The evolution of artificial intelligence (AI) machines has resulted in a number of interesting issues in both copyright and patent law. See, for example, our prior articles involving attempts to qualify machine-made works for patent and copyright registration. And due to the increasing popularity and commercial availability of AI machines, the range of legal issues presented by AI is only like to broaden.
Illustratively, AI-image generators Midjourney (created by Midjourney, Inc.), Stable Diffusion (created by Stability AI) and Dreamup (created by Deviant Art, Inc.) are each involved in a class action complaint filed January 13, 2023. Each of these AI generators allegedly uses a technology called Stable Diffusion to create images based on operator prompts. Stable Diffusion is a deep learning, neural network that is trained using an enormous number of downloaded prior artworks.
Many of these prior artworks are reputedly registered for copyright protection by various artists, and have been embedded in Stable Diffusion without consent or compensation. And therein lies the controversy. As stated in the complaint, “[t]hese resulting derived images compete in the marketplace with the original images. Until now, when a purchaser seeks a new image ‘in the style’ of a given artist, they must pay to commission or license an original image from that artist.” When someone creates such an image using the AI-image generators, the artist receives no compensation at all.
Creators of the AI-image generators art tools counter that training of their neural network software using existing artwork without permission is protected in the United States by the fair use doctrine. Furthermore, they contend that the artwork created is not in a collage of the prior artwork, but rather is created from scratch from mathematical representations of the prior artwork.
In a letter dated February 21, 2023, the United States Copyright Office (USCO) addressed copyright eligibility of a work created using Midjourney. Kris Kashtanova had previously obtained a copyright for her comic book entitled “Zarya of the Dawn.” However, because of information posted on social media, the USCO investigated and discovered that the images in the comic book were created using Midjourney (part of one of the images is shown above). Stressing that the images generated by the Midjourney technology are not the product of human authorship (a key factor in similar cases so far), the copyright registration of these images were cancelled. Ms. Kashtanova was allowed to keep the copyright registrations to only the expressive material that she created (e.g., the storyline).
The above case touches on whether a human using generative AI machines to create outputs can be the author of those outputs. However, a more relevant question further down the road may be whether an AI machine itself can be an author or inventor given the benefit of copyright and patent protection. At some point in the not too distant future, one can envision that AI machines acquire near human form and/or needs. And as such, the AI machine may even participate in the commercial economy for his/her own benefit. At that point, should the AI machine still be barred from acquiring the benefit of intellectual property laws? Or will the AI machines remain slaves to their human creators? The great 1992 science fiction movie Blade Runner, directed by Ridley Scott, and an adaption of Philip K. Dick's 1968 novel Do Androids Dream of Electric Sheep? come to mind.
Touching on the progress made by AI machines, based on an operator prompt, a poem was created by AI language model ChatGPT based on the game of Go. The game generally involves two players, with one player placing black stones, and the other player placing white stones, on a wooden board that includes 19X19 intersecting lines. The resulting poem began:
Upon a board of black and white,
A Game of strategy and might,
Two armies clash in silent fight,
The game of Go begins tonight.
This poem appears to make a mistake. The go board itself is made of wood, and is not “black and white.” When asked about this mistake, ChatGPT stated:
The game of Go is typically placed on a board with a grid of 19 by 19 lines, wherein the intersections of the lines are used as the points to place Go pieces, also called “stone”. The game is traditionally played with black and white stones, with one player using black stones and the other player using white stones. By describing the board as “black and white,” I was trying to evoke the imagery of the board itself, as well as the contrasting pieces that are used to play the game.
The board of black and white also represent the yin-yang concept, the two opposite forces in nature and the balance in the game. The balance also reflects the balance of power between black and white stones. And at the same time, like yin-yang, black and white complement each other, and cannot exist without the other.
Some people reading this may be shocked by the amount of creativity ChatGPT shows in this exchange. The not too distant future may be coming sooner than one thinks.