South by Southwest is known for showcasing creativity: in film, music, comedy and other forms of art. But with this year's conference focusing heavily on generative AI, there was no better place to see the collision of human creativity and machine creation up close.
What did I find? Technologists who see machine learning making it easier for people to bring ideas to reality. Artists who see AI as an existential threat to their craft's already precarious financial stability -- and a shortcut through vital steps in the creative process.
ai-atlas-tag.png
There's more, though: Artists who see AI as another tool in the palette, a way of experimenting with new forms of art that we haven't seen before.
"The best-case scenario is that we enter a time when people are collaborating and creating in ways that we couldn't imagine before," said Holly Herndon, a musician and artist who has pioneered the use of machine learning to create new art forms.
That's one possible future, but there's still much more to be seen before we know what AI means for creativity. Here's a glimpse.
Gen AI as an art tool
Generative AI can turn ideas into reality faster and more efficiently than a person might be able to do. But there's one key step it's missing, said Hannah Elsakr, global head of new business ventures at Adobe and founder of the company's Firefly generative AI for enterprise.
"The models don't have ideas," Elsakr said. "Only we have ideas."
Tools like Adobe's Firefly are just that, tools. Elsakr compared the advent of AI image generators to the launch of Adobe Photoshop, which changed how visual artists did their work but didn't take the creativity out of it. What AI tools can do, she said, is make it easier to turn good ideas into art.
"I believe that AI actually is lowering the entry point for all of us to become creative," she said. "I'm not the best hand-drawn artist but now I can envision things in my mind's eye by using Firefly as a tool."
A similar concept lies behind a tool called SoCreate, which I found on the floor of the SXSW Expo. The app allows users to upload a story script text and images and the AI model will turn it into something that resembles a cross between a slideshow and an audiobook. At SoCreate's booth, AI-generated pictures of pirates or cars racing around the streets of Los Angeles aligned with AI-generated voices reading the user's scripts. It's not turning your ideas into a movie but turning your words into something a little more audiovisually appealing than words on a page.
Founder and CEO Justin Couto said the goal is to provide people who don't have the tools or skills to make films with a way to see their ideas come to life, just a bit.
"They just don't have the equipment or even traditional skills to do that," he told me. "But they have the skills to tell amazing stories and so we're enabling those storytellers to see their visions come alive."
Also on the Expo floor was an app called Soundraw, which allows you to choose from some musical presets to create a song. I tinkered around with it, tapping a few buttons to get an AI-generated acoustic R&B song with dreamy vibes. The app worked quickly and adjusted when I removed the guitars from some measures or added drums to others. A tap on the screen could change the rhythm or the key.
Soundraw is intended to provide creatives and musicians with backing tracks or background music, COO Tao Romera told me, not to displace the music industry. The idea is that a Youtuber, for instance, could dial in a specific track with a specific mood for a specific length of time and use it as an audio track for a video. A rapper could generate beats to perform over. The licensing agreement of Soundraw even requires users to modify the product in some way before distributing it, Romera said.
A multitrack music generating app on a tablet on the show floor of a convention.
Soundraw's app allows you to create music quickly using generative AI. Jon Reed/CNET
Romera said Soundraw's model was trained on music made in-house by the producers who work for the company, not on copyrighted music. Some models are trained using copyrighted music and create complete songs from a prompt. "They really want to replace the whole music business, the artists they're feeding their AI with, which is so ironic," he said.
Playing around with the app as the self-described world's worst guitarist, I could envision myself creating instantaneous backing tracks to practice over, like a jam partner whose schedule is always open. But the music didn't feel, on its own, all that real. I'm not the only one who gets that sense from AI-generated music.
"If there's not a real person behind it who had feelings and expressed those feelings through that story, I am pretty skeptical whether people are going to be able to connect with that," Romera said. "Because, like, what am I connecting with? This was just created by a machine."
Is AI art a threat to human artists?
One ballroom over from the SXSW Expo and its AI tools is a completely different world. It's called Flatstock, a show of posterboard artists and screen printers. You can easily watch your paycheck roll up and vanish into a cardboard tube, with musicians' show posters in a dozen styles.
