Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
Graphics AI The 2000 Beanies

Artist Appeals Copyright Denial For Prize-Winning AI-Generated Work (arstechnica.com) 75

An anonymous reader quotes a report from Ars Technica: Jason Allen-a synthetic media artist whose Midjourney-generated work "Theatre D'opera Spatial" went viral and incited backlash after winning a state fair art competition-is not giving up his fight with the US Copyright Office. Last fall, the Copyright Office refused to register Allen's work, claiming that almost the entire work was AI-generated and insisting that copyright registration requires more human authorship than simply plugging a prompt into Midjourney. Allen is now appealing (PDF) that decision, asking for judicial review and alleging that "the negative media attention surrounding the Work may have influenced the Copyright Office Examiner's perception and judgment." He claims that the Examiner was biased and considered "improper factors" such as the public backlash when concluding that he had "no control over how the artificial intelligence tool analyzed, interpreted, or responded to these prompts."

As Allen sees it, a rule establishing a review process requiring an Examiner to determine which parts of the work are human-authored seems "entirely arbitrary" since some Copyright Examiners "may not even be able to distinguish an artwork that used AI tools to assist in the creation from one which does not use any computerized tools." Further, Allen claims that the denial of copyright for his work has inspired confusion about who owns rights to not just Midjourney-generated art but all AI art, and as AI technology rapidly improves, it will only become harder for the Copyright Office to make those authorship judgment calls. That becomes an even bigger problem if the Copyright Office gets it wrong too often, Allen warned, running the risk of turning every artist registering works into a "suspect" and potentially bogging courts down with copyright disputes. Ultimately, Allen is hoping that a jury reviewing his appeal will reverse the denial, arguing that there is more human authorship in his AI-generated work than the Copyright Office considered when twice rejecting his registration.

This discussion has been archived. No new comments can be posted.

Artist Appeals Copyright Denial For Prize-Winning AI-Generated Work

Comments Filter:
  • Artwork produced by generative AI is done by ingesting a large body of (presumably) original work, and the regurgitated back. It is NOT original work. If copyright is to be granted to an AI generated work, it should be granted to each of the artists of the original work.
    • Re:Is this theft? (Score:4, Informative)

      by chuckugly ( 2030942 ) on Monday October 07, 2024 @11:36PM (#64847261)

      Because no artist ever studied the work that preceded them before creating their art. Right?

      • Because no artist ever studied the work that preceded them before creating their art. Right?

        That is pretty much the basis of it all. There is very little new under the sun. For my photography, I'm heavily influenced by Brett Weston.

        So art is pretty much tweaking things, other than the group of goofballs that try to see what they can get away with, and the idiots they can convince that they are creating art - think Damien Hirst https://www.thecollector.com/d... [thecollector.com] Yeah, kill an animal, put it in a glass box and watch it rot away.

        Now of course, art is what you can get away with, and that twatwaffl

      • In my opinion, this line of thought ought to be irrelevant when it comes to deciding what copyright law should look like.

        Copyright is an enforced monopoly on the *distribution* of a creative work.

        How a human being created the creative work or obtained the copyright (rights can be traded) is largely irrelevant, unless they broke a law while creating it. Be it by using someone else's material and falsely representing it as their own, or because they engaged in some other illegal activity as part of the produc

    • Re:Is this theft? (Score:4, Informative)

      by rolfwind ( 528248 ) on Monday October 07, 2024 @11:48PM (#64847287)

      Yeah, I think the human mind works the same way. Are you gonna claim that anybody that went through school owes infinite copyright to previous generation starting with use of the alphabet and number system?

      • That's the question, isn't it? Are we just machines or is there something special about us, and ostensibly other living creatures by extension? So far, it's unproven either way, and anyone who says different is selling something (probably either AI software on one hand, or a holy book on the other.)

        So far the law has taken the view that we are special, and that is frankly the only thing that makes sense. Humans need protection from oppression, software doesn't (although humans need software to have certain

    • by Rei ( 128717 )

      Artwork produced by generative AI is done by ingesting a large body of (presumably) original work

      Correct

      and the regurgitated back

      Incorrect.

  • It claims (from TFA) that the prompt used is a copyrightable work. Midjourney's output is clearly a derivative work based on a copyrightable original. How can that not be copyrightable itself? The claim could be made (in this case) that Midjourney modifies the original prompt in ways that are hidden from the user and the resulting work is a collaboration. Fair enough. However, it would not be broadly applicable to all synthetic images. If I use a local installation of Automatic1111 or ComfyUI, I have compl
    • The claim could be made (in this case) that Midjourney modifies the original prompt in ways that are hidden from the user and the resulting work is a collaboration.

