in Articles

AI: Disrupting bonds between music, artists and audiences

by Wif Stenger

How is AI impacting the music industry? Three experts from Finland tell us it will shake our fundamental assumptions about artists’ identities, their unique creative visions and relationships with listeners.

“A grotesque mockery of what it is to be human.”

That was singer-songwriter Nick Cave’s response earlier this year when asked about song lyrics generated by artificial intelligence. Using generative AI in the composition and recording process has drawn similar responses. Like it or not, though, this is the new reality, as evidenced by this year’s new releases from the Beatles and the Rolling Stones, and the proliferation of apps promising to churn out “royalty-free classical music”.

This year, Canadian electropop star Grimes launched an interactive AI toy, her latest single “I Wanna Be Software” (“Upload my soul…you can make me however you like”), and a generative AI music platform that allows anyone to use a “GrimesAI voiceprint” in approved collaborations. Although she charges a fee for that, Grimes tweeted that she supports “open-sourcing all art and killing copyright” – which is easier to say when you have an estimated net worth of $12 million than if you are a struggling new artist.

Grimes was inspired by an AI-generated hit featuring deepfake, simulated voices of Drake and The Weeknd, which went viral before being yanked from streaming platforms after a copyright claim.

While AI has gone mainstream in the music industry this year, it’s hardly new. Pertti Grönholm, an electronic musician and university lecturer in history at the University of Turku, points out that consumers have been offered completely virtual artists since at least 2007, when Japanese ‘virtual idol’ Hatsune Miku began a highly successful recording and ‘performing’ career. Her ‘voice,’ which appears on over 100,000 songs, is cobbled together from micro-samples from a real vocalist using Vocaloid software, which was originally developed in 2000. In late 2022, it was renamed Vocaloid:AI.

“It’s possible that consumers will be offered more and more completely virtual artists as well as music and visualisations generated by AI. On the other hand, virtual artists can be combined from features, mannerisms and styles of living or dead performers,” says Grönholm.

“We may see deceased musicians ‘resurrected’ with the help of AI, continuing their lives as virtual artists. Such use of AI might weaken our interest in pop music and artists in general. Despite this, live music will probably retain its attractiveness,” predicts Grönholm, who has researched the history of electronic music and whose band E-Musikgruppe Lux Ohr releases albums on Finnish and German labels.

Pertti Gronholm Of E Musikgruppe Lux Ohr  Live @ Ambient Festival Pori 2023 09 08  Image By Tero Liedes C 2023
"We may see deceased musicians ‘resurrected’ with the help of AI, continuing their lives as virtual artists", notes Pertti Grönholm. Photo: Tero Liedes

Creating music – or generating sound?

Artificial intelligence is one thing, but how about artificial creativity? The endless possibilities of AI will force us to question and redefine creativity, along with the complex relationships between artists, their work and their audiences.

“Generative AI combines old ingredients, rearranging pieces of previously generated content into a new order. So, it’s natural to ask whether anything new can emerge through that,” says Pii Telakivi, a post-doc researcher in practical philosophy at the University of Helsinki, who has published several articles on AI in recent years.

“On the other hand, humans, including artists, also combine the old. We have an unconscious storage of ‘data,’ representing our entire learning history within our environment. So, combining the old is perhaps not as such an argument for discarding something as not creative. The outcomes might appear as creative and innovative, even if the ingredients are second-hand and processes not ‘creative’ in the sense that we have traditionally understood that,” she says.

Kuva Pii Ja Robotti Cata Portin
"Combining the old is perhaps not as such an argument for discarding something as not creative", says Pii Telakivi. Photo: Cata Portin

Composing, especially for a mass audience, has always been a balancing act between tapping into the familiar while creating something new and innovative.

Generative models “are only able to create something truly novel in very limited ways, for example by making combinations of previously used styles,” writer another University of Helsinki researcher, musicologist and machine-learning expert Otso Björklund in his article in this special issue of FMQ. The latest results of using generative models to create music “have not been as promising as with text and images,” he notes.

Even with text such as song lyrics, they are still relatively crude and superficial, as Cave pointed out.

For instance, just for fun, I asked ChatGPT to write a rap lyric about AI in the style of MF Doom and Madlib. The results were rather cringeworthy: “Machine intelligence, like a lyrical beast / In the silicon jungle, where every line’s a feast…AI algorithms, spitting rhymes in binary/In the language of the future, where the rhythm’s never secondary”.

A request to craft lyrics about AI in the style of a Great American Songbook standard was slightly more successful, even showing some wry Cole Porter-like humour: “You’re the algorithm of my heart’s refrain…You’re the Turing test of a love sublime” – referring to Alan Turings 1950 test of a machine’s ability to show intelligent behaviour indistinguishable from that of a human.

Therein lies a key question: will listeners be able to tell whether music is “the real thing” or an AI artifact?

“A listener can’t possibly know whether a song was produced with the assistance of AI or not,” notes Telakivi. Of course, in live concerts, the expertise of the musician remains unchanged, even if the creation process involved generative AI.”

“When one can copy the style and voice of a specific artist by using existing recordings of the artist to create new material, which will of course make it more difficult to protect one’s intellectual property and rights,” says Prof. Tuomas Virtanen, leader of the Audio Research Group at the University of Tampere.

“It can be difficult to detect or prove reliably that somebody has ‘stolen’ your voice or playing to create new material. And currently there is no established legislation or ways to manage rights in this situation.”

The copyright issue

In late 2023, the EU agreed on the principles concerning transparency obligations for AI developers as part of its planned AI Act, the world’s first legislative proposal of its kind. It was cautiously welcomed by two major music industry groups, the IFPI and GEMA.

Dr. Tobias Holzmüller, head of the performance rights organisation GEMA, said that “those intending to offer generative AI in Europe must be able to explain what contents they used to train it.” The major risk for composers and performers is that AI will rip them off, using their work as raw material with recompense.

“So far, the EU proposal does not seem to recognise the risks or define status related to intellectual property, creativity, authenticity, person’s right to their own voice, appearance and so on,” says Grönholm. “From this perspective the proposal clearly needs some reworking.”

“For example: are a performer’s voice, mannerisms, and accent their personal property? Creative personalities, their works and working methods could increasingly become the targets of simulation, theft and exploitation,” he points out.

A frequently mentioned downside of AI is that it tends to reinforce stereotypes and clichés, and music is naturally not immune to this.

“I don’t think there are currently easy ways for users of AI technologies to avoid this, since the data that is used by AI developers affects such stereotypes, and users can’t directly control this,” says Virtanen. “This will require increasing awareness of such biases and pushing developers of AI technologies to avoid them. This is a broader problem in AI technologies, which requires ways to measure better such biases, for example.”

“AI may favour styles of music that are esteemed in Western countries, for instance, so the content it produces is likely to be dominated by those styles. Users need to be aware of this,” says Telakivi.

Tuomas Virtanen
Tuomas Virtanen works as the leader of the Audio Research Group at the University of Tampere.

AI – your personal assistant

During this interview, it has become evident that the risks are clear. But what about the benefits?

“When skilfully used, generative AI can become a useful tool that enhances and improves efficiency and creativity,” says Telakivi. “It can lower the barrier to start and ease the fear of a blank page. With its help, one can generate numerous variations of a specific theme within seconds. The composer can experiment and fiddle around with various styles, instruments, and vocal expressions in a way that wasn’t possible before. AI tools can also assist with laborious and time-consuming technical tasks.”

From the artist’s perspective, suggests Grönholm, AI could be used as a personal assistant, trainer, mentor and sparring partner.

“It can help us analyse our own working methods, helping us become more aware of unconscious preferences, challenges and strengths,” he says. “And it could surely be helpful in learning music theory, playing skills and so on.”

“AI can also generate new ways of listening to music, for example novel technologies that create the sense of immersion of being at a concert even if you’re sitting on your sofa, and even interacting with music so that it adapts based on user behaviour,” says Virtanen.

Grönholm agrees, predicting that “algorithms may observe our states of mind, our emotions and the stress state of our bodies, and recommend music suitable for different situations”. However, he warns, “in this way our minds could also be exposed to political, ideological and commercial manipulation through music.”

Telakivi sees the AI revolution as part of a broader sea change in creativity.

“In the TikTok era, younger people in particular have become used to creating content themselves, and music industry will have to adjust. It has been ruled by a few big companies, but generative AI tools will remove their gatekeeper role, as anyone can create professional-sounding music in seconds and sell that to streaming services or upload it on social media. Social media has already changed the fan culture, and AI will swirl up even more how music and fans come together.”


An unpredictable future

To sum up, in Grönholm’s view, “there is a true risk of losing diversity and richness of human creativity and also many cultural idiosyncrasies and marginalities. However, I’m still somewhat optimistic, as people may grow more critical and change their consumption behaviour. In the best-case scenario, we’ll get even better musical and visual content to experience, as long as the music creators harness AI and algorithmic applications as tools of their own authentic creative power.”

Clearly, AI will reshape how we consume and conceptualise music.

“Technology is not separate from the rest of society, and it is impossible to predict the future because there are so many variables. For example, the impact of social media was quite unpredictable 15 years ago,” notes Telakivi.

“However, the significance of authorship in commercial music will probably become less clear since everyone can tailor music to their own preferences,” she says. “On the other hand, fan culture and the inherent meaningfulness embedded in art will probably continue to lead people to appreciate the human artists behind their favourite music.”

Featured photo: Data flock (digits) by Philipp Schmitt. Souce: Wikimedia Commons