involves very little creative control over the actual composition of the art (even if there may be other creative elements), so it's a relevant analogy.
ONLY if both have an agreed term that the copyright transfers that meets the statutory requirements. Even if both parties agree that it was a work for hire, there must be a written contract stating this. Just because someone works for you fulltime is not enough.What are you talking about? Your own Wikipedia link directly contradicts your claim. A full-time employee (as ewelch describes) is not considered the author of a work for the purposes of copyright, when that work is created within the scope of their employment.
What color paint? How large a brush? How much do you load the brush? How big a canvas?The Copyright Office's letter repeatedly emphasizes the degree of creative control that the artist exercises over the artwork's composition. As I understand their analysis, throwing paint at a canvas involves very little creative control over the actual composition of the art (even if there may be other creative elements), so it's a relevant analogy.
If several independent people input identical prompts each using the same generative tool, which then produces identical outputs, should they all claim independent copyright on the work products? Each will be unaware of others' work products. Seems unlikely to hold on that basis. However, if they each then amend common work products rendering them unique, a good question is will they have any ownership of the derivative work.
This sounds like a fun test case for the Supreme Court.
If a person was to look at hundreds of images, and then paint something inspired by the images they reviewed, why is that so different than having AI perform that step? Isn't the end result the same? Images have been reviewed, and something was created. If the images in the article aren't copyrightable, is the copyright office saying it's ok for an organic brain to review and paint but, that same organic brain, cannot use AI to perform the same task? And where does the Copyright office get the authority to make that decision?
And, where is the difference between telling Photoshop to perform some task to an image using AI vs. Midjourney doing a bunch of the work directly?
Artists sue other artists for copyright infringement. If a work is significantly similar to pre-existing art, then the plaintiff can win the suit. The same rules should apply with AI assisted art.
The challenge with AI is that we don't know there the prior art comes from and what percentage of the prior art was used in the generation of the images.
I suppose there's some legal threshold/percentage of where a work becomes copyright infringement. The AI could be trained to use prior art below that threshold to avoid copyright infringement. I suppose it could even print a report as to how the image was generated.
ONLY if both have an agreed term that the copyright transfers that meets the statutory requirements. Even if both parties agree that it was a work for hire, there must be a written contract stating this. Just because someone works for you fulltime is not enough.
Even my link says that generally copyright goes to the person doing the work by default UNLESS there is a specific exception. And in the work for hire case, it's that there is a contract spelling that out to meet the statutory requirements.
The opposite is true as well, I can hire someone even temporarily, to create a work and I get the copyright transfered to me.
In the case of a work made for hire, the employer or other person for whom the work was prepared is considered the author for purposes of this title, and, unless the parties have expressly agreed otherwise in a written instrument signed by them, owns all of the rights comprised in the copyright.
I don't think there will ever be 100% no-gray-area answers to these kinds of questions. But I don't think it hurts to consider the large amounts of black and white on the opposite sides of that gray area as basically solved.To further this idea, if Photoshop had the ability to take text inputs in natural language (draw a fine line between the head and the hand, etc), would that work be copyrightable?
Or to use a more real example, using Context-Aware fill, is the part of the image that was edited with Context-Aware fill copyrightable? Or the Healing Brush, or any of the other automatic tools in Photoshop
This definitely adds a twist to business that are going to depend on AI generations. The thing I don't understand is what the line will be on when the artist's work become predominate. The guy in this article for example: https://meincmagazine.com/information...has-a-confession-his-photos-are-ai-generated/ -- he heavily retouched the photos he generated. Is that enough to earn copyright? If not, where is the line? I think the Copyright office has made a somewhat sensible ruling (not what I expected) but I think they have opened a gigantic can of worms now about where copyright starts and stops for this technology.Personally, I agree that copyrights should not be issued for AI-generated images. As for stories, that's more problematic, unless one wants to be pedantic and say that machines can't hold copyrights, and someone who generated the story is not the holder of the copyright, either.
Then we get into the "who owns the copyright, the camera's owner or the monkey who took the picture?" question.
AI's aren't entities who can own copyrights, but they're also "cameras" that can generate images based on data from other sources.
I wouldn't be at all upset if it's decided that nothing AI generated can be copyrighted. That would, at least, take all the wind out of the sales of AI-generated art and probably please the hell out of the artists and authors who actually work for a living.
Her creative process ended with the text prompt. That could be copyrighted assuming it meets the required threshold for creative work (i.e. isn't just "picture of a girl"). As the law stands, if you gave a similar prompt to a human artist, they would own the copyright (which would transfer if a work for hire). But a machine generated art does not currently qualify for a copyright, which means there is no copyright for her to acquire.I realize the issues with copyright for AI generated (or supported) images, but fundamentally she engaged in a creative process. That seems like it should be copyrightable...
It comes down to two points:Benj Edwards said:It's possible that the ruling may eventually be reconsidered as the result of a cultural shift in how society perceives AI-generated art—one that may allow for a new interpretation by different members of the US Copyright Office in the decade ahead.
If it's simply "helping," is there an example of Kashtanova's unassisted artwork for this project?Benj Edwards said:Kashtanova's attorney responded to the letter in November with an argument that Kashtanova authored every aspect of the work, with Midjourney serving merely as an assistive tool.
Right, but there’s nothing stopping someone from setting up a locked-down instance like you’re describing and opening it up to public use…so the question remains.You'd probably need to run your own stable diffusion instance and control all the seeds in order to repeat output.
