Dall-E 2 can create images from a few words, but is the product yours? Your digital artwork generated by AI may not be copyrighted

The creative nature of neural networks leads some to question whether it might not be useful to amend existing US laws that grant copyright protection only to works created by humans.

Generative models capable of automatically producing paragraphs of text or digital art are becoming increasingly available. People use them to write fantasy novels, marketing copy, and create memes and magazine covers. The content automatically generated by the software is about to flood the internet for better or for worse with the marketing of artificial intelligence technology. Can any of this be legally protected under copyright laws*?

Take, for example, the cover of the world’s first artificially intelligent world*: a photo of a giant astronaut walking on the surface of a planet against a dark sky scattered with what looks like stars and gas, such as that produced by the DALL-E 2 OpenAI model. Creative Director Karen Cheng described experiencing several script prompts to guide DALL-E 2 in producing the perfect picture.

For the record, DALL-E 2 is OpenAI’s proprietary AI system that can create images from a few words or edit and retouch existing images in the same way. For example, a fox in a tree wave might show an image of a fox sitting on a tree, or a wave would show an astronaut with a cake in hand well, you see where this is going. The program not only creates a picture in one style, but you can add different art techniques on demand, insert drawing patterns, oil painting, plasticine model, woolen knitting, cave wall painting, or even like a 1960s movie poster.

In addition to the technology’s ability to produce images only in text prompts, the Dall-E 2 has two other smart technologies: internal drawing (the name given to a technology that reconstructs damaged images or fills in missing parts of an image) and variants. Both applications work similarly to the rest. Dall-E apps, just one touch.

With Inner Drawing, you can take an existing image and add new elements to it or modify parts of it. If you have a picture of the living room, you can add a new rug, a dog on the couch, change the picture on the wall, or even introduce an elephant into the room because someone might think of it, you see.

Variations is another service that requires an existing image. Insert a photo, illustration, or any other type of image and Dall-E’s Contrast Tool will create hundreds of versions of it. You can give him a picture of one of the teletubbies, and he will duplicate it, and make duplicates. An old painting of Samura will create similar images, you can even take a picture of some of the graffiti you see and get similar results.

You can also use this tool to combine two photos into one. Mix a dragon and a corgi, or a rainbow and a bowl to make colorful pots.

So it took Karen Cheng to find the right text prompt for Dall-E 2 to create the image. Then she said a photo of DALL-E 2 to create the magazine’s final stylish cover. However, who owns the copyright*? Who is the author of the picture?

The answer likely depends on how much human intervention is needed to create something, said Mike Wolfe, copyright attorney at Rosen, Wolfe and Hwang: To create a work, there are still ways to some degree of copyright protection. Even with a highly capable AI, there is likely plenty of room for human creativity. If the AI ​​helps create a song and make an audio line, but the creative professional makes it more complete by filling in the gaps to make a coherent piece of music, then that same work is likely eligible for copyright. On the basis of human paternity.

In practice, this would mean that the melody or bass line could be used freely by a third party, since those parts were machine generated and not copyrighted, but people couldn’t literally copy the entire song, Wolf noted. But in reality, it may not be easy to separate human and machine work. Going back to the global cover example, it’s not entirely clear which parts of the image were created by DALL-E 2 and which parts by Cheng.

Designers often have little effect on the model’s output, especially for more visual systems like DALL-E 2. The resulting image is intact. Under US copyright law, these images are not technically subject to copyright protection. Only the original works of the author are taken into account. To qualify as a work of “authorship,” the work must be created by a human, according to a report from the United States Copyright Office.

DABUS case

One man, Stephen Thaler, founder of Imagination Engines, a Missouri-based software company, discovered the hard way.

On November 3, 2018, Thaler filed an application to register a copyright claim on a work. The author of the work is identified as the Creativity Machine, with Thaler listed as the plaintiff. In his application, Thaler left a note with the office stating that the work was created independently by a computer algorithm running on a machine and that he sought to register this computer-generated work as a work for hire with the owner of the creativity machine. In a letter dated August 12, 2019, a registration specialist at the Copyright Office refused to register the application, believing it lacked the human authorship necessary to substantiate a copyright claim.

Thaler then asked the Bureau to reconsider its initial refusal to record the work, arguing that the human authorship requirement was unconstitutional and not supported by law or case law.

After reviewing the work in light of the issues raised in the first application, the office re-evaluated the allegations and again concluded in March 2020 that the work lacked the human authorship required to support a copyright claim. A creative contribution or sufficient involvement of a human author in the work. The office also said it would not abandon its longstanding interpretation of copyright law, Supreme Court and lower court precedent that “a work does not meet the legal and official requirements for copyright protection unless it was created by a human author.”

In his May 2, 2020 reconsideration request, Thaler reiterated his arguments that the Office’s human authorship requirement is unconstitutional and unsupported by case law. The second application reiterates the same arguments as the first, and broadly advances public policy arguments that the Office should register copyrights in machine-generated works, as this would further the fundamental objectives of the law on copyright, including the constitutional justification for protecting rights publishing.

In response to the Bureau’s citation of case law relating to human authorship, Thaler maintains that there is no binding authority to prohibit copyright for [les uvres gnres par ordinateur], that is, copyright law already allows non-human entities to be authors under the principle of remuneration work; Finally, the Copyright Office “is currently relying on non-binding court opinions from the Golden Age to answer the question of whether [les uvres gnres par ordinateur] can be protected.


Modern entrance to heaven

The work of the Creativity Machine, as described above, is titled Modern entrance to heaven (Literally a hadith entry to heaven). It is part of a series described by Thaler as a simulated near-death experience, in which an image-processing algorithm recreates an image to create hallucinatory images and a fictionalized account of the afterlife. Most importantly, AI is supposed to do this with minimal human intervention, but the Copyright Office has not granted it the copyright.

But the US Copyright Office rejected his application.

The United States Constitution grants Congress the power to protect intellectual property in Article I, Section 8*: to promote the progress of science and the useful arts, by guaranteeing for a limited time authors and inventors the exclusive right to their writings and discoveries. .

Thaler’s attorney, Ryan Abbott, believes that the US Copyright Office erred in denying his application for copyright registration of the AI. AI is able to produce functional creative output in the absence of a traditional human author, and copyright protection for works created by AI is essential to enhance the production of socially valuable content. He previously said that such protections are required under current legal frameworks.

But not all jurists agree with him. The onus should always fall on the creator to prove that the copyright they acquire is for the benefit of the public. In my opinion, this charge is not worn for machines. Wolf said that currently granting rights to works created by AI does not seem to make us richer or more advanced.

Do we want machines treated equally before the law*? There doesn’t seem to be much enthusiasm for that. But as you see more and more impressive results from these really powerful systems, I’m sure the math can change.

Sources: US Copyright Office, Karen Chengworldwide

And you?

Can we really talk about concepts generated by artificial intelligence? To what extent?
In the case of DALL-E 2, who do you think the author of the created work should be credited with? why ?
Should AI be able to obtain patents? why ?
If not, who should benefit from the concepts it presents?
What is the situation in your country?
Who should be considered the inventor of an AI product?
Can we claim (protect) the copyright of an AI-powered product?

See also:

Dall-E 2: The AI ​​image creator developed by OpenAI is available in beta. It can produce a wide range of images from a few words
An engineer is suing the US Patent Office for refusing to grant patents to his artificial intelligence, arguing that only natural persons are entitled to do so.

Leave a Comment