      At that point, you might as well claim that a compiler partially owns your code as a collaboration. I don't even buy that part.

  • This sounds to me like 'prompt engineer' (god, i hate that term so much. did these prompters go to an abet accredited institution? did they pass any exams or have any training whatsoever?) are trying to claim copyright over others hard work, (artists, authors, computer scientists, and actual computer engineers who earned their degrees and did lots of hard math to create the ai whose buttons these prompters are mashing blindly.
    • The important aspect of the engineering degree is not the accredited institution.

      The important part of engineering is the stamp and the associated responsibility. The stamp says "This design obeys commonly understood engineering principles and I certify it's not going to kill anyone if implemented as specified and take full responsibility for those outcomes."

      There is none of that in so-called software engineering. Taking that concept one step further into the sociological realm, I don't think there's anyth

      • You are confusing CE/PE certification with engineering. The difference between aa ABET Software engineer and a Software developer, is that a software engineer has training in math, including calculus and advanced math, physics, engineering principals and other physical sciences. This is extremely helpful in making software that interfaces with the real world, such as the software in streetlights, cars, medical equipment, rockets and aero planes. These are the exact same classes that other engineering stud
    • by Rei ( 128717 )

      You're right. When I paint a painting, I shouldn't get the copyright. Rather, the people who designed and engineered the paintbrush should hold the copyright.

      Nikon should be a mass copyright holder, owning the rights to every photo ever taken with a Nikon camera.

      This makes sense.

    • You diminish the value of words, or even a single word.

      "Dreams" by Langston Hughes
      Hold fast to dreams
      For if dreams die
      Life is a broken-winged bird
      That cannot fly.

      Add "nude" to pretty much any Stable Diffusion prompt and it changes it a whole lot.

      The same holds for LLM prompts, if you know the right words.

  • We need less (Score:4, Insightful)

    by MpVpRb ( 1423381 ) on Tuesday October 08, 2024 @12:18AM (#64847331)

    IP laws, not more

  • Can we all please stop referring to people who use AI tools as artists? If your work is wholly AI generated and you added nothing but some prompts, you are not Da Vinci. I don't know if you should even be considered as equal to someone who colors inbetween the lines of a diner menu with those free crayons.
    • What if you created artwork from a mathematical formula, or used one to modify a basic image to make it visually interesting? What if you use a CNC machine or 3D printer to make a sculpture?

      • Can you copyright a fractal?
        • You can't copyright math per se, but a representation based on math is the question. If the underlying formula is obvious in the manifestation then it looks more like math than art-- but where is the line?

    • What's the difference between someone using one of these systems and using a synthesizer? Given the same synthesizer model and input settings, isn't the variation then purely based on the notes played? What if those notes are played by a sequencer and not a person or generated programmatically? Is that person not an artist?

      I agree with your point regarding so-called artists that use AI tools but I would also contend it's not as clear-cut as it may appear.

    • by Rei ( 128717 )

      Can we all please stop referring to people who use AI tools as artists? If your work is wholly AI generated and you added nothing but some prompts, you are not Da Vinci. I don't know if you should even be considered as equal to someone who colors inbetween the lines of a diner menu with those free crayons.

      This sounds familiar. [csus.edu]

      A revengeful God has given ear to the prayers of this multitude. Daguerre was his Messiah. And now the faithful says to himself: “Since photography gives us every guarantee of e

    • by Tablizer ( 95088 )

      "Esthetics Engineer"?

    • Art isn't limited to just the ability to create aesthetically pleasing works. It's not even limited to creating static works.

      Certainly a piece like "The Treachery of Images" [wikipedia.org] isn't interesting because of its creative visual style. It's still art because of the expression of ideas. Even if you don't like it. If it was originally created by an AI image generator it wouldn't lose anything artistically.

  • If I use some AI tool to replace the background on one of my images, I still own the full copyright. If I use some AI to generate a full image from my prompt and then do some post-processing, I own no copyright. Can someone define the exact threshold where I lose the rights? To me, the distinction is very vague.

  • by tiqui ( 1024021 ) on Tuesday October 08, 2024 @02:55AM (#64847483)

    then you understand both why he lost, and why he must ultimately lose at the end of whatever string of appeals are deployed.

    In the United States, it's a basic HUMAN right (what some would call a "God-given" right... in other words a right that even trumps majority rule because it was not granted by some group of humans) to have control over what you create on your time, with your tools and your creativity. In order to improve society and encourage creative people to give the public access to the results of their efforts, the founders of the country established the "copyright" (a thing, which may be thought of as a right to make copies of a work). By default, the copyright for any work belongs to the person who created the work, but that person may sell, trade, barter, etc that right to another party (like a book publisher, or movie distributor). For as long as the copyright is in force, only persons holding/granted that copyright have the legal right to make copies of a work. The bargain struck is that in exchange for government power enforcing that copyright on behalf of the creator of a work, after the copyright expires, the work becomes public domain and any other person my copy the work and use it or even base other works on it. The presumption is that the creator of the work has had his chance to profit from that work and the expiration witll encourage him to create more works.

