In a letter to Helen Keller, Mark Twain famously writes:
“For substantially all ideas are second-hand, consciously and unconsciously drawn from a million outside sources, and daily used by the garnerer with a pride and satisfaction born of the superstition that he originated them.”
If one were to take Twain’s word for it, there is nothing original about any human creation, and all products of human creativity are derived from their predecessors. So too can be said about art that is not human. If no human art is original, one must not take issue with the fact that AI also references millions of instances of human art to generate new art. If there is no such thing as artistic originality, then all arguments that seek to undermine AI art for being phony and derivative cease to hold water. If human intelligence can get “inspired” by other art, why is it that art made with artificial intelligence is accused of “theft” of intellectual property?
These are a few of the arguments that proponents of Generative AI might posit in favour of technology that uses human art as training data to produce an infinite number of images given a single written prompt. And they would seem valid if we did not know that Artificial Intelligence does in fact heavily rely on extensive and well-documented sets of training data available for scrutiny and analysis, whereas the same cannot be said about human intelligence. The murkiness of the human psyche must then provide for the benefit of the doubt to human creators, until the sources of their creativity can be made more evident. Until the claim that human creativity works in the same way as artificial intelligence can be proven true or false. Until the apologists of uncredited data usage for Generative AI have a way to prove that human intelligence adds nothing of value to the source it derives from, throughout successive iterations and regenerations. That Shakespeare added nothing to Arthur Brooke’s The Tragical History of Romeus and Juliet when he based Romeo and Juliet on it. That the several versions of the oral Indian epics have in fact added nothing of note to the original’s depiction of the human experience.
At the outset of this article, it seems necessary to declare its underlying assumption: that people accept and recognize the value of human-made art over AI-generated alternatives. The difference between the two mainly lies in the circumstances of their production and the manner of their consumption. AI art has no actual intellectual intentionality and all readings of authorial intention must make way for an apathy that only values art for its vapid utility - that it fills space and does little more. The skill, patience, and investment one would require to make art undoubtedly surpass the expertise required to write an adequate prompt.
“Good Artists Copy, Great Artists Steal”
However, some would claim that there is definitely something to be said about AI art and its merits and it has found many legal sympathisers in the past few years. In 2022, Jason Michael Allen entered and won the Colorado State Fair’s yearly competition and submitted an image that he created with the help of Midjourney, called Théâtre D'opéra Spatial (Space Opera Theater). In 2023, The United States Copyright Office denied it copyright protection claiming that human authorship was the primary requirement for any art to be considered under the copyright law. Allen was dissatisfied with the court’s ruling and claimed that the amount of work he had put into this image was “far from de minimis”, having revised his prompts over 624 times and having altered the image hence produced with the help of Adobe Photoshop and the Gigapixel AI tool. What Allen does not address however is that his work would have been impossible without the contribution of several artists whose art Midjourney’s generative model uses, and by demanding copyright protection for his work, he denies the role of those artists in his creation, essentially stealing their work and claiming it as his own.
Source: Wikimedia Commons
Irrespective of one’s stand on AI art and its fairness, one cannot deny that it is everywhere and it is becoming more difficult by the day to tell AI art from its human-made alternatives because that is the end goal of any advancement in artificial intelligence - to be able to blur the line between human and machine capabilities. The machine, however, does not exist without the human - and it uses human intelligence to be able to do what it does. It seems unjust, therefore, that artists should make no compensation at all for their art which can be freely used to generate AI art for commercial purposes.
What Do We Owe Our Artists?
In 2023, Wang et al. tried to quantify the value of art with a WTP (Willingness to Pay) index and found that, when it comes to AI art, consumers tend to prefer art that has been created with the help of prompts that invoke the name of a human artist (provided the artist is well-known enough to matter). They substantiate their claim that consumers are willing to pay more when they know that said artists are receiving a royalty. A commercial motivation to regulate AI art is not satisfactory and does nothing to demystify the question of intellectual property rights and consent. These findings do however serve to support the claim that human contribution adds to aestheticism - ambiguous as it may be as a quality. And since human contribution is valuable and indispensable, it should receive monetary compensation.
Inclusion of an artist’s work in training datasets used by generative AI therefore needs to be governed by principles that would allow for some agency on the part of the artist in question. This can be achieved in two ways:
To begin with, an artist should be able to revoke permission for it to be used in a dataset.
If they allow it,
they could be paid upfront for their contribution to training the model, or
they could be paid according to the value that their art provides to the dataset.
The latter is of course harder to quantify because black box AI models typically do not and cannot reveal which instance of data it is more heavily reliant on for each pass. Some may even claim that the compensation, once divided among all the contributors, will be insignificant. Hypothetical valuations, according to Pasquale et al., serve to prove that although the compensation may be small, it is in no way trivial and it is also suggested that these amounts may be paid annually instead of just once.
The Plague of Plagiarism
Fair compensation for artists may seem like a new question that has arisen in the wake of AI, but in truth, the question has always been there and the legal frameworks determining fair use are far from adequate, and the advent of AI will hopefully draw attention to these matters of intellectual property and rightful ownership.
Of late, two significant unions in Hollywood - Writers’ Guild of America (WGA) and Screen Actors Guild - American Federation of Television and Radio Artists (SAG - AFTRA) demanded policies that afford them informed consent and fair compensation for AI-generated art. This is not a niche question invoked by critics of AI to try and inhibit the inevitable advancement in AI by head-shaking speculations and idle intellectualisation, but a subject that entire livelihoods depend upon. Policies that regulate the use of AI and govern artistic creation using Generative AI are the need of the hour and need to be implemented before it is too late. We need to find a way for human originality to coexist with artificial intelligence before we start depending on AI too much and human art loses all its patrons. We need to pay the Piper before AI drowns in its own outputs and takes our reality with it.
Comments