Digital Crescendo
AI is making headlines in many creative industries, including music — how is Winnipeg’s music scene responding?
Listen to this story:
Through the dimly lit purple glow in his home recording setup — a simple audio interface, a mic, a guitar, and pedals — Winson Lam, the guitarist of Winnipeg hardcore punk band Nuclear Man, sits in front of his monitors holding his burgundy Squier Affinity Jazzmaster on his lap.
“If you’re really good at art, you’re not going to be replaced with AI,” says Lam.
Before today, Lam had never experimented with artificial intelligence (AI) in his music.
AI — software that mimics the human mind to generate images, text, and other creative work by scanning seemingly infinite amounts of data — has become a popular topic in all creative industries, including music. Big-name artists around the world, like Paul McCartney, have shared that they’re open to using AI. Others like Nick Cave have strong feelings against it.
AI isn’t a new concept in the music scene. In 1951, a major contributor to computer science named Alan Turing built a machine that generated three simple melodies. But now, over 70 years later, AI has become much more powerful. AI-powered tools come in many forms and can create things that verge on the uncanny. There are programs that can create music videos that were never produced, like this one that a fan created of a Lana Del Rey song, artist collaborations that don’t exist, deepfakes, like Freddie Mercury covering The Beatles, and more.
So how are Winnipeg musicians responding to AI?
Winnipeg musician Liam Duncan, known best as Boy Golden, says he’s not worried about what AI will do to the music industry.
“Interesting, but not interesting to me,” he says. “I’m interested in art, and I’m interested in the artists that make it. I think for most people who are interested in art the same way that I am, it’s not going to be very interesting or moving for them to hear music that was not made by a person.”
Duncan says he’s done some investigating into the software and can’t see himself using it as a creative tool in music-making, mainly because he already doesn’t use digital technology until the final process. Instead, he relies on tapes for recording. However, he has used AI for video creation.
“I have mixed feelings about it because in some ways I’m taking work away from a human animator,” says Duncan. “In another way, though, I could never afford a human animator, because I’m just a little baby artist who doesn’t have a bunch of money to spend.”
He makes short music videos to post on YouTube and Instagram of his new songs. He says that all he needs is a green screen, and AI does the rest for him.
Roman Clarke, a Winnipeg artist and multi-instrumentalist, weaves the sounds of soul through groovy pop-rock songs and has mastered singing and drumming simultaneously.
“I don’t really have a ton of experience with AI and music, but when I think about it, I actually use a ton of software,” says Clarke.
One software he uses is called “Soothe2,” a mixing tool that identifies problem frequencies and takes them out of the song for you.
“AI is still so new, and I don’t know where the line is, what’s AI and what’s not?” says Clarke.
The key difference between traditional software versus AI-powered software is that software with no AI-powered features is programmed to perform tasks using fixed instructions, according to an article by Intelligent Block. AI uses machine learning to imitate human thoughts and consider the context when giving answers. It is always adapting and learning for continuous improvement.
Clarke says that if someone asked him if he wanted to use an AI-generated tool, he would likely decline because of the stigma around it. That doesn’t mean he is necessarily against it.
“The software is created by a person to do a specific thing, and to me, that’s like the same thing as an instrument,” says Clarke.
He thinks that because a person interpreted and analyzed the data to make software, it just feels like an extension of humanity. People seem to fear AI, but it might turn out to be an essential instrument.
“We don’t have to be so limited to just what we can do as a single person or as a group of people, so I think there’s a purpose,” says Clarke.
There are many AI tools that serve different purposes. AIVA can produce a variety of musical work and offers full usage rights, Splash can produce customized songs quickly, and LANDR is for mastering, which is the final step of audio production.
When I asked Nuclear Man’s Winson Lam to experiment with AI to create a song, he looked up an AI music generator on Google. After testing three programs, he ultimately selects a program called “Beatoven.ai” to create his first song with the help of AI.
