Commons:Deletion requests/File:DALL-E sample.png

This deletion discussion is now closed. Please do not make any edits to this archive. You can read the deletion policy or ask a question at the Village pump. If the circumstances surrounding this file have changed in a notable manner, you may re-nominate this file or ask for it to be undeleted.

File:DALL-E sample.png

The content of OpenAI's blog is not under a free license. The MIT license applies to the DALL-E source code, not to the images generated by OpenAI. TilmannR (talk) 13:32, 6 May 2022 (UTC)[reply]

In particular:

"As this is an experimental research platform, you may not use generated images for commercial purposes. For example:

"You may not license, sell, trade, or otherwise transact on these image generations in any form, including through related assets such as NFTs.

"You may not serve these image generations to others through a web application or through other means of third-parties initiating a request."

-- https://labs.openai.com/policies/content-policy

That last clause, taken at face value, is a completely Wikimedia-incompatible (not to mention explicitly web- and social media-incompatible) restriction. But it's clearly wrong because the first sentence of that Content Policy says, "Do not attempt to create, upload, or share images that are not G-rated or that could cause harm," which clearly implies that they intend for G-rated, harmless images to be uploaded and shared. Someone should email support at openai.com and ask them to clarify their policy, preferably allowing commercial use (similar to how MS-Paint users are allowed commercial use of the images they create.)

As it is, I recommend keeping it at least until a non-contradictory OpenAI image content policy is forthcoming. The spirit of the current policy is clear even though the letter is self-contradictory. 2601:647:4D00:2C40:0:0:0:75DA 05:02, 18 May 2022 (UTC)[reply]

Neither the content policy nor the MS-Paint analogy are applicable, because these giraffe-dragon images are not user-generated content. They were generated by OpenAI for their blog. (Go to https://openai.com/blog/dall-e/#animal-chimeras , click on the box with the giraffe-turtle, replace "turtle" with "dragon" and you get exactly the same images in exactly the same order.) TilmannR (talk) 16:39, 18 May 2022 (UTC)[reply]

Keep The image is created by an artificial intelligence. As such, there is no copyright protection. The content policy of OpenAI is not valid on copyright grounds because OpenAI does not hold the copyright to the image. If anything, it's more of a request than a legally-binding license requirement, considering the work is in the public domain by default as a work of an AI. See Template:PD-algorithm. Di (they-them) (talk) 11:48, 24 May 2022 (UTC)[reply]

PD-algorithm applies to "work done entirely by an artificial intelligence or computer algorithm", but this image is based on a human-authored prompt, which required plenty of creativity, using specific phrases and repetition to make desirable results more likely.

There is another active deletion request with a discussion between Di and me about the same topic:

TilmannR (talk) 03:59, 25 May 2022 (UTC)[reply]

@TilmannR: Then the human owns the copyright to the prompt, not the images generated from the prompt. You cannot copyright an idea, only an expression of an idea, and the expression has to be created by a human in order to have copyright protection (at least in the United States). Nosferattus (talk) 22:28, 4 June 2022 (UTC)[reply]

@Nosferattus: Your argument implies that all digital photographs are in the public domain, because it was the camera and not the human who created the image. Unless someone presents a law, regulation or court decision, which establishes a copyright-relevant difference between a camera and a generative AI, this image must be considered "created by humans" and is therefore eligible for copyright protection. TilmannR (talk) 11:37, 5 June 2022 (UTC)[reply]

That will certainly not happen until an AI makes law. I am not sure of the copyright status here, but an AI is clearly quite different from a camera. Yann (talk) 11:51, 5 June 2022 (UTC)[reply]

Note that I didn't claim that cameras and AI were indistinguishable in every way. I claim that the differences between them (e.g. whether they capture photons) have never been relevant for determining copyright eligibility. TilmannR (talk) 13:47, 5 June 2022 (UTC)[reply]

Keep - No human author, thus no copyright protection. (Creating a prompt is not authoring a work.) Nosferattus (talk) 22:30, 4 June 2022 (UTC)[reply]

Please note that the people at OpenAI did not merely create the prompt, but they went through all the steps necessary to convert the prompt into a concrete image, which is what makes them the authors of this work. TilmannR (talk) 11:40, 5 June 2022 (UTC)[reply]

No, they created DALL-E and they create the prompt. They did not convert the prompt into a concrete image, as that is the entire purpose of DALL-E. You seem to be unwilling to accept that an artificial intelligence created the image, not a human. Nosferattus (talk) 14:36, 5 June 2022 (UTC)[reply]

Obviously I'm not denying that an artificial intelligence (specifically DALL-E) was involved in the creation of this image. But humans were also directly involved by giving their prompt to the AI. Which criteria do you apply to determine that the human involvement is negligible? TilmannR (talk) 15:00, 5 June 2022 (UTC)[reply]

@Nosferattus: Sorry, I forgot to ping you in my previous reply.

Also "You seem to be unwilling to accept [...]" sounds as if I'm simply ignoring your arguments for the lack of human authorship. So let me be more explicit:

By consistently applying the camera analogy, statements such as "They did not convert the [scene] into a concrete image, as that is the entire purpose of [a camera]." become thoroughly unconvincing. TilmannR (talk) 16:34, 5 June 2022 (UTC)[reply]

@TilmannR: All of the creativity involved in creating a photograph is directly controlled by the human: the angle, the exposure, the shutter speed, the time of day, whether to use a filter, etc. The depiction of the object itself is not what is necessarily copyrighted; it is the creative expression applied to that depiction. In the case of AIs, nearly all of the "creativity" directly involved in the creation of the image is from the AI, not the human. There is a small amount of creativity indirectly coming from the prompt, but this is de minimis, as how that prompt is expressed is completely up to the AI. I hope you can see the distinction. Nosferattus (talk) 18:31, 5 June 2022 (UTC)[reply]

Kept: no valid reason for deletion. Public domain as not created by a human. The teacher / students analogy is convincing. --Yann (talk) 19:47, 5 June 2022 (UTC)[reply]