"David Bowie doing a somersault."
In the time it took to type the prompt, an image of "Heroes"-era Bowie loads on the mounted TV, his legs thrown over his head in weightless suspension.
The image isn't credited to a photographer or graphic designer. "Floating Bowie" is an image conjured by DALL·E 2, an AI system that generates images from text descriptions. Think of it as a limitless visual Google for your imagination. If you wanted to see Taylor Swift playing Jenga against Alexander the Great, DALL·E 2 could make that happen.
For Toronto indie pop band shy kids, the system is among a handful of AI platforms they used to create the visuals for their upcoming album, house cats (out July 14 on Everything Forever).
"It just feels like another tool in the toolkit," the band's Patrick Cederberg tells Exclaim! from the band's office space in Toronto's Ossington neighbourhood. "It's blowing a lot of people's minds, but for us, I think we're finding how to best work it into the process that's already existed for as long as we've been working."
Over the past few months, AI has dominated the discourse within creative industries. Image-generation systems, voice-cloning apps and lyric-writing programs have artists digging their heels into either side of the "friend" or "foe" line in the sand. In the former camp, you'll find Grimes, who is open-sourcing her voice to facilitate cloning and offering a 50/50 split of all resulting song royalties.
Nick Cave, meanwhile, is among the vehemently anti-AI. When a fan sent him song lyrics generated by ChatGPT in his likeness, the Aussie rocker called it a "grotesque mockery of what it is to be human."
Cave's sentiment echoes the concern of many creatives. Will AI be a crutch or a collaborator? A creative architect or a thief? Many AI systems pull from existing content to create composite art, raising urgent questions around copyright infringement — and the future of artistic collaboration and which roles might become obsolete. Beyond that, AI poses societal risks of algorithmic bias and misinformation, as well as threats to privacy and security. The internet might get a kick out of hearing a voice-cloned Eminem rapping about cats, but the technology is drawing artists, labels and streamers into a whirlpool of muddied legalities.
Amid the growing panic, shy kids offer an optimistic balm. Consisting of Patrick Cederberg, Walter Woodman and Matthew Hornick, the trio are also award-winning filmmakers who have long been melding art with technology.
"I think so much of this is about idea generation," Woodman explains. "It's not like the first thing I do is go to my computer and say, 'Make me a movie,' and then I go get a coffee while it's making it. It's like, you are doing something and you hit a wall. How do you get over that wall? Sometimes you go for a walk. Sometimes you smoke a joint. Sometimes you use AI."
In these ways, staring at the blank page has never been easier. Traditionalists may lament this shift in the creative process, but writer's block becomes less of an issue when you can iterate content almost instantaneously.
Along with generating art from scratch, artists can also take existing content and feed it through AI systems after the fact. This was the case with shy kids' recent video for "fresh off a feeling," which shows a group of otherworldly creatures dancing in what looks like a continuously flowing impressionist painting.
"We wanted to be aliens," Woodman explains, "but we didn't have the budget for Avatar."
Their James Cameron hack was to treat already-recorded footage with AI using Stable Diffusion, another text-to-image model. They filmed themselves dancing in their office space, edited it into a video, and then ran it through the AI system. Similar to DALL·E 2 and Midjourney, Stable Diffusion allows users to enter artistic prompts. In their case, they entered words like "cinematic colour correction," "oil painting" and "Louis Wain" (an artist known for his anthropomorphized cat paintings.) They did this meticulously, shot by shot, gradually imprinting stylistic cues onto their existing footage.
The results aren't instantaneous. It can take days to run video footage through these programs, and if you don't like something, you have to go back into the system and make the changes. Woodman likens it to baking a soufflé: you put all the ingredients together, shut the oven, and wait.
But the soufflé comparison shouldn't suggest an easy-bake guarantee for success. shy kids' Matthew Hornick emphasizes the difference between the enthusiasts (like a music journalist prompting DALL·E 2 for David Bowie doing a somersault) and the developers like Cederberg, who has been learning these systems for months and coding in order to achieve more customized, refined results.
In this way, the relationship between the artist and AI is more collaboration than autopilot. Case in point: the AI-generated Drake and the Weeknd song "Heart on my Sleeve." When the viral track was released, people were so stunned by the cloning accuracy that they disregarded the fact that an anonymous ghostwriter had to produce, write and sing the track before putting it through the AI system.
