The future of art: USC’s first AI + art competition


Images generated by competitors Nivedit Balam (left) and Anna Shaposhnik (right) using AI program Midjourney and prompts that began with only a few words. (Nivedit Balam | Ahmanson Lab) (Anna Shaposhnik | Ahmanson Lab)

PhD candidates Matthew Fontaine and Samuel Teets, master’s student Nivedit Balam and junior Anna Shaposhnik were able to produce unbelievable masterpieces with just words, such as “wind,” “glass” and “float,” during the first AI + Art Competition held by the Ahmanson Lab Oct. 19.

With the rise of digital media, it’s no surprise that artificial intelligence can make artistic wonders out of mere thoughts. 

AI art merges technology and the human psyche by interpreting text prompts and utilizing algorithms to produce unique pieces drawn from preexisting media, according to the Curtis Fletcher, Director of the Ahmanson Lab. 

“This technology, for the first time in human history, allows us to create images and videos of anything we could possibly imagine,”  Fletcher said.

The first of these AI Art competitions was hosted by the Ahmanson Lab,  part of the USC Sidney Harman Academy for Polymathic Study. Participants utilized Midjourney, one of many artificial intelligence programs, to generate surreal artwork with text prompts.  

Competitors went head to head to see who could come up with the most impressive illustration. As long as the three assigned words were included in their prompt, players could generate almost any textual description imaginable.

The competition was broken into rounds, each of which began with one competitor having three minutes to develop a prompt including the assigned words and generate four images, ultimately selecting their favorite one to upcale and submit for judging.

After the first player’s generation, opposing players could modify up to half of the previous player’s prompt and generate their own image with the goal of improving on the other’s; modification continued until each player had generated three new prompts and images — three “Volleys” back and forth. 

What began as three flat words could eventually develop into a phrase such as “Horror silhouette windmill rim light backlit foggy moonlight illuminated from above by the aurora borealis wintry spindly derelict haunted spirit capture horror film still painting whispering quiet suspense abstract in the style of greg rutkowski ar [sic] 9:16.” 

After six final pieces were generated for a round, a popular vote on the best image decided the winner. The same process was repeated in the second round with three different words — “weather, “hill” and “fly” — and two new competitors. 

The third and final round was a match between the two winners of the previous rounds, Fontaine and Balam, and consisted of only two volleys with the words “winter,” “book” and “draw.” Somehow, Pikachu came into the mix during the final volley. When it came down to the final competition, Fontaine credited his win to the risks he took by using original art styles of the Deep Dream technique as a modifier. 

“I took a risk during the competition by using that modifier, but I lucked out that it turned into something really good,” Fontaine said. “It makes the style very unique because I’m not copying a famous art style, but I do like blending two artists’ styles over time.”

The competition opened the window to a unique audience-artist interaction not typically seen with AI art generation. Waiting for the images to develop was part of the anticipation, and audience members were able to interpret each image and express their own fascinations and emotions. 

Anna Shaposhnik, a junior majoring in arts, technology, and the business of innovation and a participant in the AI + Art competition, noted the feeling of community the event created. 

“People came together, they created stories around those images,” Shaposhnik said.

There is constant debate over whether or not AI art can be classified as “artistic.” Some argue that art requires originality and physical creation, claiming that AI makes it “almost too easy to generate a really great work of art,” as Fontaine pointed out. 

“So it kind of takes away the skill aspect of learning to be an artist,” Fontaine said. “Some people have argued that you’ve replaced that with taste, because now what matters is your own taste in art. It’s like a flip barn.”  

The AI art community is continuously evolving and expanding. Fletcher commented on the ever-changing landscape of AI, and posited that “the lab is, in some ways, a perfect hub for these discussions to take place… trying to gauge what the community is for this and then  building off of that.” 

Fletcher cites the lab as a place not only for the intersection of different disciplines and interests, but between artists and technologists. The competition is a launching pad for students and staff members interested in what AI has to offer and its possible implications. This event embraced the connection between the human imagination and technological advancements. In doing so, it sparked conversation regarding the ethics and extent to which we should rely on artificial intelligence in art — and further in life. 

Undoubtedly, AI art generation takes its own creativity and careful attention. Shaposhnik explained that specificity and familiarity with artists and their visual styles are the most important part of curating the story you want to tell through a prompt. 

One of the most interesting qualities of AI-generated art is how it allows for “unartistic” people to participate. Though it may appear simple to create a fantastical, dreamlike art piece with the right software, that only makes the deceptively tricky practice more difficult to master. 

 “It by definition lowers the barrier of entry for people who can’t create in [traditional] ways, right?” mentioned Fletcher. “But then it also gives superpowers … there are plenty of artists who are now embracing this.” 

Besides, this is a time period that already embraces remixed, remade content. Shaposhnik supports this new art form as a believer that the inspiration we draw from other images prohibits an artist to be “truly original.” While technology can now replicate and mimic distinct art styles, it also has the ability to remix and hybridize styles to create a completely new aesthetic. 

AI caters more to creativity of the imagination, and thus requires a different skillset to produce aesthetically captivating images.

“It is just shifting the part where you get to be creative,” Fletcher said. “Like I actually can’t produce very good images, because I don’t know, I don’t know the right ways to describe the stuff. Someone with a PhD in English, it turns out, he’s really good at it … So there’s a skill set there.” 

There is no telling what the future looks like for AI-generated art. The possibilities are endless. No matter one’s standpoint on the AI art phenomenon, technology will continue to advance and exceed human capabilities. Mats Borges, the project specialist at the Ahmanson Lab, believes it is crucial to encourage discussion around the ethics and implications of rapidly developing technology because it is already so integrated into our communities. 

While AI-generated media is foreign to many, Fletcher believes it will exponentially improve and develop due to its ability to alter “our appearance, how we sound, and our identity.” 

“In some ways, but not really in real time. AI is going to make it so that we can consistently alter that stuff in real time,” said Fletcher. Reality and technology will continue to converge and blur together, and rather than trying to stop it, people need to be educated on it and weigh its benefits and harms. 

This competition is only the beginning of a future of possibilities for AI art, according to Borges.

 “It’s gonna go everywhere,” Borges said. “It’s going to be part of our lives, like, how photography is now in our cell phones.”