Literature Review
Over these past 3-4 years, Generative Artificial Intelligence has seen rapid growth in both capability and usage. GenAI has thoroughly integrated itself in the day to day life of many students and professionals, from generating text to art. Because of such widespread use of GenAI, the ethics surrounding its usage, especially when it comes to art, are often called into question by many. While some see it as entirely harmless, others,in what seems to be a majority, view it as theft and a blight on human creativity. Ultimately the question boils down to “To what extent is the use of AI generated art acceptable, if at all?”. To examine this ethical dilemma, along with taking a look at how exactly a model like Gemini is trained, a focus must also be placed on the potential consequences of the widespread use of AI generated content.
The core of the argument for AI art being restricted to personal use (generating wallpapers, adding to presentations etc) lies in how a GenAI model like Google’s Gemini is trained. For models Google Gemini and the well known ChatGPT, large data sets comprising information from all over the internet, ranging from articles to art created by talented professionals, are fed to the AI. The model then learns from this vast array of information so that it can produce appropriate responses when prompted. Artists have brought many lawsuits alleging “companies trained AI tools using data lakes with thousands — or even many millions — of unlicensed works ” (Appel, et al). By using AI art to publicly promote something or make content for monetary gain it can be argued, as the artists do from the lawsuits detailed in this article, that one would be directly contributing to art theft and profiting off another’s work. This then brings up the question of copyright infringement, with courts ruling that “ If…AI’s works are unauthorized and derivative, substantial infringement penalties can apply”(Appel, et al). While it may not necessarily be the case that each individual image is an unauthorized derivative, it is difficult to differentiate when an AI image was trained on unauthorized data versus when it is. Because of this copyright “gray area” as some might call it, it is generally recommended that art produced by generative AI be restricted to personal use to avoid the ethical question of whether or not the user is gaining from stolen works.
Another facet of the ethical dilemma are the potential consequences of the widespread use of AI art/content. While part of the debate surrounding the ethics of AI art is the aforementioned issues of stolen art for training and copyright, an equally important aspect is the impact AI art can have on the world as whole, with particular emphasis on misinformation, deep fakes, etc… Because of the robust ability of generative artificial intelligence when it comes to not only art but content creation in general, “the ability to prepare an agenda via disinformation and false narrative building increases”(Vyas 4.1). While the majority of people who would engage with an art creation tool wouldn’t use it for the express purpose of spreading misinformation, using such a tool beyond personal use could unintentionally contribute to the current misinformation epidemic plaguing the broader internet. As to not violate the honesty principle when it comes to the ethical use of any tool, the only real way to use AI art in an ethical manner would be for personal use. “AI aiding the mass generation of low-quality yet plausible content” (Vyas 4) is one of the core reasons as to why using AI art ethically can be difficult in the first place. Taking this academic article’s argument in tandem with the first source which talked about copyright issues, AI art potentially violates multiple ethical principles. Restricting use of AI art to be purely personal avoids these ethical issues. Beyond simply just misinformation lies the issue of Deepfakes (using real people in AI images or content) in order to “produce counterfeit content” amplifying “the risk of privacy violations and identity theft” (Al‑kfairy, et al 4.2.4). While criminal deepfakes are a more extreme topic when it comes to ethics in AI art and AI in general, the underlying principle of the unethical use of AI art still stands. This once again highlights how AI generated art/content can violate the honesty principle of ethics as shown in the previous source. Honesty is ultimately the glue that keeps society together, when this honesty is broken people begin to distrust each other implicitly which only serves to harm both society and the individual. “Deepfakes can serve as a potent tool for spreading misinformation, further muddying the waters of fact and fiction.”(Al‑kfairy, et al 4.2.4). When taking a look at each of these three sources, each presents a different manner in which the widespread use of AI art can be unethical. In addition to violating the honesty principle, it could potentially violate copyright as well or the justice/fairness principle. Taking into consideration the ethical dilemma that is presented by these sources, the only true way to use AI art in an ethical manner would be to restrict it for personal use.
When utilizing the instructions for GenAI art on this website, or when using AI art in general, it is important to keep in mind the ethical concerns discussed so far. Engaging with any tool or service ethically is ultimately what keeps society functioning, and AI art generation is simply just another tool that must be used in the same ethical manner.