Cloning aside, creative intervention will be vital in preventing the artist from being lost in the artificial.
"As an artist, you are everything from the story you tell, to your taste and how you execute it. You aren't the final result — you are the process," Cederberg explains. "We also have those thresholds for: is there enough of us in this? I think we would shy away from just doing something for the sake of ease."
These AI systems are also undeniably more accessible than the alternative. Sans AI, the video for "fresh off a feeling" would've required rotoscoping, an animation technique where filmmakers trace over motion picture footage. The process can take months of labour and thousands of dollars. For cash-strapped creatives, cost efficiency is an AI upside.
"The democratization of these things shouldn't scare you. In fact, it's going to be very clear who is themselves and who is not themselves. So are we ourselves? Well, that's a better question," Woodman posits. "I hope we have something unique to say. But I want people to see these things. I want you to know how easy it is to generate that image because it's a good thing. I think it's a good thing."
Ottawa-based artist and producer Thor Simonsen has been grappling with the risks and opportunities presented by AI. Simonsen is the founder of Hitmakerz, a record label based in Iqaluit that specializes in Indigenous artists. His team has been using ChatGPT to draft press releases, write funding applications, decode legal documents and even plan national tours.
"This morning, when I was mapping out a tour for Latin American artists in Canada, I just named the cities we want to go to and I said, 'Give me 25 venues in each city and include their contact info and phone number and email,'" Simonsen tells Exclaim! "I could have done that using Google because it's all publicly available, but it saves me hours of going through all that stuff."
Many are concerned that AI will further separate the art from the artist, but Simonsen sees it as a tool to bring them closer together. A means to an end. By reducing the administrative aspects of the job, the hope is that an artist has more time to dedicate to their craft.
As for the roles behind that administrative work, he predicts that it isn't simply a matter of AI taking people's jobs within the music industry. Rather, it'll be about people who know how to use AI taking the jobs of those who don't.
"It's someone who knows how to use the printing press versus someone who knows 18 scribes. One of them is going to be quite a bit more effective," he argues.
Inevitably, the more that artists integrate AI into their creative process, the more challenging it will be to decipher whether you're listening to, or competing with, an AI product. Will artists disclose this information, or will AI get stamped with an Auto-Tune-like stigma that discourages artists from being transparent?
Simonsen recalls someone proposing a fix at a conference in Colombia earlier this month: artists could add "100% AI free" disclaimers to their bios. "Suddenly, that's like the 'organic' of music," he suggests.
Recently, Simonsen teamed up with Nunavut artist Agaaqtoq to perform a song entirely written and composed by ChatGPT. They instructed the system to deliver a catchy pop song in Inuktitut with an ABAB rhyming pattern. They wanted a track about a man traveling through space to find other musicians, and they asked ChatGPT to provide English translations and chord progressions. The result was "Gitaralauq Sulusiaq," loosely translated as "Guitar Players, Let's Dance."
While Simonsen and Agaaqtoq were impressed by the system's creative capacity, its imperfections were exposed. Much of the AI-generated Inuktitut was pure gibberish. Agaaqtoq, born Abraham Eetak, acknowledges that this is probably due to lack of Inuit involvement in the development of AI — yet another sign of the infancy of these technologies, and the requirement for human interference at some level.
Eetak recognizes the exciting potential of AI as a tool for cultural exchange and idea generation, but after his brief flirtation with ChatGPT, it won't be a fixture in his process.
"I feel like it's just going to take away from my skill. Maybe I'll get too lazy and write a whole album using AI and just make it way too easy. And that takes all the fun out of it," he tells Exclaim! "I like the challenge of writing a song about a topic, whatever that may be. Getting into that process of just going back and forth in my brain with how I want the song to be. With AI, I just can't do that."
Like Eetak, some artists will dip their toes into the AI waters and decide they've had enough. Some will wade right into the deep end, while others will opt to yell "shark!" from the presumed safety of the beach. Simonsen, meanwhile, opts for the nuance between threat and opportunity.
"When you see a painting, it might move you in some way, and I definitely think AI can do that," he reflects. "But I think what humans care about is the context and the story behind it, and what it means to them. If AI creates something, it might be beautiful and technically fantastic, but if it doesn't have any connection to that person, and if it doesn't have a meaning, then I don't think it's valuable. I don't think that's the purpose of art."
