4 ways AI is helping musicians—and the entire music industry
When we give a machine values and it solves a calculation for us, that’s simply computing. When we give a machine data and it learns from its experiences and then makes recommendations, that’s artificial intelligence.
So what happens when we give AI one of the most human of art forms: music? Quite a bit, as it turns out.
AI uses machine learning models to produce new patterns and correlations based on the data it was trained from. In the case of music, almost 100 million recorded songs exist. Many scores of scores provide a deep base of data that’s hard to beat, and plucky researchers have taken note: AI’s ability to learn and iterate on its knowledge can change the way musicians work. And now, it’s impacting the entire music industry.
New and different sounds
For anyone concerned that robots are coming for their music job, keep in mind that AI-generated tunes tend to lack originality—take the less-than-fab song "Daddy’s Car," which was trained on 45 Beatles tunes, as an example. Although AI can produce solid tunes and perhaps a few hits, it won’t make your parents demand you turn your music down the way Elvis Presley, the Beatles, the Sex Pistols, Jean-Michel Jarre, Grandmaster Flash and the Furious Five, and Nirvana have—at least, not any time soon.
“The best artists and the most thrilling breakthroughs in music have been people who came up with a new and different sound,” said Toby Butterfield, a partner in Moses & Singer's intellectual property and litigation groups, at the Mondo.NYC conference on AI and the music industry.
But that doesn’t mean AI has no play. From recommending artists to literally playing second fiddle to a human composer, here are some ways AI is turning the music industry on its ear.
1. Making music
Western music has 12 notes and 24 major and minor scales. Add in tempo, harmony, and articulation, and you have millions of potential songs waiting to be created. Enter AI.
Thanks to AI, musicians and non-musicians have access to sophisticated yet simple-to-use creation, remixing, and learning tools like Flow Machines, Pixel Player, Amper, Melodrive, Boomy, Landr, and more. These AI-powered tools generate seemingly limitless tunes for you to “play” with, to spark your creativity. Or they can even be creative on your behalf: Musician Taryn Southern recorded an album composed entirely by AI.
Better yet, these tools can be found at varying price points, including free, so now the band whose drummer is also the driver who is also the PR guy can afford them.
Beyond composition and collaboration, AI vocalizes music, too. It took more than 100 iterations, but electronica composer Ash Koosha built an AI, named Yona, to not only compose music but also sing lyrics.
In a dystopian future: Yona will declare herself an independent sentient and retire to a llama farm.
2. Recommendations everywhere. Will AI make them more accurate? Or just scary?
Online retailer Amazon may have popularized the recommendation algorithm, which suggests items that may be of interest to you based on prior purchases. Now, recommendation algorithms are a high note of music streaming services.
Music streaming services like Pandora, Spotify, Apple Music, and SiriusXM employ different vectors—including rhythm, genre, and user geography—to suggest new songs to their clientele. Recommendation AI does this in part by culling songs from playlists created by other users who share your musical taste, something known as collaborative filtering.
How many playlists? In Spotify’s case, more than 2 billion.
Listeners need all the help they can get separating the audial wheat from the chaff. According to analysts at BuzzAngle Music, 534.6 billion songs were streamed via online services in 2018. If you consider the average song is 3.5 minutes, that’s roughly a bazillion-jillion minutes of music listened to. And thanks to AI, no human will have to filter out Rebecca Black’s “Friday” for us.
Prior to this, audiophiles received recommendations from their friends, the radio, and the obnoxious guy in the record store.
In a dystopian future: Pandora’s recommendation software will recommend you drop your boyfriend in favor of the obnoxious yet totally cool record store guy.
3. AI-driven promotion
In 2018, 2.65 billion people used social media, a number expected to increase to 3.1 billion by 2021. That’s a whole lot of data for AI to draw from.
Artificial intelligence can derive insights on your fan base that go beyond “also listens to Jonas Brothers.” With an AI that can analyze music fans' activity, you can determine that a fan has a penchant for the Paleo diet and prefers surfing over jogging. Hello, tie-in advertising and cross-promotions.
AI can also identify which social media influencers have the most impact when they tweet your work, as any good AI media company can explain when you sign its contract.
Note: As a bonus, social media influencer and musician Lil Miquela is herself artificially intelligent.
In a dystopian future: Your social media sites will converge into one government-backed mega-corporation. You won’t be able to get a job until you buy at least one recommended song.
4. Predicting future hit songs and artists
According to Butterfield, AI “may also be useful for recording companies to analyze what is likely to be popular with consumers.”
AI uses songs that appear on multiple playlists to determine which songs you prefer. Likewise, knowing which songs you skip on an album or playlist is equally informative.
After determining your favorites, Butterfield said, producers “can then marry that with demographic information.” This way, “recording companies [may] analyze whether a particular composition is likely to be popular with consumers.”
Buttferfield added, “It’s conceivable that AI applications will augment the traditional roll of A&R [artists and repertoire, or talent scouts], those who decide what band is worth signing.”
Soon, A&R scouts could simply take their AI to a bar and walk away, leaving HAL-9000’s musically inclined sibling the task of making business decisions after listening to four hours of J-Pop.
In a dystopian future: There is clearly no way this could go wrong.
Although composing your own music is becoming more simple than ever before, it leads to a complex issue: rights.
An AI is trained on prior works: It’s artist A who provides a springboard for artist B. While artist B helps create new music, shouldn’t artist A be compensated for their work, too?
“What we have is this landmine of potential legal issues that have yet to be fully explored," said Candice Cook Simmons, a business strategist and managing partner at the Cook Law Group who also spoke at the Mondo.AI conference.
“We want intellectual property and the protection of intellectual property to maintain two things: the freedom of creativity but also the right of the creator to be compensated for their work," she said. "We want free flow of ideas—we want to serve as inspiration—but we don’t want to confuse inspiration with the right to create derivative work without paying the owners.”
Simmons said the music industry needs to be proactive if it wants to stay ahead of lawsuits. She suggested that companies shape their policies in advance, “not playing scared but playing smart.”
This applies to artists, too. Musician Holly Herndon skirted the issue of IP rights by doing it herself, literally. As part of her dataset, Holly Herndon trained her AI on recordings of herself and her band, as well as members of her audience. The results are as avant-garde and startling as they are legally prophylactic.
A question for you philosophers in the audience: Human artists are also trained on the music of their peers and predecessors. If they are then inspired to compose new music, that’s a creative act. But if an AI does it, is it truly creative if it’s not genuinely adding something new? Or is it?
In a dystopian future: Musicians will be philosophers who will also be patent lawyers.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.