    The emergence of corporate monsters like Disney, and the ability to bribe... errr... "lobby " and "support" politicians to manipulate the duration of the copyright and the idea of massive immortal corporations highly-dependent upon copyrights have introduced lots of distractions/confusion to the arguments. At the core, however, is the concept that "copyright" is a HUMAN right.

  • Pretends to make art.

  • Something should be copyrightable if it can be considered created in a uniquely creative way.

    i.e. Can some random lay person playing with prompts replicate the picture in such a way that copyright law would consider the output infringing on the other image? If so then it's probably not worth copyrighting. If on the other hand it takes actual skill of someone who knows what they are doing to get to that point then the result should be subject to copyright.

    • Not even the artist can recreate the image without a stored set of parameters, because these images are extrapolated from (pseudo-)randomness. The nature of the prompts is selective, not creative. Unless you can't play five songs in the same order as on my mix-tape without paying me royalties for my artistic genius, these images deserve no more copyright than the output of /dev/urandom.

      • by Rei ( 128717 )

        Please learn what a seed is.

        • Oh intelligent one, please enlighten me. I mentioned stored parameters and pseudo randomness, but the meaning of these words eludes me. I beg your forgiveness.

          • by Rei ( 128717 )

            AI artists can recreate the exact same work by using the exact same seed value and inputs. So your entire argument makes no sense. They DO have the "stored set of parameters". In fact, these parameters are often stored in the image metadata, thus causing the image itself to contain the means of its own creation. Some go even further: with ComfyUI, it's common to store entire image generation pipelines in the metadata of an image, to the point that you can load the image as a workflow and have hundreds o

            • Those are stored parameters, numbnut. Unless they can remember the seed value and inputs verbatim and guarantee that the "AI" they used hasn't been retrained, they can not recreate their "art". The prompt is not enough, and certainly not a prompt that looks very very similar. So what this boils down to is: Anyone who has all that data can recreate the image or nobody can. The artist didn't create the image. He chose it randomly from more images than could be looked at by all of humanity until the heat death

              • by Rei ( 128717 )

                It's in the image itself. They do have it.

                Furthermore, you absolutely CAN reverse it from the image without any metadata. That's literally what unCLIP and its ilk are. I can unclip an image of you and put you into different situations, without having to train a custom model on it.

                • Which part of "anyone can or nobody can" do you not understand? The point of this discussion is that there allegedly is a creative element to the "art" and to prove it, consider that other people can't recreate the "art". But without the stored parameters or the image itself as a base, neither can the "artist", and with the data, anyone can, so this is a bullshit test.

                  • by Rei ( 128717 )

                    That's a nonsensical standard you're promoting. The exact same thing could be said about, say, a 3d artist's work. Either you have their .blend file or you don't.

      • Not even the artist can recreate the image without a stored set of parameters, because these images are extrapolated from (pseudo-)randomness.

        That is false, these images are seeded and thus can be recreated. But I didn't say perfectly recreate. I said recreate in a way that would be considered copyright infringing. If you ask Davinci to redraw the Monalisa the brushstrokes would look slightly different too.

        Unless you can't play five songs in the same order as on my mix-tape without paying me royalties for my artistic genius, these images deserve no more copyright than the output of /dev/urandom.

        Funny you bring music into this given that artists are paying each other royalties over simply notes not even using the same timing. I.e. All royalties from Bitter Sweet Symphony go to the rolling stones for the simple fact that a completely di

  • Here's a solution: make the prompts inputted to the AI copyright, but the artwork itself in the public domain. Surely then everyone is happy - the author gets what he wants (copyright protection for HIS part in the whole thing) and we aren't left with any complicated "is it or isn't it?" questions.

  • ...don't tell anyone how it was made.

  • by omnichad ( 1198475 ) on Tuesday October 08, 2024 @01:07PM (#64849047) Homepage

    If AI art isn't copyrightable, there are a lot of online image (or music, or video) generators out there who say you need a subscription to use the generated works commercially.

    If you generate it yourself, this means the most they could come after you for is a TOS violation. But if you find the picture in a Google Image search and copy it (knowing it was for sure AI generated), then it's a free for all.

Force needed to accelerate 2.2lbs of cookies = 1 Fig-newton to 1 meter per second

Working...