Blurring the Lines: When AI Creates Art Is It Copyrightable?
Stan Gibson and Jessica Newman
In October of 2018, Christie’s sold the work Portrait of Edmond de Belamy for $432,500. None of this, including the price, would be notable if not for the claimed artist. The claimed artist was not a person but an algorithm, bearing its creator’s signature:
Obvious, a Paris-based collective, created the work using what is known as Generative Adversarial Networks, or “GANs.” Creating works using GANs requires inputting thousands of images as well as developing and making adjustments to the algorithm.
The sale, which garnered attention and outrage alike, served as a flashpoint in the discussion of artificial intelligence and art. As Ian Bogost of The Atlantic wrote, “[t]he image was created by an algorithm the artists didn’t write, trained on an ‘Old Masters’ image set they also didn’t create.”
In the aftermath of this sale, Sotheby’s followed suit with the sale of Mario Klingemann’s Memories of Passerby I (2018) in March of 2019. The work, which sold online for roughly $42,000, uses multiple GANs trained on thousands of portraits from the 17th to 19th century to generate an infinite stream of real-time portraits for the viewer. Klingemann refers to the neural networks used to create an endless stream of new images as “the brushes that I’ve learned to use.” Memories of Passerby I may not have generated the same buzz as Portrait of Edmond de Belamy, but it raises many of the same questions.
Although the use of technology to create art is nothing new, GANs, and similar neural network software, increasingly blur the lines when it comes to creation, authorship and ownership. The use of technology such as GANs to create art results in works that are not authored by individuals in the traditional sense. The question then becomes, are these works copyrightable under U.S. Copyright laws, and if so, to whom does the copyright belong? Continue reading