AI Is Learning How To Compose Music

Songwriters and producers know how much work and practice goes into the craft of writing music. However, artificial intelligence may soon change that. AI music is a quickly-expanding field with many projects and companies developing revolutionary new means of music composition. The famous Abbey Road Studios, known for hosting legendary bands such as the Beatles and Pink Floyd as well as contemporary stars such as Adele and Kanye West, is running a startup incubator called Abbey Road Red aiming to foster links between the music industry and music technology companies. On the other side of the pond, startup accelerator Techstars Music, based in Los Angeles, has been partially funded by major labels Sony Music and Warner Music Group and included two AI startups in its program this year. Among some other companies in the growing field are Humtap in San Francisco, Jukedeck in London, Groov.AI in Mountain View, while Google and Sony’s Computer Science Laboratories (CSL) in Paris have similar endeavors titled Magenta and Flow Machines, respectively.

Automated music composition brings with it questions as to what will happen to professional human songwriters. Ed Newton-Rex, CEO of Jukedeck suggests that AI music is not meant to become better than professional musicians, but rather, is meant to be useful for people. “Music is so subjective,” he said to The Guardian. “It’s a bit of a false competition: there’ is no agreed-upon measure of how ‘good’ a piece of music is. The aim [for AI music] is not ‘will this get better than X?’ but ‘will it be useful to people?’. Will it help them?” AI music may not quite replace professional film composers, but it could be useful for YouTubers who need some background music for their videos. Jukedeck, for example, produces a song based on selected genre, tempo, mood, instruments, and song length and sells royalty-free licenses for $0.99 for individual or small businesses, and $21.99 for larger companies.

Music industry consultant Mark Mulligan said that this type of AI music focuses on sonic quality rather than music quality, and having pleasing chord progressions, the right balance of instrumentation, and the appropriate quantity of builds and breaks renders it sufficient. “AI music is nowhere near being good enough to be a ‘hit’, but that’s not the point,” he said to The Guardian. “It is creating 21st-century muzak. In the same way that 95 percent of people will not complain about the quality of music in a lift, so most people find AI music perfectly palatable in the background of a video.”

Rather than generating new material, some AI-music startups are developing tools that do otherwise. British tech startup AI Music, one of the AI firms taking part in Abbey Road Red, is developing a tool that will “shape-change” songs that already exist anywhere from adjusting their tempo to match someone’s walking pace to automatically generating remixes. In another approach, Australian startup Popgun has created an AI dubbed “Alice” that, by listening to thousands of songs by experienced playlists, engages in a call-and-response duet with human pianists. