File talk:Théâtre D’opéra Spatial.jpg

 

 

Not Public Domain

This image is not in the public domain, and the author did not give their consent. All rights of images created at Midjourney by subscribers belong to the person who created them. Additionally, this author made substantial changes to the images after creation at Midjourney, as well. (I am a guide at midjourney, and the author, Jason Allen, is right now complaining to us.)

It should be removed.

Gejyspa (talk) 11:16, 8 September 2022 (UTC)

I replied within the deletion discussion, however to echo my response I have typed it below:

The image I uploaded was an image generated by an artificial intelligence program, known as Midjourney, from within the United States. Therefore it is not a copyright violation per Template:PD-Algorithm and [1]

Secondly he has gone on record that after he has got the desired output from the Midjourney algorithm he upscaled it to a higher resolution using another artificial intelligence program. (Gigapixel A.I.)[2]

For sake of clarity here is the direct quote. “…I chose my top three and had them printed on canvas after upscaling with Gigapixel A.I.”

For these reasons the artwork falls under the public domain, as both tools used to make such art were simply artificial intelligence programs.

This is a faithful photographic reproduction of a two-dimensional, public domain work of art. The work of art itself is in the public domain for the following reason: Template:PD-Algorithm The official position taken by the Wikimedia Foundation is that "faithful reproductions of two-dimensional public domain works of art are public domain". This photographic reproduction is therefore also considered to be in the public domain in the United States. In other jurisdictions, re-use of this content may be restricted; see Reuse of PD-Art photographs for details.

ALittleLighter (talk) 15:40, 8 September 2022 (UTC)

It appears mobile messed with my inputs a bit. I fixed the tags below:

This is a faithful photographic reproduction of a two-dimensional, public domain work of art. The work of art itself is in the public domain for the following reason: This file is in the public domain because it is the work of a computer algorithm or artificial intelligence and does not contain sufficient human authorship to support a copyright claim. The United Kingdom (legislation) and Hong Kong (legislation) provide a limited term of copyright protection for computer-generated works of 50 years from creation. Legal disclaimer Most image-generating AI models were trained using works that are protected by copyright. In some cases, such models can output content with major copyrightable image elements which are identical to or derivative of the original training data, making these outputs derivative works. Accordingly, there is a risk that AI-generated media uploaded on Commons may violate the rights of the authors of the original works. See Commons:AI-generated media for additional details. The official position taken by the Wikimedia Foundation is that "faithful reproductions of two-dimensional public domain works of art are public domain". This photographic reproduction is therefore also considered to be in the public domain in the United States. In other jurisdictions, re-use of this content may be restricted; see Reuse of PD-Art photographs for details.

ALittleLighter (talk) 15:42, 8 September 2022 (UTC)

While a large portion of the art was generated by Midjourney and the Gigapixel AI, the author also manually edited the image to remove artifacts and improve the image quality. As such, the claim that this is an algorithmically generated piece and thus public domain is not correct. A rather similar AI generated piece may exist, but it has not been released. The author mentions the use of Photoshop to clean up the AI images in this source: https://www.washingtonpost.com/technology/2022/09/02/midjourney-artificial-intelligence-state-fair-colorado/ SomeguyDave (talk) 18:20, 8 September 2022 (UTC)

Thank you for the reply.

I do seek clarification for the added work to this piece of art in particular. The source cited mentioned broadly that photoshop was mainly used to "remove visual artifacts" of the algorithmically generated piece of art, but doesn't seem to mention directly to what extent this entails. (It does mentions once case, a missing head for a central figure, but doesn't directly attribute it which of the three pieces he submitted... hence the ambiguity.)

Secondly if photoshop was used to purely remove visual artifacts, and enhance the overall quality of the algorithmically generated piece of art, I would argue that it may still fall under public domain through it being a "faithful reproduction of a two-dimensional, public domain work of art." ( As pointed out above, via embed, the official position taken by the Wikimedia Foundation is that "faithful reproductions of two-dimensional public domain works of art are public domain". Algorithmically generated pieces of art, from within the United States, is public domain. )

I will note that if the manual editing process for this work of art in particular was more in-depth, to the point where it would fall outside being a "faithful reproduction of a two-dimensional, public domain work", then I will concede that I uploaded this image in error and be in full agreement that it should be removed as it would not be public domain. ALittleLighter (talk) 19:47, 8 September 2022 (UTC)

Thank you for your reply.

I do not think that the extent of the editing is relevant. Any edit, even a single pixel, on top of what the algorithm produced, is an act of artistic authorship layered onto what the AI created.

Additionally, the copyright case that you reference refers to AI autonomously creating artwork with the AI itself receiving rights: "The Board accepts as a threshold matter Thaler’s representation that the Work was autonomously created by artificial intelligence without any creative contribution from a human actor" (Section III Discussion: https://www.copyright.gov/rulings-filings/review-board/docs/a-recent-entrance-to-paradise.pdf) The situation with Midjourney is quite different, as the algorithm is not autonomous. In this case, the artist provided a prompt, received images, selected the next generation of images to feed the algorithm, modified the prompt, and repeated this process hundreds of times over several weeks before upscaling and editing the final result. This constitutes a straight-forward creative contribution from a human actor which removes this piece from the public domain. SomeguyDave (talk) 20:19, 8 September 2022 (UTC)

Thanks again for further reply.

