The production is slick, the drums are spot-on, the vocals sound great, but the titan of Newfoundland and Labrador’s music scene has two songs created less than a year apart: It Could Be Worse. I felt something strange about “Tales of The Atlantic.” 1 minute with powerful algorithms.
Bob Hallett, a founding member of the Newfoundland folk-rock band Great Big Sea, said, “That’s wrong because this is a country singer. The lyrics don’t rhyme.” “That sounds kind of weird.”
Hallett had just finished listening to a cheerful song. The song was created using a generative artificial intelligence tool called Suno, with prompts that can express any of Great His Big His See’s jams: Celtic, folk, vibrancy, passion, and more.
Hallett said they missed the point. On a scale of 1 to 10, he gave these songs his 2, with 10 being the band’s hit song “Ordinary Day”.
But experts say valuations could rise quickly. Jimmy Lin, professor and director of the Artificial Intelligence Institute at the University of Waterloo, said technologies like Suno are advancing rapidly and the results will only get better.
Suno is one of several companies developing generative artificial intelligence software that allows users to create original songs using text prompts. Users can create instrumental tracks and songs with lyrics either generated by the program or provided by the user. But if a user provides copyrighted lyrics (for example, the first line of Nirvana’s “Smells Like Teen Spirit”), the program won’t create the song.
Suno also sounds like the work of other artists, such as “Heart on My Sleeve,” an AI-generated song that used unauthorized similar vocals by Drake and the Weeknd, which caused controversy in the music industry last year. I don’t even make songs with that in mind. . When Suno was asked to write a “Great Big Sea song about cod fishing,” the resulting song had a pensive, Celtic feel, but the band’s authentic It didn’t sound like a typical song at all.
Google is developing similar software called MusicFX. You can sample this software from the AI Test Kitchen site. And last month, Adobe announced Project Music GenAI Control, which it describes as “an early generative AI music generation and editing tool.”
In December, Microsoft introduced a Suno-powered song generator for its Copilot chatbot, a program that uses artificial intelligence to simulate conversations with users.
The technology behind these programs is similar to the technology that powers ChatGPT chatbots.
These programs use large datasets to “train” algorithms, or step-by-step processes, to predict the next process from any starting point, Lin said. So while a chatbot trained on text can predict the next word in a written response, a music generation program is trained using sound to predict the next “acoustic sequence”, he said. .
In December, the New York Times sued Microsoft and OpenAI, the owner of ChatGPT, for using its stories to train its programs.
Lin said companies running AI platforms that produce music could run into similar problems if they train their algorithms on the work of artists who have not given consent or received compensation for the use of their music. He said that there is a sex.
“Whether this is fair use is an open question,” he said in an interview. “It will resolve itself. Courts always do that.”
Suno’s website does not indicate what data it used to build the program, and the company did not respond to requests for comment.
Hallett said he wouldn’t be surprised if the algorithm learned from some Newfoundland bands. He said the songs he produced had some distinctive sonic signatures, such as tight melodies and the intense strumming of an acoustic guitar, which he and his fellow producers had learned over the years from the Ennis Sisters and Shaniga. This is something he developed while recording albums by bands such as Knock and The Fables.
However, he was unfazed by the use of music for training in these programs, noting that artists have been battling for years with platforms like YouTube and Spotify that are already taking a big bite of musicians’ incomes.
“There’s a bit of a sense of surrender about everything…that’s very hard to police,” Hallett said. “Creative work is really about increasing concert sales or finding commercial introductions. Even at the highest levels, people don’t actually make money selling records. there is no.”
Lin said AI-generated songs will likely be used by advertisers who need catchy jingles for their commercials. And given the rapid evolution of these tools, he believes that could start happening soon.
“We’re not talking about years or decades. We’re talking about months,” Lin said.
But Hallett said humans would be best served by people who want to use music to connect with their audience.
“It’s easy to be afraid of AI,” he says. “But we’re all attracted to honesty in music. We want to hear people telling real stories and delivering real emotions. And computers can’t do that.”
This report by The Canadian Press was first published March 3, 2024.
With files from Associated Press