And this work? It's all done by human hands. Maybe a touch of Photoshop. But not by generative AI tools like Midjourney. That's by design: The Poster Institute, which put on the show, doesn't allow AI-generated art. Terrence "Tuffy" Ryan, a San Francisco-based artist and president of the institute, said the AI policy stems from the fact that artists haven't been compensated for their art being used to create a tool that competes with them.
"Our biggest concern is that it's taking money out of our pockets and it's a real threat to our entire industry," Ryan said.
Artists I talked to were not fans of generative AI tools and the designs they create but that doesn't mean they haven't investigated. Ryan said he experimented with image-generating tools last year. The results were … boring. He found the work to be homogenized, combining disparate styles to form something lacking distinct characteristics. Even when it gets the usual number of human fingers and toes correct, it gets repetitive in its detail work.
"I think when you see something that's AI, there's just that uncanny valley thing going on, where, like, sometimes you might even know what is in it, but you're subconsciously picking up on whatever that is," he said.
Taylor Adams, a poster artist from Alabama, also said his experiments with it were underwhelming. "It feels like I'm just pulling a slot machine and something random's coming out," he said. "And no matter how hard you paint over it, it doesn't do what I want it to do. By the time I'm done with it, I could have just drawn it in the first place."
Another artist, Barry Blankenship, said he hasn't used the tools and won't. His concerns stem from the use of copyrighted images to train the models. "In my opinion, I don't see any ethical AI on the market today," he said.
In the minds of many artists, these tools try to find shortcuts through essential components of creating art. The time and practice it takes to form a style are what make art compelling, Blankenship said. In the study of art history, for example, time isn't just spent on how an artist chose colors or applied paint to the canvas but on that person's life and existence. You can't fully appreciate Vincent Van Gogh's work, for instance, without understanding his mental and emotional state (and the time he cut off his ear).
"If you're just taking away from the process and just removing the process entirely and just basing it off of existing artwork, how does that help an artist?" Blankenship said. "How does that help the world? I just don't see it."
A booth of posters on a trade show floor.
Poster artists sold their work in the same Austin Convention Center where speakers and tech companies discussed the use of generative AI to create images at SXSW 2025 in Austin, Texas. Jon Reed/CNET
What do these artists want? For one, they want to have a say in how the future of art takes shape. Adams said it's better for AI companies to involve artists up front to reach a system that works for everyone rather than waiting for the inevitable lawsuit that might lead to a big payout.
"It is easier to just pay that fee rather than try and get something working ahead of time," he said.
Something new with AI
There is another option. Some artists are finding creative ways to use AI not to replace the tools to make traditional art but to make something that would've been unfathomable a few years ago.
The musician Will.i.am, in a conversation with Qualcomm CEO Cristiano Amon, called these people "hyper-creators," people who are doing entirely new art. He compared it to the change brought to the music industry by recording -- which was totally different from the printed music world of classical composers. These artists will be "just reimagining the concept of a song. What is music?"
Holly Herndon's work gives a glimpse at what this looks like. Herndon's art tackles the meta aspects of generative AI. She trained an AI model on her voice and then allowed people to sing through it. It's called Holly+.
Herndon calls this type of work "protocol art." It's more akin to performance art than it is to releasing an album or selling a painting.
"It's not about trying to control the process as something that the artist owns alone," she said. "It's about creating a sort of rule set where you want people to take the idea and fork it and run with it."
Herndon worked with machine learning for years before OpenAI's ChatGPT launch in 2022 set off the current generative AI boom. Despite her affection for machine learning as a tool, she has serious concerns about how these models are trained and what this means for working artists, especially those trained on others' intellectual property.
"If we don't come up with a solution that really honors creators and IP holders in this generative AI ecosystem, then we're headed toward a situation where the internet becomes less open and free, becomes more locked down, and we have to have a more defensive approach," she said.
It isn't the AI that makes Herndon's work so interesting. It's the fact that she's using these technologies to amplify herself and her own creativity. The tech experts at SXSW know that AI-generated art that misses the human element won't resonate with people.
"Humans are complicated, and AI systems will not replace us," said Oji Udezue, who has led product teams for several tech companies. "They can't love, they can't trust, they can't do so many things that actually make us human, and we can smell that out."