IIRC, that monkey case was decided as, the monkey was the author of the picture for copyright purposes, but since a monkey can't hold a copyright, the pictures are public domain.Personally, I agree that copyrights should not be issued for AI-generated images. As for stories, that's more problematic, unless one wants to be pedantic and say that machines can't hold copyrights, and someone who generated the story is not the holder of the copyright, either.
Then we get into the "who owns the copyright, the camera's owner or the monkey who took the picture?" question.
AI's aren't entities who can own copyrights, but they're also "cameras" that can generate images based on data from other sources.
I wouldn't be at all upset if it's decided that nothing AI generated can be copyrighted. That would, at least, take all the wind out of the sales of AI-generated art and probably please the hell out of the artists and authors who actually work for a living.
Arguably in the case of Midjourney and other stable-diffusion software that operates in similar fashion, the creative input was the prompt given to the AI. Typically this would have to be iterated multiple times to get the sort of output that the artist is looking for. If another person provided exactly the same inputs to the algorithm, they should get the same generated image.“the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author”
The original images originated from a machine. Doesn't matter how creative she was. There will need to be a new law or court decision. This is such a novel situation (AI/ML that can output decent products) we need to have the best 'thinkers' in the country get together and decide how it should be handled.
And as these tools improve, getting faster, more complex, and more capable, I fully expect most professional artists are going to be relegated to the 'director' role, employing various AI tools to do the work that used to be done by artists. It's not going to eliminate artists so much as it's going to relegate them to more of a niche role, not unlike many other jobs that only exist in a dramatically paired down role in the modern world (like blacksmiths or stagecoach drivers).ANY prompter's contribution to the A.I. image generation process is purely ART DIRECTION.
It looks like I got tripped up with the "within the scope of work" portion. A photographer would have their photography work reassigned, but apparently if they say wrote code while working there but it wasn't part of the assigned duties persay it wouldn't be transferred.No, that is not the law. 17 USC 201(b) says the following:
It's the opposite of what you say. The employer owns the work by default, a contract is required to countermand that default. Repeatedly insisting that you are right will not change what is written in the statute book.
If you use an AI tool to create pieces and details that YOU then stitch together into a finished product, that should meet the bar necessary to be eligible for copyright. The AI isn't handling placement, composition, lighting, etc. You are.Does this mean an otherwise original artwork can become "tainted" if an ML model was used to say... populate some detail in the background using latent diffusion and then become ineligible for copyright protection? Kind of like doing sampling in music production.
Arguably in the case of Midjourney and other stable-diffusion software that operates in similar fashion, the creative input was the prompt given to the AI. Typically this would have to be iterated multiple times to get the sort of output that the artist is looking for. If another person provided exactly the same inputs to the algorithm, they should get the same generated image.
I wonder if an argument could be made that it's not the image itself that is copyrightable, but rather the specific combination of prompts given to the AI?
The act of applying an "unsharp mask" or "gaussian blur" operation on a digital image is purely mathematical. The creativity is in specifying which part of the image needs the operation applied, how strongly, with what radius, etc. In other words, the inputs to the algorithm.
"made by a human"?This is a counter example to the argument that AI will completely displace people for creating art or any other copyrighted material. Or an argument why AI works shouldn't be granted copyrights. If you want to monetize AND protect a creation, it needs to be made by a human.
As a thought experiment, how much AI do you thing Disney would rely on if it puts a cloud over their content?
Right so, you can create a sketch of an image, put it into img2img diffusion and get a fully rendered photograph or painted work, or whatever style you prompt for. You're dictating the composition yourself, but the AI is pulling in all the details from latent space to fill out the details. I guess purely using a text prompt without any image source fails the test. but adding a source image as input passes the test?If you use an AI tool to create pieces and details that YOU then stitch together into a finished product, that should meet the bar necessary to be eligible for copyright. The AI isn't handling placement, composition, lighting, etc. You are.
No, that is not the law. 17 USC 201(b) says the following:
It's the opposite of what you say. The employer owns the work by default, a contract is required to countermand that default. Repeatedly insisting that you are right will not change what is written in the statute book.
What these "prompt engineers" do is the equivalent of shopping on Amazon for new drapes. Type in the search bar the description of the style of drapes you want, then look through the results until you find what you were more-or-less looking for.ANY prompter's contribution to the A.I. image generation process is purely ART DIRECTION.
I wonder if an argument could be made that it's not the image itself that is copyrightable, but rather the specific combination of prompts given to the AI?
This still doesn't prevent Disney from using AI to generate "filler" art and backgrounds (so they won't have to pay humans to do it), while keeping most of the content human-generated so it is copyrightable. Which is what will probably happen.This is a counter example to the argument that AI will completely displace people for creating art or any other copyrighted material. Or an argument why AI works shouldn't be granted copyrights. If you want to monetize AND protect a creation, it needs to be made by a human.
As a thought experiment, how much AI do you thing Disney would rely on if it puts a cloud over their content?
That's not any different than providing it a text prompt. None of your 'work' is present in the final product. An AI is responsible for all of it.Right so, you can create a sketch of an image, put it into img2img diffusion and get a fully rendered photograph or painted work, or whatever style you prompt for. You're dictating the composition yourself, but the AI is pulling in all the details from latent space to fill out the details. I guess purely using a text prompt without any image source fails the test. but adding a source image as input passes the test?