In the time it took to type the prompt, an image of "Heroes"-era Bowie loads on the mounted TV, his legs thrown over his head in weightless suspension.
The image isn't credited to a photographer or graphic designer. "Floating Bowie" is an image conjured by DALL·E 2, an AI system that generates images from text descriptions. Think of it as a limitless visual Google for your imagination. If you wanted to see Taylor Swift playing Jenga against Alexander the Great, DALL·E 2 could make that happen.
For Toronto indie pop band shy kids, the system is among a handful of AI platforms they used to create the visuals for their upcoming album, house cats (out July 14 on Everything Forever).
"It just feels like another tool in the toolkit," the band's Patrick Cederberg tells Exclaim! from the band's office space in Toronto's Ossington neighbourhood. "It's blowing a lot of people's minds, but for us, I think we're finding how to best work it into the process that's already existed for as long as we've been working."
Over the past few months, AI has dominated the discourse within creative industries. Image-generation systems, voice-cloning apps and lyric-writing programs have artists digging their heels into either side of the "friend" or "foe" line in the sand. In the former camp, you'll find Grimes, who is open-sourcing her voice to facilitate cloning and offering a 50/50 split of all resulting song royalties.
Nick Cave, meanwhile, is among the vehemently anti-AI. When a fan sent him song lyrics generated by ChatGPT in his likeness, the Aussie rocker called it a "grotesque mockery of what it is to be human."
Cave's sentiment echoes the concern of many creatives. Will AI be a crutch or a collaborator? A creative architect or a thief? Many AI systems pull from existing content to create composite art, raising urgent questions around copyright infringement — and the future of artistic collaboration and which roles might become obsolete. Beyond that, AI poses societal risks of algorithmic bias and misinformation, as well as threats to privacy and security. The internet might get a kick out of hearing a voice-cloned Eminem rapping about cats, but the technology is drawing artists, labels and streamers into a whirlpool of muddied legalities.
Amid the growing panic, shy kids offer an optimistic balm. Consisting of Patrick Cederberg, Walter Woodman and Matthew Hornick, the trio are also award-winning filmmakers who have long been melding art with technology.
"I think so much of this is about idea generation," Woodman explains. "It's not like the first thing I do is go to my computer and say, 'Make me a movie,' and then I go get a coffee while it's making it. It's like, you are doing something and you hit a wall. How do you get over that wall? Sometimes you go for a walk. Sometimes you smoke a joint. Sometimes you use AI."
In these ways, staring at the blank page has never been easier. Traditionalists may lament this shift in the creative process, but writer's block becomes less of an issue when you can iterate content almost instantaneously.
Along with generating art from scratch, artists can also take existing content and feed it through AI systems after the fact. This was the case with shy kids' recent video for "fresh off a feeling," which shows a group of otherworldly creatures dancing in what looks like a continuously flowing impressionist painting.
"We wanted to be aliens," Woodman explains, "but we didn't have the budget for Avatar."
Their James Cameron hack was to treat already-recorded footage with AI using Stable Diffusion, another text-to-image model. They filmed themselves dancing in their office space, edited it into a video, and then ran it through the AI system. Similar to DALL·E 2 and Midjourney, Stable Diffusion allows users to enter artistic prompts. In their case, they entered words like "cinematic colour correction," "oil painting" and "Louis Wain" (an artist known for his anthropomorphized cat paintings.) They did this meticulously, shot by shot, gradually imprinting stylistic cues onto their existing footage.
The results aren't instantaneous. It can take days to run video footage through these programs, and if you don't like something, you have to go back into the system and make the changes. Woodman likens it to baking a soufflé: you put all the ingredients together, shut the oven, and wait.
But the soufflé comparison shouldn't suggest an easy-bake guarantee for success. shy kids' Matthew Hornick emphasizes the difference between the enthusiasts (like a music journalist prompting DALL·E 2 for David Bowie doing a somersault) and the developers like Cederberg, who has been learning these systems for months and coding in order to achieve more customized, refined results.
In this way, the relationship between the artist and AI is more collaboration than autopilot. Case in point: the AI-generated Drake and the Weeknd song "Heart on my Sleeve." When the viral track was released, people were so stunned by the cloning accuracy that they disregarded the fact that an anonymous ghostwriter had to produce, write and sing the track before putting it through the AI system.
Cloning aside, creative intervention will be vital in preventing the artist from being lost in the artificial.
"As an artist, you are everything from the story you tell, to your taste and how you execute it. You aren't the final result — you are the process," Cederberg explains. "We also have those thresholds for: is there enough of us in this? I think we would shy away from just doing something for the sake of ease."
These AI systems are also undeniably more accessible than the alternative. Sans AI, the video for "fresh off a feeling" would've required rotoscoping, an animation technique where filmmakers trace over motion picture footage. The process can take months of labour and thousands of dollars. For cash-strapped creatives, cost efficiency is an AI upside.
"The democratization of these things shouldn't scare you. In fact, it's going to be very clear who is themselves and who is not themselves. So are we ourselves? Well, that's a better question," Woodman posits. "I hope we have something unique to say. But I want people to see these things. I want you to know how easy it is to generate that image because it's a good thing. I think it's a good thing."
Ottawa-based artist and producer Thor Simonsen has been grappling with the risks and opportunities presented by AI. Simonsen is the founder of Hitmakerz, a record label based in Iqaluit that specializes in Indigenous artists. His team has been using ChatGPT to draft press releases, write funding applications, decode legal documents and even plan national tours.
"This morning, when I was mapping out a tour for Latin American artists in Canada, I just named the cities we want to go to and I said, 'Give me 25 venues in each city and include their contact info and phone number and email,'" Simonsen tells Exclaim! "I could have done that using Google because it's all publicly available, but it saves me hours of going through all that stuff."
Many are concerned that AI will further separate the art from the artist, but Simonsen sees it as a tool to bring them closer together. A means to an end. By reducing the administrative aspects of the job, the hope is that an artist has more time to dedicate to their craft.
As for the roles behind that administrative work, he predicts that it isn't simply a matter of AI taking people's jobs within the music industry. Rather, it'll be about people who know how to use AI taking the jobs of those who don't.
"It's someone who knows how to use the printing press versus someone who knows 18 scribes. One of them is going to be quite a bit more effective," he argues.
Inevitably, the more that artists integrate AI into their creative process, the more challenging it will be to decipher whether you're listening to, or competing with, an AI product. Will artists disclose this information, or will AI get stamped with an Auto-Tune-like stigma that discourages artists from being transparent?
Simonsen recalls someone proposing a fix at a conference in Colombia earlier this month: artists could add "100% AI free" disclaimers to their bios. "Suddenly, that's like the 'organic' of music," he suggests.
Recently, Simonsen teamed up with Nunavut artist Agaaqtoq to perform a song entirely written and composed by ChatGPT. They instructed the system to deliver a catchy pop song in Inuktitut with an ABAB rhyming pattern. They wanted a track about a man traveling through space to find other musicians, and they asked ChatGPT to provide English translations and chord progressions. The result was "Gitaralauq Sulusiaq," loosely translated as "Guitar Players, Let's Dance."
While Simonsen and Agaaqtoq were impressed by the system's creative capacity, its imperfections were exposed. Much of the AI-generated Inuktitut was pure gibberish. Agaaqtoq, born Abraham Eetak, acknowledges that this is probably due to lack of Inuit involvement in the development of AI — yet another sign of the infancy of these technologies, and the requirement for human interference at some level.
Eetak recognizes the exciting potential of AI as a tool for cultural exchange and idea generation, but after his brief flirtation with ChatGPT, it won't be a fixture in his process.
"I feel like it's just going to take away from my skill. Maybe I'll get too lazy and write a whole album using AI and just make it way too easy. And that takes all the fun out of it," he tells Exclaim! "I like the challenge of writing a song about a topic, whatever that may be. Getting into that process of just going back and forth in my brain with how I want the song to be. With AI, I just can't do that."
Like Eetak, some artists will dip their toes into the AI waters and decide they've had enough. Some will wade right into the deep end, while others will opt to yell "shark!" from the presumed safety of the beach. Simonsen, meanwhile, opts for the nuance between threat and opportunity.
"When you see a painting, it might move you in some way, and I definitely think AI can do that," he reflects. "But I think what humans care about is the context and the story behind it, and what it means to them. If AI creates something, it might be beautiful and technically fantastic, but if it doesn't have any connection to that person, and if it doesn't have a meaning, then I don't think it's valuable. I don't think that's the purpose of art."