Winnipeg’s folk singer-songwriter Madeleine Roger lights up with excitement as she talks about the depth of her love for the songwriting process. She says she can’t imagine giving up something that keeps her awake with excitement. Then I ask her about AI, and her enthusiasm vanishes.
“The first word I think of is scary, but I think because I am a little bit terrified of what it means for the future perceptions of music from an audience perspective,” says Roger, pausing. “But I think that’s also coming from my own problems of being fearful of change and robots.”
Roger has experimented with AI in writing. She says she was curious about how it worked as a generative tool and explains that she was both amazed and horrified at the speed and execution of an idea. Roger used ChatGPT and said after she entered the prompts it came up with something in mere seconds that would usually have taken her an entire day to create. She says she asked it to write about a koala that goes on an adventure and eats granola because she was curious to see how specific it could get from a silly prompt.
“It came up with this narratively brilliant story that was in perfect rhyming pattern and had five verses,” says Roger.
Though amazed, she still doesn’t want to be tempted to use it for her general, everyday creative process. She said that one of the most compelling things that happens in an artistic journey is finding your own artistic voice, figuring out your style, and growing from what you’ve created in the past. Creating a sound that is unmistakably your own from diving into your own experience and imagination is something that AI cannot do, and Roger doesn’t want to stop herself from finding that for herself by using an AI-powered tool.
“I’m sure there’s going to be really beautiful, wonderful, fantastic things that come out of it, in the same way that there’s going to be potentially really destructive side effects or some loss, too,” says Roger.
To the point that AI pulls from content that already exists, Roger says that humans do too because we are influenced by everything we ingest.
Roger says she is working on an album with the working title “Dystopian Lullabies.” It features a variety of songs for and by people at the end of the world, and she does want one song on the record to be written by AI as part of the story. It will represent a future where computers are doing the storytelling. While doing some early experimenting with this, she’s been able to use ChatGPT to write dystopian poetry from different characters’ perspectives, like Shakespeare’s.
Beatoven.ai, the program Lam is using to write his song, does not write words based on prompts, rather it generates melodies based on them. You begin with selecting a mood, with options like sad, scary, fearful, and joyful, and then pick a genre for your song. From there you can prompt it with more specific instructions.
“If you say, ‘make me a Led Zeppelin song,’ it might give you something super random — it’s not super accurate,” says Lam.
He says you can really get into the prompts by giving it a duration, vibe, era, and occasion. He said he chose this generator because of the text prompts, while some of the others just gave you something with the click of a button, and that’s it. He ended up asking for a song that sounds like “Youngest Daughter” by Superheaven and went from there.
Copyright is a major question for both musicians and people using AI. AI works by analyzing large combinations of existing data and learns from the patterns within it. It runs through millions of tasks in seconds and improves each time, but it doesn’t tell users where it pulled the data from. So how can musicians protect their copyrighted music from being used by AI? How can musicians ensure they are not violating copyright law when using AI as part of their creative process?
Michael Duboff, a Canadian entertainment lawyer who specializes in music and other kinds of media, says he often sees musicians using AI to push the boundaries with what the technology can do. Going too far and extending into the area of deepfakes, which means using AI to replicate someone’s voice or image to create something that never happened, are violations of personality and intellectual property rights use. An example of this is a “collaboration” between The Weeknd and Drake. A song called “Heart on My Sleeve,” posted on TikTok by user Ghostwriter977, blew up the internet for the day. It turned out it was only a fake collaboration created with AI using the artists’ voices.
Another legal consideration is ownership. When someone uses AI as part of music, who owns the final product?
Duboff says someone who creates music will become the copyright owners of that music. If it’s an original piece of music they’ve recorded in some form and it doesn’t violate anyone else’s copyright, they become the copyright owner of that recording. Then, they can do anything they want with it, whether that’s putting it online, selling it, or licensing it to other people. Other people cannot use that recording without the copyright owner’s consent.
