AI has these educators weary, and hopeful

AI-generated image created with Stable Diffusion distributed on Twitter. Courtesy of Alex J. Bowers.

When news hit Twitter that the winner of a Digital Arts competition at the Colorado State Fair used an AI-generated image for his submission, emotions ran high. 

Many called Jason M. Allen, the man behind the AI image, a cheater. Some groaned of the existential danger AI meant for artists. The outcry on social media led to news outlets such as VICE and The New York Times picking up the story. But when Holyoke University Professor of Digital Art and Design Marianna Dixon Williams heard the news, she laughed.

“I just find it funny on a personal level,” she said. 

Williams was originally trained as a classical painter in art school, though her work has varied greatly since then. Her early experience with art was learning to paint landscapes and portraits, honing her skills to make them as lifelike and beautiful as possible, because that’s what she thought the goal of art was. She said it wasn’t until she was in art school that she had an epiphany: Art has greater meaning beyond aesthetical beauty.

“I think a lot of people say, ‘That’s beautiful. It’s a great piece of art,’” Williams said. “But if you ask people: What’s at stake in this work? Why are we making this? What is the meaning of the images we’re taking in? People are like, ‘I don’t know. I don’t really draw.’ I just find it funny.”

Williams categorizes herself as a New Media artist, her work combines traditional art forms with emerging technologies to simulate environmental loss and growth in areas like the Arctic Circle. She still paints, but over the years she’s expanded her tools far beyond the brush. She implements animation, architectural fabrication, 3D graphics, and any other technology she can find to expand her art.

“I think I just had this dissatisfaction in my work,” Williams said about experimenting with tech. “I have a deep love for painting. But in my own practice, in the search to create real experiences, I found that I needed to learn a lot of digital tools.”

In the classroom, she offers workshops on ways to incorporate technology into art, with tools like Adobe products, 3D modeling and coding programs. Recently, she began showing her students the potential of AI, both through image generation and other more technically advanced tools. Now she says her classes are attracting students who would normally never explore the arts.

“I’m getting more art students in my class who wouldn’t have normally taken a drawing class,” Williams said. “I’m getting a lot of STEM students, and a lot of students who say ‘I really love art, but my family doesn’t want me to be an artist.’ They can come to their families and say, ‘Oh my gosh, I learned Python in my art class!’” 

Python is one of the most commonly used computer programming languages.

Machine-learning programs that generate images, commonly referred to as “AI Art” on social media, are a rapidly advancing technology that is effortlessly simple for its users regardless of their experience in art or coding. Using text descriptions of the images, the machine scours its available datasets, cutting up and combining sometimes hundreds of images together, adding in any stylistic effects the user requests. Users can also upload their own images for the AI to edit, and becoming familiar with the language AI uses to categorize its creations can result in extremely high quality and stylish art pieces. However, most AI generated images are taken from images publicly available on the internet, including the work of other artists. AI manipulating the work of other artists without permission has caused many to label the technology as art theft.

The method these machine learning algorithms use to generate images has sparked discourse over what rights an artist has over their work when released publicly. Many programs are sophisticated enough where the AI is capable of replicating an artist’s exact style, such as what happened with Deb lee. In a TikTok video shared by the artist, they claim that someone took their artwork, used an AI program to modify and recreate their style into a new image, and distributed the new artwork across social media without crediting Lee’s original work. Stories like Lee’s highlight how easy it is for users to recreate artists’ work, and create resentment of both the AI and its users.

Recently, new programs like DALL-E, Midjourney and Stable Diffusion were made available to the public. These programs create high quality images generated entirely by machine-learning, and they have been a source of both awe and controversy. While fear and anger from artists run rampant on social media, artists and educators like Williams are encouraging their colleagues to embrace AI in their work.

With a new generation of artists raised with rapidly evolving tech coming into the field, AI will likely become very prominent in the art and media we consume. Williams said that while fear is a normal reaction, she doesn’t want that fear to close people off to the potential that AI has for artists. Text-to-image programs are certainly the most popular use of machine learning, but it is far from the only artificial intelligence used in the arts.

The Boston Museum of Science opened a series of exhibits this past September that showcased the role AI tech currently has, and will continue to have, in our society. One of these pieces, titled “Data Choreographics”, is an interactive exhibit that invites visitors to dance and have their movement recorded. The installations then generate images based on the users’ recorded movements, displaying how AI processes data in a form comprehensible to the human eye.

Williams says works like Data Choreographics wouldn’t be possible without machine learning.

“It had to have been an AI piece because it was both a critique of AI and something that reflects how we live and move through the world,” Williams said.

For consumers and artists alike, this technology can expedite laborious, time-consuming work, but Williams is well aware of the possible impacts of overreliance on AI. Although for Williams, the greatest concern is how AI could stifle creativity.

“If we trust the AI too much to determine value for us, are we still thinking for ourselves? If we only rely on AI systems, will it make everything really generic and streamlined?” Williams said. “I would hate for artists to feel pushed into one mode of making art where AI has to be part of your work.”

While the technology has been in the works for years, it was not until this past year that a massive anti-AI art movement formed online. However, there are experts in the digital arts who see this technology as a natural progression for tech in art.

Bobby “AB” Osborne is a professor at Augusta University teaching animation who has worked in the animation industry for the past decade, and wasn’t surprised when the AI art trend went viral.

“I worked previously at an engineering place that used a lot of artificial intelligence machine learning technologies,” Osborne said. “They were using it back in 2015. So I knew the wave was going to happen.

In Osborne’s art class, the discussion around AI art became a hot button topic. He found students were not resistant to the technology, but were dismissive of the quality of art produced.

“It could be good art or bad art, that’s up to whoever is paying for it,” Osborne says. “But the idea that it could potentially give artists a new payment method, a new way to not be a starving artist. To me that always seemed kind of exciting.”

