Set to perform at the innovative music festival and conference Sónar+D, Ville Haimala and Yaboi Hanoi explain how they've utilised AI in their music to create something entirely new.
"We've seen the introduction of AI into culture creation and there's been a moral panic. Is it going to steal all our jobs?" the electronic artist Ville Haimala says within moments of us starting our conversation.
Haimala may accept that AI can "do what we do but just faster and cheaper", but he makes the contention that a lot of the AI-generated "culture" you find on social media sites is representative of AI "aiming to the most vanilla, average human output."
"It tries to mimic the most common medium human behaviour," Haimala says. "I always wanted to see what happens if I push beyond that."
Haimala isn't an anti-AI luddite. One half of the bracingly abrasive Finnish electronic duo Amnesia Scanner who have embraced innovative technology in their music. He is speaking to me ahead of his new solo project 'Hyporeal'. Haimala is currently preparing for its debut in Berlin, where he's based, before a performance at Sónar+D, the industry event that runs concurrently with the Barcelona-based festival in June.
As part of Amnesia Scanner, Haimala has produced dark glitchy tracks with technology riven by obtuse experimentation. Their most recent album 'HOAX' brings their claustrophobic club atmosphere and splits it apart through a second-half by Freeka Tet which reimagines the album noise cancelling headphones.
'Hyporeal' takes the technical experimentation to another level. Haimala has spent two years working with AI models to "try to find the AI native expression". While others use AI to voice back their own demands, Haimala is trying to give a voice to the machine-learning generators behind it all.
To do this, he feeds inputs into a combination of models, then creates feedback loops by asking them to continually modify their own inputs until their output is truly their own. "I lead it into a maze where it can't find its way back to the human anymore," Haimala says. He suggests that 'Hyporeal' isn't totally his project and he's more the producer for the AI artist. "I'm sorta like Rick Rubin sitting on the couch telling the AI 'no, you're still doing something that's not you'," he jokes.
To find meaning in AI-generated art, we have to move away from the slop that is being shoved down throats on social media. With each new update to ChatGPT or another image generator, gleeful technocrats hail the end of artistic careers.
Haimala believes that the rise of technology will bring us to a more spiritual understanding of human-made art. What AI will replace is the steady stream of bland background art that is increasingly consumed. "There's a big demand for background functional music for working out, focusing, and advertising that can all blend seamlessly together. I think AI is a much more fitting tool for this."
Spotify was recently shown to be producing AI-generated playlists as a means of creating "music we benefited from financially" by Liz Pelly in her book "Mood Machine: The Rise of Spotify and the Costs of the Perfect Playlist".
Where AI can exist in tandem with humanity is in finding new ways to tell human stories of our lives which are now indelibly linked to technology. "I still believe in music as a vehicle for telling stories," he says. "AI can just assist and bring people who don't have the traditional musical skills to express themselves."
He suggests a world where people with poetic ideas can utilise AI to forge gaps in their musical knowledge. "Listener driven music scenes where music is created almost like memes." Musical virtuosity will never be replaced, but the form can be expanded to include more methods of expression, instead of dulling out voices as some fear AI will do.
Another musician who believes that AI can be used create unique music instead of middle-ground content is Lamtharn 'Hanoi' Hantrakul, or ญาบอยฮานอย. The Thai musician/AI researcher is best known by his stage name Yaboi Hanoi. In 2022, he won the AI Song Contest with his track 'Enter Demons and Gods'.
'Enter Demons and Gods' was the summation of a career started with a double-major in applied physics and music at Yale before working at Google and TikTok. Using his personal life to combine his interests in physics and music, the final piece was to connect it to his Thai culture.
"I wanted all of these things to kind of come together and make a very tangible manifesto," Hanoi says. "How can machine learning be used to empower music in a way that is true to the melodies and the tuning systems of music from Southeast Asia, and especially music from Thailand."
The result is a fascinating song that bends listeners expectations, deviating away from western approaches to harmony but also not sounding like a traditional Thai piece of music. No matter your background, 'Enter Demons and Gods' will sound both familiar and unfamiliar at once.
Most generative AI content is created to average out all content available. The result often means creating a piece of art that sits at precisely the most generic middle ground of all possible examples. Typically, due to the proliferation of western media, this skews outputs to resemble white European art.
What Hanoi did was to create entirely new AI programs that would focus specifically on the unique cultures that he cared about. He trained AI to play in music that didn't rely on equal temperament, the semitone-based note system devised from pianos. "I felt like I'd landed in this really beautiful space where these melodies could only be made by approaching it from an AI-driven machine learning style of synthesis.
Since winning the AI Song Contest, Hanoi has continued to innovate his AI programs to continue to synthesise Thai and Asian approaches to harmony with new projects incorporating dance choreography and mythical storytelling.
Hanoi believes AI can be used to create a new chapter in culture. "If a machine learning model is trained on pieces of Thai history and Thai architecture and Thai design and Thai melodies, it becomes a living and breathing artefact of that culture. Just in the same way that you would treat a national dress or national pot that was made in the early fifth century BC."
"My aesthetics are very much driven from a point of respecting the culture and respecting the history," Hanoi says.
Both Hanoi and Haimala are connected in their use of AI to forge new cultural languages instead of flattening them out as we've seen in countless social media trends such as how everyone approximated Hayao Miyazaki's beautiful art style recently.
Hanoi, like Haimala, is also set to perform at the upcoming Sónar+D. While the main festival will see sets from DJs like Peggy Gou, Eric Prydz and Skrillex, Sónar+D is more like a conference for people in the music industry fascinated with the future of technology.
Sónar+D has been at the forefront of bridging AI with music. Their first AI speaker was in 2016 and in 2021, they developed the AI Music Programme with the S+T+ARTS programme from the European Commission.
Where Sónar+D is different from a normal tech conference is, thanks to its link to the main music festival, the priority is always how it can work for artists. "We are Sónar, so we're always going to be on the side of artists and creators," Andrea Faroppa Cabrera, Head of Sónar+D says.
This year's programme focuses on three branches, "AI + Creativity", "Futuring the creative industries", and "Worlds to come". Through these, Faroppa Cabrera thinks the festival can shed light on how artists are able to evolve alongside technology in ways exemplified by Haimala and Hanoi.
"Technology is made by humans," she says. "We tend to forget the physicality of these things. At the centre of it all, artists can show us the different ways of using and abusing technology."