First, to set a baseline of clarity, under United States copyright law artwork originating from text prompts are technically not subject to copyright protection. Only "original works of authorship" are considered. "To qualify as a work of 'authorship' a work must be created by a human being." This is according to a US Copyright Office's report. . Therefore the process of sending in a prompt, receiving an output of images, then refeeding the algorithm with those images and a modified prompt would likely not meet the threshold for the final outputted work to have copyright protection. (As you are only modifying the prompt, and refeeding previously outputted images into the algorithm.)

With that in mind we come back to the extent of the manual edits made to the final output from the Midjourney Algorithm. The reason the extent of the editing is relevant is that the Wikimedia Foundation holds that faithful reproduction of two-dimensional public domain work is still within the public domain. Hence why I am seeking clarification for this piece of art as the source cited mentions that the manual work done was to "remove visual artifacts". The question is if "removing visual artifacts" falls under a "faithful reproduction" and therefore public domain in the eyes of the Wikimedia Foundation, or if these edits are more in-depth and wouldn't fall under that that interpretation.

Currently I cannot seem to directly locate a source that could be used as a baseline in clarifying when changes made to a public domain work would be seen as substantial enough to be worthy of copyright protection. However looking at a document from the Copyright Office, in particular a section that covers derivative sound recordings, it notes the following: "Mechanical changes or processes, such as a change in format, declicking, or noise reduction, generally do not contain enough original authorship to warrant registration.". Therefore "one pixel edits" likely wouldn't be seen as substantial enough to count as a derivative work by the Copyright Office, and why determining the extent of the manual alterations to the algorithmically outputted image is important. ALittleLighter (talk) 21:47, 8 September 2022 (UTC)

I have found a source that also covers derivative works more broadly. As noted by the Legal Information Institute: "Overall, one cannot simply change a few words in a written work for example to create a derivative work; one must substantially change the content of the work. Along the same lines, a work must incorporate enough of the original work that it obviously stems from the original." Therefore minor edits, like as you put it "one pixel edits", would likely not reach threshold needed to be qualified as a derivative work worthy of copyright protection. ALittleLighter (talk) 21:54, 8 September 2022 (UTC)

Last addendum, and likely the last time I try to use formatting commands on mobile, the two cited sources are as follows:

[1] https://www.smithsonianmag.com/smart-news/us-copyright-office-rules-ai-art-cant-be-copyrighted-180979808/ [2] https://www.iflscience.com/ai-generated-artwork-wins-state-fair-competition-leaving-human-artists-unhappy-65189 ALittleLighter (talk) 15:46, 8 September 2022 (UTC)

The problem is that this has not been tested in a court of law. What we DO know from the case cited is that the copyright cannot belong to an AI. But that is not the case there. Mr Allen is the owner. While it cannot be denied that an AI created the artwork based on the prompt of Mr. Allen, it was not jsut some single prompt. He worked with the bot over the course of many weeks, refining his prompt almost one thousand times, and using variations of previous results, to get the result he wanted [op cit]. This was not some simple throwaway two minute thing that the original uploader tries to assert here. The law here is at best ambiguous, and cannot be presumed to be in the public domain. Gejyspa (talk) 01:42, 9 September 2022 (UTC)

Thanks for the reply. (I will post this as well in the other thread to keep both updated)

First and foremost I am not trying to assert that Mr. Allen got this output in a matter of "two minutes". I make it clear in my previous post, under the "How The Art Was Made" section if you are viewing this under the deletion request, that the process took at least several weeks.

Secondly, as you have confirmed in your reply, this this piece of art was designed by the Midjourney based off a extensive text prompts provided by Mr. Allen. You have stated that he has refined his prompt "thousands of times" and used variations of the past results, provided by Midjourney, to fine-tune the results from the Midjourney algorithm. The length of time that it took for this process, as well as the subsequent text prompt revisions, are irrelevant in the eyes of the US Copyright Office when determining if a work is eligible for copyright protection.

As stated above, and reposted here for clarity along with its sources: Under United States copyright law artwork originating from text prompts are technically not subject to copyright protection. Only "original works of authorship" are considered. "To qualify as a work of 'authorship' a work must be created by a human being." This is according to a US Copyright Office's report.

Therefore the process of sending in a text prompt to an algorithm, receiving an output of images from that algorithm, then refeeding said algorithm with those past images along with any modifications to the text prompt would not meet the "Human Authorship" requirements for copyright protection as set out by the US Copyright Office (As the final outputted work itself is still algorithmically created). The length of time spent refining the algorithm's outputs, in and of itself, does not change the fact that the final outputted work still lacks the required "Human Authorship" for copyright protection as set out by the US Copyright Office. (As again to qualify as a work of 'authorship', in the eyes of the US Copyright Office, the work itself must be created by a human being.)

For the reasons listed above I believe that it is a safe presumption to consider this work as public domain, due to the final work lacking the required "Human Authorship" as defined by the US Copyright Office (Which in turn makes it ineligible for copyright protection). ALittleLighter (talk) 09:32, 9 September 2022 (UTC)