Code to Canvas
Creativity and controversy in generative AI art
—By Sofie Markle
Artificial Intelligence (AI) has emerged as a force in our world, changing technology, education and science as we know it. It’s a topic that is simultaneously fascinating and terrifying. In particular, AI’s impact on the art world is sparking debates about creativity, authorship, and the future of artistic expression. Part of what it means to be human is our ability to build skills that can be used to discover and create. For artists, these skills can take decades to develop. Countless hours spent building techniques and a personalized style. Now all this can be done with a click of a button through Generative AI art. As AI-generated art gains prominence, it’s crucial to explore both the opportunities it offers and the challenges it poses for traditional ideas of art and originality.
The world of Generative AI
At this point, many of us are familiar with programs such as ChatGPT. You ask the computer to respond to a question and you’re answered within seconds. It can summarize novels, write poetry and even make an illustration of a studious University of Michigan cat.
What makes Generative AI different from other forms of AI? Generative AI is a form of artificial intelligence that is focused “entirely on content creation,” centered around creating new content such as images, music, text and other forms of media. Generative AI systems are trained on large data sets, learning how to form new content through identifying patterns and creating new variations from them. Hence, content created through Generative AI produces more realistic results and resembles more human features.
The presence of AI in the artistic world began in the late 1960s with experiments through the use of algorithms to produce visual art. Throughout the next several decades, advances in computer graphics and digital tools began to advance, but it wasn’t until the 2010s that AI art became more mainstream. Machine learning models were introduced, such as Generative Adversarial Networks (GANs), changed the way machine art could be created. GANs have two networks that compete against each other: a generator and discriminator. The generator learns to create new data from samples and the discriminator distinguishes what data from the generator is real or fake. This built the framework for Generative AI and changed the way machine art could be created. Fast forward to 2018, an artwork generated by AI, “Edmond de Belamy,” was sold at an auction for $432,500. The algorithm was given 15,000 portraits between the 14th and 20th century. The “painting” departs from the human concept of what a portrait should be, making the final result appear almost contemporary. This was a turning point in the art world, raising questions about ownership and copyright, as well as challenging the ideas of creativity and artistry.
Creative beings without conscious
AI can produce work that pushes boundaries, with designs that are innovative and unique, signs that may point towards resemblance of human creativity. Along with its ability to rapidly produce such ideas, one might consider it a creative power. However, a machine’s “creativity” ultimately differs from that of a person. It lacks the conscious intent and emotional depth. Asad Khan, a computer science major who holds an interest in artificial intelligence, agrees that AI is not complex enough to produce at the level of humans. “Humans are distinct from AI in the nature that they can break down concepts into multiple levels of abstraction and find novelty that is not super intuitive.” However, he believes that it may be possible in the future with a level of AI known as Artificial General Intelligence (AGI). This hypothetical intelligence could match or surpass human cognitive abilities and performance. Khan says, “I believe at this point, AI will be able to produce novelty from observing the world at a rate as good, if not better than humans.”
Many artists view AI as a tool for creativity, rather than a creative being itself. Refik Anadol is a reputable artist who utilizes this new media to his creative abilities. His work explores relationships between architecture, media arts and machine intelligence. He stretches the creative potential of AI and the digital space through stunning visuals that entirely transform spaces. Constantly in motion, giving the observer new colors and shapes to watch endlessly. During his lecture at the Penny Stamps Lecture Series, AI with a Thinking Brush, he discusses his desire to involve machines in his work: “There is a world we don’t see. We perceive. And there’s a world where we can use machines to reconstruct new ways of seeing.”
AI could also be used as a tool to help artists in processes that can be very tedious. Kat Callahan, a student of University of Michigan’s STAMPS, School of Art and Design, has recently worked with AI as part of an internship this past summer. She used AI art, or AI “mock-ups” as they called it, to present a quick draft to clients. It saved the working artists a lot of time from having to start from scratch on other programs such as Photoshop, which can be very time intensive. She doesn’t think it changed her way of working creatively, but said it made the process much faster and easier, especially for something that was just an idea and meant to provide a quick visual aid. “Being a creative person, you can come up with ideas pretty easily. Even when I’m stumped, I know that an idea is going to come and then you can take that idea and refine it and then more ideas will come. I’m not really anxious about my creative abilities so it didn’t change anything.”
Machine thieves or innovators?
As AI-generated works continue to gain popularity, debates and controversy also rise. While it may seem exciting to make something in the style of Vincent van Gogh, a 19th century painter, doing so in the style of a living artist raises more complex ethical questions. These AI-generated works use a practice known as data scraping, meaning they use large amounts of image and artwork through the internet. This includes work from modern artists, who may not be aware that their work is being used to train AI models. In her article, The TikTok challenge: How AI art won social media, Neelam Tailor speaks to artists at the forefront of the AI art trends. “Some platforms obtain permission from every artist included in their sets, but the most popular programs, Midjourney and DALL-E, were both trained on LAION-5B—a dataset of 5 billion images downloaded from the internet without their creators’ consent.” Many artists have begun to notice stolen work in these programs. Even smaller artists, such as illustrator Julia Bausenhardt, found countless examples of her work on algorithms being used by the AI model, Stable Diffusion. In her blog, she explains why this was a very scary moment for her: “As an artist, the intellectual property right is the most important right I have. If I can’t manage the rights to my art and decide who is allowed to do what with it, it becomes worthless.”
Copyright law is a crucial part of protecting artists, giving them control over the distribution of their work. AI creations that are heavily influenced or resembled the style of a specific artist could be seen as derivative work. Under copyright law, derivative work would need permission from the original creator. However, AI may be using combined information from several sources. And as stated on The US Copyright Office page, “[First] copyright protects original works of authorship, including original pictorial, graphic, and sculptural artwork. A work is original if it is independently created and sufficiently creative.” The use of the word “original” indicates that pieces drawing inspiration from a particular work or style, but not a replica, would make it a new composition.
This challenge of determining who gets credit could have a significant impact on artists financially. Many artists gain income from licensing their images for commercial use. An artist’s style is a key part of their branding. However, if AI art can replicate products using their style, then companies may prefer to use AI rather than the artists, especially in fields like graphic design and illustration. An artist’s style is not protected by traditional copyright laws; however there are several initiatives pertaining to protecting those with copyrighted works. In 2023, the Copyright Office began an initiative to examine copyright law and policy issues pertaining to AI. Currently, the Office has released Part 1: Digital Replicas, providing legal clarity on works that mimic human creators. There is of course a long way to go, and this subject does not only apply to artists. Khan says that current solutions to advancing these machine learning (ML) models is to throw in as much data as possible. “This can challenge the state of the art in performance, but there is obviously an inherent cost to this strategy.” A great deal of this data contains sensitive information, including private user data. Although there are initiatives to reduce this cost, much of our privacy has already been invaded and used in current ML models.
The uncertain future
Generative AI art is a fascinating intersection between technology and creativity that challenges the artistic process. While AI is expanding the world of art, enabling access to artistic tools to the larger population, it may have a significant impact on living artists. There are many unanswered questions. And while it’s critical to ask ethical questions and find solutions, we must also adapt to the new technologies being presented. The world of AI art will continue to evolve, prompting us to find new solutions and rethink what art is and what it means to be an artist. However, nothing can truly replicate the authenticity and depth of human creativity. As Callahan aptly puts it, “There’s no artistry to it [AI art], no human touch to it.”
Cover Photo: Magic Studio generated image; text prompt: “Artificial Intelligence”