How much ‘art’ is there in artificial intelligence? Photo: imaginima, Getty
“When I first heard of AI imaging tools, I actually was scared. I mean, it's exactly what clients would want—work that is quick, cheap, and quality. AI would win over any of us freelancers,” Risa Barcelona, a freelance graphic designer based in Manila, Philippines, told VICE.Artificial intelligence systems that generate vivid images in response to simple text or visual prompts have cropped up in recent years. Tools like DALL-E and Midjourney, for example, can create everything from absurd hypotheticals and porn to realistic faces of fake people and self-portraits in a matter of seconds. It even made award-winning art, pissing off artists and prompting one Twitter user to say “We’re watching the death of artistry unfold before our eyes.”
AI champions and concerned creatives say these imagery systems could displace graphic designers, illustrators, and other professionals in the creative industry. As Barcelona said, the systems produce work more quickly and inexpensively than a human designer could, with arguably comparable quality. But this conclusion may be reductive, as it neglects the intangible but principal role of human creativity in making art.
These tools work by “tapping into the internet’s subconscious” and using massive datasets of text and images to illustrate whatever prompts are fed into them. The results are often bizarre yet eerily accurate and realistic. But there are drawbacks.One thing AI-generated images have been criticized for is perpetuating sexist and racist conventions. For example, including prompts like “CEO” might generate more images of white-passing men, while using “personal assistant” may generate more images of women. In the project’s documentation on GitHub, OpenAI, the research lab that created DALL-E, said that “models like DALL-E 2 could be used to generate a wide range of deceptive and otherwise harmful content” and that the system “inherits various biases from its training data, and its outputs sometimes reinforce societal stereotypes.”
AI systems can also be used to produce or support fake news. In September, for example, a tweet with photos of houses that were said to have been flooded by a typhoon was reportedly retweeted thousands of times, only for the person who posted it to admit that the photos were fake and generated with AI.Copyright concerns currently shroud AI-generated art because the systems mine original art from the internet without credit or compensation to the artists. This means that AI-generated images will be hard-pressed to replace things like digital ads, because design and advertising agencies and regulating bodies would likely not be inclined to take the risks of being sued for producing unoriginal work for their clients.
“I think it's a great tool, internally, when brainstorming or just sharing visuals to friends,” said Barcelona, who has tried generating AI art but hasn’t fully incorporated it into her creative process. “But it can never be published and claimed as your work of art, because at the end of the day, is it really yours?”In a project he calls Robomojo, Melbourne-based audio-visual artist known as Vincenzi uses AI imaging systems to generate new posters for popular films.“I'll feed the film title into various AI tools, or sometimes just the premise, and see what it generates. Often the results are horrible but sometimes they're genius, so it's always fascinating to see where the AI will take an idea,” said Vincenzi.The artist takes the results and refines them himself, adding text for the movie titles and credits, a feature that is currently out of the AI systems’ capabilities. The results lean into the ways the titles and plots of the films can be misinterpreted rather than provide accurate representations, said Vincenzi. For example, one of Robomojo’s versions of the poster for the movie The Human Centipede features a perturbing image of a man with centipedes appearing to grow out of his body. Its take on Dirty Dancing, on the other hand, shows two pigs standing on their hind legs, dancing.
While the results are absurd, they do make a point about the utility of AI systems. Barcelona, the freelance graphic designer, said that the systems can be especially helpful for people who want to communicate their ideas visually instead of through words. It’s like telling someone about a dream—but instead of telling them, you’re showing them. As Robomojo’s film posters show, AI systems can also produce images with unconventional and unexpected points of view. “At the moment, AI is fast becoming a required tool for creative professionals, particularly if they can use it to assist in large volumes of work,” said Vincenzi. “There's going to be an explosion of output very soon, most of it really bad, but new specks of beauty to be found within that mass detritus.”As AI technology evolves, so do its applications. People can now use AI to generate videos from text, a development that Vincenzi thinks will significantly impact social media.“An Explore page could potentially reveal options to watch new reimagined memories with friends and family. We'd have the option to explore and view a parallel universe of us having started a family with a long-lost friend from high school. And watch a video of our grandparents meeting for the first time, which was never actually recorded. And then of course there will be the option to watch potential future experiences,” he said. “Beyond the ethical implications, the influence this technology could eventually have on our lives is profound.”
AI is also being used to compose music. “Bored With This Desire To Get Ripped,” for example, is a hilarious AI-generated song whose lyrics are auto-populated by a combination of Morrissey's words and customer reviews of P90X on Amazon. Some songs even sound like they were recorded by well-known artists, although the artists never actually recorded them. One of these is “Deliverance Rides,” an amusingly accurate fake Metallica song made using an AI bot.“The creative industries of the future will likely be AI-led, which may be somewhat depressing,” said Vincenzi. “But if you're a creative thinker then you'll be in a better position to harness it to your advantage.”
For Vincenzi, the potential technological shift marked by AI-generated art is not one that necessarily does away with people, but provides a new appreciation for people’s creative process. A common critique of AI-generated images at the moment, said Vincenzi, is that they all have quite the same look. That’s obviously no good for artists trying to express unique visions and differentiate themselves from other creatives. “But this only emphasizes that, first and foremost, the key point of difference is having a strong guiding idea. Now that anyone can make a high-end image with little effort, it's those with a creative mind and developed ideas that will be able to captain their ship to exciting unchartered waters.”
“The creative industries of the future will likely be AI-led, which may be somewhat depressing. But if you're a creative thinker then you'll be in a better position to harness it to your advantage.”
Therein lies what for Barcelona is the inherent advantage of human creatives over AI systems. “It's your originality and personality that make an artwork great, and in my opinion, AI can never replace that,” she said.But even if AI could one day mimic an artist’s originality, Barcelona questioned whether creatives would have anything left to enjoy. “When creating the work, as an artist, the process really matters, and if you cut the process to inputting text and getting the result right away, then where's the fun in that?”Follow Romano Santos on Instagram.