He says the issue with AI is that it scrapes as much information and data from the internet as it can, including copyrighted materials. When AI generates something new, it doesn’t tell us if it’s based on a copyrighted work or not.
“These programs as a tool are phenomenal,” says Duboff. “They can help cut down costs and increase efficiencies for people and for businesses.”
He says, however, to be mindful and practice care when using it as a generative tool to avoid trouble with copyright laws.
“I do have some worries, both as a musician and a lawyer. Whether it will dilute the creative process of humans and cut down on the workforce,” says Duboff.
Overall, Duboff says that AI, when used lawfully, can be a cool tool. There are endless ways to manipulate it and create something that, without AI, might need software, pedals, and effects.
There are no laws around AI in Canada right now. However, Canada is among the first countries to propose a law to regulate AI and it has developed a Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, according to the Government of Canada. The proposed framework would ensure safely developed AI systems for use in the country and accountability for the businesses that develop and use the programs.
Sydney Krause, who has a background in business and branding, spoke at one of Manitoba Music’s Sound Waves Music Meetings, called “AI: Your Creative Partner.” She takes pride in her ability to take creative approaches to finding solutions and has started using AI as a tool. In the meeting, she says that she tests about five new AI programs every day.
There are several misconceptions about AI, and Krause says they need to be addressed to foster a better understanding. A major one is that AI will make human labour obsolete.
“While AI has the potential to automate certain tasks, it is unlikely to completely replace human labour. Instead, it can augment and enhance human capabilities,” says Krause.
She says there are a few different ways AI can be used as a tool: creative collaboration, meaning AI can facilitate collaboration between artists by providing virtual assistants that help brainstorm ideas; artistic style transfer, where AI algorithms can transform images or videos into different artistic styles; and AI generation tools, like the one Lam is using.
Lam is making progress on his song. He now has a base track.
“I picked this one because there’s a little bit of flexibility, you can change the emotion, genre, or tempo after if you want. It also gives you four different options to choose from,” says Lam, pointing at the list of different versions on his monitor. “You can also take out certain parts of what it generates, like the chords, bassline, and percussion.”
He plays the song the program generated. Beatoven.ai allows you to listen to the chords, melody, bassline, and percussion separately and remove any elements you don’t want. He clicks and listens to each one. He says he’s gotten the same drum beat a few times with different prompts when testing out the program before and feels like the program is limited in the sounds it can produce.
Bluebloods is a collaborative R&B band from Winnipeg led by producer, guitarist, and songwriter Conrad Sweatman.
“I’m among those who don’t feel serious artists pursuing originality in music should feel threatened by AI,” says Sweatman.
He sees AI as mostly a “glorified search engine,” but it is capable of producing impressive outcomes with the right prompts.
“In music, all of this is rather gimmicky and fun,” says Sweatman. “It’s innovative technologically, but not expressively; it’s producing convincing knockoffs, not daring originality.”
He compares it to hip-hop and EDM producers using samples in original ways. When the final product is good, we know the producer deserves more credit than the actual sample. He says he can imagine AI dominating in any musical field where the rules around it are established.
“It will be ace at producing generic music for commercials, dentist offices, the sounds Air Canada pipes at us when we’re on hold,” says Sweatman. “It will ascend to much higher heights, as it already has, by creating very elegant Bach-style fugues, chorales, and contrapuntal music.”
He also says that until AI possesses some form of consciousness, he’s not holding his breath.
A survey by Ditto Music with over 1200 responses found that 59.5 per cent of independent musicians in 2023 used AI for their music. A survey I conducted to find out how many of Winnipeg’s local musicians have used AI revealed that 33 per cent of the 28 respondents have so far. Of those who have used it, lyric generation was the most popular tool, followed by production, music composition, remixing, and sampling other songs.
