Artificially Intelligent Artists?

By: Kiara D. Benac

Although art generated by algorithms has been around for years, the emergence of new programs that use machine learning to transform words into images within seconds, such as DALL-E 2, Midjourney, and Stable Diffusion, has become not only trendy but also controversial.  Historically, controversy has followed the introduction of new art-making technology.  For example, 19th-century painters criticized the invention of the camera, calling it “art’s most mortal enemy,” and 20th-century artists dismissed digital editing programs, viewing them as void of human skill.  Despite the pushback over the years, new art-making technologies have become commonplace by making it easier to unlock human creativity and bring one’s creative visions to life.

However, an issue arises specifically with the ethics of Artificial Intelligence generating artwork (“A.I.-generated art”).  Whereas prior art technologies served a role as an instrument or a tool analogous to a brush or canvas, the recent A.I. art programs diverge by making many of the creative decisions independent of the user; the computer program generates the work itself.  These platforms use machine learning software that first gathers millions of online images and then uses algorithms and pattern recognition to create new images from a string of text.  The programs may explicitly learn from artists who have their artwork available online and who did not provide permission for the program to use their art to train the algorithm.  Thus, artists who oppose these A.I. art platforms view them as a high-tech form of plagiarism.  Dr. Andres Guadamuz from the University of Sussex tweets AI-generated artwork in the style of other artists to highlight the legal and ethical issues that A-I-generated art poses.  Guadamuz expects that A.I.-generated art, especially those that replicate artwork by using an artist’s name in a text prompt, will lead to lawsuits.

These new A.I.-generated art programs present an interesting issue for copyright laws.  As the distinction between artwork made by a human versus a computer starts to blur, who should the United States Copyright Office (“USCO”) consider to be making the artwork: the platform, its creator, its user, or no one?

Article I, Section 8 of the United States Constitution grants Congress the power to protect intellectual property.  The USCO subsequently grants copyright protection for human-created works, protecting “original intellectual conceptions of the author.”  Although U.S. copyright law does not have an explicit guideline for non-human created works, it does not register a claim if it determines that a human being, or a group of humans in a corporation, did not create the work.  For example, courts have rejected copyright protection for non-human creations, such as when a monkey took a selfie and for a book of supposed divine revelations.  However, copyright applies to works where a human uses a machine as a tool, such as a camera.  Although skeptics argued that photography was the work of a machine devoid of creative expressions, the court in Burrow-Giles Lithographic Co. v. Sarony, ruled that copyright can apply to photos because they are “representatives of original intellectual conceptions of [an] author.”

Current copyright law, however, has not changed with the introduction of A.I.-generated artwork.  The USCO has now twice denied Stephen Thaler’s, Ph.D. and founder of Imagination Engines Incorporated, claims for A.I. generated artwork, initially in February 2019 and again in 2022.  Thaler attempted to gain a copyright for being the creator of an A.I. machine, the “Creativity Machine,” that generated a piece without any human help.  Despite Thaler’s argument that the “human authorship” requirement for copyright protection was unconstitutional, the USCO maintained its reasoning; it reinforced copyright jurisprudence that A.I.-generated artwork lacks the necessary “human authorship” element, even though Thaler built the A.I., and was not convinced to depart from copyright legal theory.  Both the European Patent Office and the United Kingdom also rejected Thaler’s applications.  The United Kingdom, however, recognizes the programmer’s work in creating the A.I. art generating program in Section 9(3) of its copyright law and creates an exception to the solely human authorship requirement.  Notably, in 2021, South Africa became the first country to issue a patent for an A.I.-generated invention without a human inventor – it granted the patent to the Artificial Inventor Project and listed the A.I. as the inventor.

In addition to Thaler’s efforts to gain copyright for creating the Creativity Machine, designer Roope Rainisto also filed a request with the USCO for an A.I.-generated artwork.  Rainisto announced in a tweet that the USCO registered the request as an open case to have the A.I.-generated artwork, called “Original Robots,” filed as a “Work of the Visual Arts.”  Rainisto followed up in a tweet on September 29, 2022, that the USCO responded to the request by reiterating that copyright law protects “the fruits of the intellectual labor” that “are founded in the creative powers of the mind.”  However, in its response, the USCO asked Rainisto to answer some questions, such as the prompt used and the number of revisions and creative inputs that occurred during the overall process.  The USCO examiner asked such questions in order to assess whether the work is the result of intellectual labor “with Dall-E 2 A.I. technology serving as a mere assisting instrument” or whether the technology “conceived and executed the uploaded work.”  As of September 30, 2022, the USCO’s revision of Rainisto’s request is still ongoing, however it provides insight into the USCO’s current evaluation of A.I. art and exemplifies how A.I.-generated artwork is an evolving area for copyright law.