Osborne compares the rise of AI tech in art to photography, digital art and photoshop. Each medium stirred controversy and fear, similar to the response toward AI art, on how their introduction will radically shift the art world. Osborne doesn’t think the reality is as dramatic, but art students will need to decide what tech tools they want to bring into their practice.

“Once students graduate, they’re not necessarily struggling because of artificial intelligence. They just never develop the skill sets that were needed anyways,” Osborne says. “I always try to throw that out there. Yeah, you can draw. What other skills do you have in the toolbox?”

Machine-learning algorithms have an impressive ability to replicate an artist’s style and technique, but users shouldn’t expect the technology to meet exact expectations for every use. Osborne sees AI as a useful starting tool, but terribly behind when trying to create anything innovative. For that reason, Osborne says, artists will be safe as long as they adapt.

“For artificial intelligence to take over our career, the client would have to accurately describe what they want. So we’re safe! The client never knows what they want,” Osborne jokes.

Commodification of art through AI runs could do more than just add uninspired art into the world. Image based machine learning algorithms are rife with biases ingrained in the data. Users of AI art programs usually find the algorithm favors white men when generating images of people. Oftentimes white men are also displayed in positions of power, such as an employer or leader of a group, when the generated image contains multiple people of various gender and race.

Alex Bowers, Professor of Data Science at Columbia University, sees those issues as the biggest priority to develop and correct as the technology moves forward.

“Every time we have a technology of this kind that then gets built into our workflows, then those biases become systemic,” Bowers said. “There needs to be a deeper conversation about whom are these technologies for? Where do the images come from? Whose work?”

Bowers is not just concerned with the influence this tech has on art. As machine learning algorithms have access to any image available online, that means a person’s photos, voices and likeness are available for AI use.

“If I’m represented in images that are in the data set, or if some of my images that I have taken on my phone or cameras and put online are ingested into the training data set, what are my rights?” Bowers said.

Currently, most of these questions don’t have any answers in the law. While no privacy laws regarding AI have been made, some precedent has been established in regards to copyright. The U.S Copyright Office has historically denied attempts to register images generated by a machine as copyrighted material due to the images lacking the required human authorship necessary to sustain a claim in copyright.

“The whole question of ‘is AI image generation art or not?’ kind of misses the boat,” Bowers said. “I think these issues of privacy rights and bias are the big ones that really need to be addressed because the technology is just going to keep moving forward.”

Proponents of AI art say the technology will open the doors for a variety of people to get into the art world. The ease of use of the programs is an easy incentive to draw new users in, and the speed at which the programs produce images makes the technology gratifyingly addicting. However, the early limitations of the artform are already apparent.

The Harlem CoLab is a game company that also runs technology tutoring sessions for high school and college students. The company creates games that center around  telling stories about societal issues such as economic inequality, climate change or racial discrimination. The company also runs student workshops to expand digital literacy and tech knowledge in traditionally underserved communities in Harlem.

Dennis Morgan, CEO of the Harlem CoLab and part time instructor at Columbia Teachers College, thinks that despite all the potential AI has, there are many barriers of access for students to properly utilize the tech. 

“Digital literacy, I think, is the real challenge,” Morgan said. “And knowing what digital tools to use is really important to being successful in the digital economy that we have now.” 

While proponents of AI art say the technology will open doors for many prospective artists, not everyone will find AI easy to use. Access to a computer, stable internet and enough tech savvy to effectively use the AI are all barriers Morgan says keeps most people from using AI in creative and worthwhile ways. Most AI image generation companies offer a free trial of the service, or give new users a handful of free image generations upon creating an account. After those are used up however, users will usually need to purchase a subscription plan to continue utilizing the AI. These limits, Morgan says, are especially felt by communities lacking the means or resources to access the technology

“Once you have the skills and you understand the technology, you can roll with it,” Morgan said. “But I feel like in underrepresented and under-resourced communities, this stuff is way too far.”  

Beyond the technological limits, utilizing AI effectively often requires comprehensive knowledge of how programmers categorize the data used by AI. For those lacking proficient tech education, AI may not be the entry point for new artists that proponents claim.

At Columbia Teachers College, Morgan teaches game design and runs student led workshops around ways to incorporate new technologies into students’ workflows. Recently, he began running workshops around AI for his students to use in their own projects.

“I’ve seen a couple other students who have been able to put together some game projects using it,” Morgan said. “I’ve seen some people using it for digital environments in early prototypes.” Morgan uses AI art himself for his own work during early production. Early in the production of his own games, he uses AI software for brainstorming potential themes and aesthetics of his products. He also uses AI as a storyboarding tool, creating quick visual representations of his product and the overall message of his games.

However, Morgan says that he would never fully substitute a human artist for an AI when the time comes to sell his work. For Morgan, his involvement with AI ends after the preliminary phases of production, and he would employ a human artist to create all assets for his products. When asked why, he compared the artistic worth of AI art to music sampling in rap music, and how he would feel about AI generated music. The difference, for him, is the human component.

“I’m a bit of a purist,” Morgan said. “I’d rather hear the struggles of a person. If someone told me a computer made that song, I would say, ‘Thank you for letting me know, I won’t be listening to that.’ I’d rather have a human story than a computer story.”

AI generated images are certainly impressive, and will only improve in the future. It remains to be seen what the long-standing impact this technology will have on our art, but educators like Morgan and Williams are hopeful that artists themselves will find a lot of value in the technology.

“If AI’s getting more people into my class and giving them more stable jobs so they can have their practice, great,” Williams said. “If it’s speeding up their workflow, great. But if it’s also de-emphasizing the content and the purpose of their work, then I think that’s a real problem.”