“I think AI is a wonderful tool that all musicians need to harness the power of,” says one survey respondent. “When humans write music, they infuse emotion into the music, and music would be lifeless without it. But I think AI is helpful for sparking that creativity in humans.”
Other survey respondents weren’t so optimistic.
“The concept of conversing with AI in an artistic sense is interesting and could be helpful in the right hands. It will inevitably get used in the wrong hands as a means of commodifying art,” says another survey respondent.
A respondent posed that art is for humans and only humans, and another worries about music coming from non-human sources.
A music fan who responded to the survey said everything’s good in moderation, AI included.
“People were skeptical when sampling started to become popular in the music industry. I think the future of AI in the music industry will have a similar evolution,” says the fan.
Sampling, repurposing a segment of a song or sound from one recording into another recording, gained popularity in the 1980s. The hip-hop genre is to thank for its creation, and now we see it in many genres. Sampling is widely accepted today, and a popular example of this is Jack Harlow’s song “First Class” where he samples Fergie’s “Glamorous” featuring Ludacris.
Another tool that sparked controversy was Auto-Tune, a software used to correct the pitch in vocals. Cher’s 1998 song “Believe” featured Auto-Tune vocals and was the start of the technology’s popularity in the music industry. A common argument was that it’s misleading to present perfect vocals, and using Auto-Tune is “cheating.” Decades later, it’s just an instrument, especially in genres like rap. Although, to some it still raises concern over authenticity and talent.
Now that Lam has the base for his song, he opted to remove the guitar chords and the melody, leaving him with only the percussion and bass to lay his own guitar riffs over. He recorded his guitar and put it over the AI-generated bassline and percussion to complete an instrumental song he partnered with Beatoven.ai to create.
“The thing I liked about Beatoven.ai was that it generated the percussion for me, because I don’t play drums. So usually when I do drums, I program it and that takes me forever,” says Lam. “So being given a drum track felt like I was playing with a drummer.”
With a laugh, he said he doesn’t think the drums sounded very good. But overall, he said using AI to generate them for him made the process easier.
“The most I could see myself using it for is demoing stuff because right now, I’m writing stuff for a few projects. I’m writing a lot of the guitar and need to program the drums,” says Lam.
He says that Beatoven.ai is the most flexible AI program he could find, and in his opinion, it is substandard because it still wasn’t flexible enough. He does note that there’s an option for a paid version of Beatoven.ai, and he was using the free version for this song.
He thinks people are making AI a bigger deal than it really is. However, his greatest worry would be the rate at which “unoriginal” or “untalented” people would make music.
“I feel like music as an art is so over-saturated with people who aren’t passionate about what they make, or are manufactured to be ‘successful,’” says Lam. “There are already so many people making music and adding people who think they can just generate their songs would make it even more difficult to find things you like.”
Listen to Lam’s AI-generated song:
I found a mix of opinions and feelings surrounding the ever-evolving role of AI in music in Winnipeg’s music scene. While only a few Winnipeg artists are actively using it now for things like production and visual storytelling, other artists may incorporate it as a tool in the future.
While our local music scene’s adoption of AI lags behind many other parts of the world, Winnipeg’s experimental spirit can be seen through Lam’s experimentation. Amid these discussions, Lam, who was reluctant to use AI for songwriting until now, might have found a way to use it in his early creative processes without letting it take away from his final products. His Squier Affinity Jazzmaster and pedals sit safely in his recording setup, not being replaced by AI anytime soon.
AI, in whatever forms it will take, is likely here to stay, and the music industry will adapt to it just like it did when Auto-Tune and sampling made their debut. Human experience and emotion are what make music so compelling. AI is not a replacement for creativity, but it can be a catalyst. The local musicians I talked to see how the software can potentially benefit new musicians who don’t have the capacity to outsource producers or music gear, or solo artists wanting to record demos, but they also reiterated their commitment to traditional ways of making music. Like we have seen through history, musicians will continue to use technology to support and push their creativity.