Although copyright law has not yet adjusted to address A.I. art, Getty Images recently banned A.I.-generated artwork fearing future copyright claims.  The September 21 announcement, which was effective immediately, reflects just the tip of the iceberg when it comes to potential copyright litigation.  As Getty Images CEO, Craig Peters, explained to the Verge, “there are real concerns with respect to the copyright of outputs from these models and unaddressed rights issues with respect to the imagery, the image metadata and those individuals contained within the imagery.”

Using A.I.-generated art platforms to create images “in the style” of another artist may not yet be legally defined under copyright laws, but it does seem to toe the line of copying someone else’s work.  Although some may make the argument that artists commonly use the style of other artists to inspire their art, such a scenario still requires the human element in the creative process, whether it be playing an instrument or painting a piece, arguably unlike when A.I. generates art.

However, could the creative process taken by a user of an A.I.-generated artwork one day be considered sufficient to meet the USCO’s human authorship requirement?  For example, consider the development of Cosmopolitan Magazine’s June 2022 first A.I. magazine cover.  Although the cover itself states that it “only took 20 seconds to make,” the digital artist Karen X. Chen spent more than a hundred hours and hundreds of attempts before getting the final image.  In a tweet, Chen describes the A.I. generator as a tool for humans to use, akin to an instrument to play where one can improve with practice.  Therefore, is there a difference between a photographer copyrighting an image taken by a camera where the photographer set up the shot, worked on the settings, and edited the photo in post compared to a user of an A.I. art-generator who selected a theme, carefully crafted the phrase, and selected a final image amongst all the searches?

Last month’s Colorado State Fair reinvigorated the discussion around A.I.-generated artwork when an A.I.-generated piece won first place in the emerging digital artists’ division.  Jason M. Allen submitted his award-winning piece, “Théâtre D’opéra Spatial,” under the name “Jason M. Allen via Midjourney” to be transparent about the fact that he created it using A.I.  Although Allen faces backlash from artists who view his artwork as cheating, the two category judges who did not know that the piece was done by A.I. at the time say that they still would have awarded him the prize.  Allen’s denial to share the prompt he inputted may signal a sense of ownership over the creative process that resulted in the A.I. art piece.

As I worked on creating the image for this blog post, I gained insight into how someone could spend a lot of thought into carefully crafting a phrase to create a specific image, or vision.  I had to stop myself from spending too much time on the platform, but I can easily see how users devote time, thought, and craftsmanship into creating their works, much like photography and other art mediums.  Although I freely share that the prompt used to create the image was “a robot painting on a canvas a baby penguin wearing a cute hat, futuristic style,” I am empathetic to artists, such as Allen, who would rather keep the curated phrase secret.

What does the future of copyright laws pertaining to A.I.-generated artwork mean for the people using an A.I. art platform to create art, for the creators of a platform, or for the artists whose art informs a platform’s creations?

One economic reason for granting copyright to the creator of the A.I. platform is that by doing so, it is more likely that companies keep investing in the technology and that the United States becomes a “global leader in promoting A.I. development.”  However, making money off a system that uses other artists’ past work seems concerning.  Therefore, copyright law may soon have to address the issue of A.I.-generated art ripping off the style of human artists.  A middle ground may involve the platforms both working with artists to protect the rights of creators and also restricting what search terms a user can input.  Ultimately, A.I. technology isn’t responsible for the ethical issues, but rather the onus is on the people creating and using the platforms.  An undeniable benefit of A.I. art generators is that it increases people’s access to making the creative ideas in their head a reality.

Copyright laws should be amendable to reflect new legal questions that arise because of innovations in online technologies.  I understand that it is not feasible for copyright laws to address all future legal questions.  However, I believe that it is reasonable to expect that copyright laws may need to go beyond its traditional jurisprudence to address the new legal questions that were not foreseeable prior to the technology’s implementation.  For example, copyright laws evolved from initially protecting writers and painters to adapting to also include photographers, a concept unforeseeable prior to the adoption of cameras.  Therefore, it is not unfathomable for copyright laws to incorporate works from A.I.-generated art platforms.  The way it should do so, however, requires further consideration of potential restraints with licensing and terms of use.  Regardless, I believe that there exists a human element in A.I.-generated art worthy of copyright recognition, whether it be the humans who created the systems, the human’s whose art inspires the systems, or the humans who use the systems for their creative visions.  The conversation regarding A.I.-generated artwork and copyright will likely progress as more people, including companies, use these platforms more frequently.  Thus, it will be interesting to follow whether the USCO one day harmonizes copyright laws to recognize A.I.-generated art platforms and future new technologies.  Although it may currently seem far-fetched, so did once copyrighting photography.

 

Student Bio: Kiara Benac is a second-year law student at Suffolk University Law School.  She is a staffer on the Journal of High Technology Law.  Kiara received a Bachelor of Arts from Wellesley College with a Political Science Major, Economics Minor, and Portuguese concentration.

Disclaimer: The views expressed in this blog are the views of the author alone and do not represent the views of JHTL or Suffolk University Law School. 

Print Friendly, PDF & Email