AI is now capable of making music, as evidenced by creations like Jukebox and Shimon.
In April 2020, for example, OpenAI uploaded a music collection to SoundCloud, all proudly created using the AI software Jukebox. Jukebox learned to produce its own music while mimicking famous artists’ styles after getting trained on 1.2 million songs and other data about genres and artists. It not only figured out how to compose songs but sing as well.
Watch Jukebox singing its own song in this video:
A month after, Shimon was officially introduced to the public. It is the creation of engineers at the Georgia Institute of Technology and was trained using a vast data set, covering progressive rock to jazz to rap music. To create music, Shimon relies on its acquired knowledge and uses algorithms to come up with its own brand of music.
But while Shimon does produce music, its creators want to show how robots can not just work for us but with us. Its unique compositions expose listeners to new musical ideas that humans may not even think about.
Watch Shimon at work in this short video:
AI in Music: Past, Present, and Future
As we’ve seen so far, AI in music is now a reality. Jukebox and Shimon are just two of the AI music composers around. There are others like Spawn, an official collaborator of Holly Herndon, and Magenta, the band Young Americans Challenging High Technology (YACHT) ’s latest album comrade.
Way before that then, we’ve all been experiencing the power of AI in music. We’ve all been prompted to listen to streaming services’ song or album recommendations, and while you may not know it, your music choices have been influenced by AI all along.
Over the years, the role of AI in music has changed. Here’s how:
Each listener has his or her favorite type of music and even artists. And they very rarely go beyond what they know and love. And so we’d often hear the same songs and albums streaming from someone’s home. Music streaming changed that. Today, Joox, QQ Music, and KuGou use AI to analyze listeners’ preferences to recommend specially curated playlists for personalized experiences. The result? Our playlists may have grown to include artists we may not otherwise have discovered on our own.
In the future, AI’s recommendation capability can be extended to studying listeners’ vitals such as heart rate, stress level, breathing, and mood through wearables to offer suggestions. Fast heart rates indicative of physical exertion like exercise could issue prompts to play dance music, for instance.
In the past, we were all limited to listening to albums we had copies of. And let’s face it, while we may love certain artists, we may not like all their songs. Thanks to streaming companies, we now have the option to download and play songs we love.
We no longer have to skip tracks on albums manually but can discard songs altogether from our playlists. We also no longer need to pay per piece, which essentially was the standard for compact discs (CDs). Each CD costs around US$15 for about 12 songs (20 if you’re lucky). Streaming services are much cheaper at US$10 a month and have more comprehensive selections to choose from. You can play any music you want an unlimited number of times and, as we pointed out earlier, get recommendations based on your preferences.
In the future, AI can enhance listening pleasure by letting us “experience” music. Aided by virtual reality (VR), we can opt to not only listen but also virtually see our favorite artists perform just for us. Who knows? Wearables can also be made to let us “feel” music in our bones, muscles, and brains.
Much as ordinary folks get inspiration for work from music, our favorite artists also get influenced by things and events around them. Movies sometimes inspire musicians to compose songs, hence you see famous artists’ songs on movie soundtrack albums. Given how connected we’ve all become, it’s no wonder that data is driving even the music industry.
Producers and artists research people’s preferences using data (powered by AI) provided by music streaming companies. That’s how they identify what kind of music sells, where to sell this, what demographics they should cater to, and so on.
AI in music also extends to audio mastering. Many tech companies now help musicians polish the quality of their sound like professional studios but at half the cost and time.
In line with biometric-driven music recommendations, AI can also affect artists’ creative process. As the beginning of this article shows, we’re actually already seeing that. Younger musicians are incorporating AI musical creations into their songs, even entire albums.
While AI in music is certainly making headway, we can’t do away with legal, specifically copyright issues. There are certainly a lot of considerations to make. For now, though, isn’t it exciting to see how much more AI in music can grow?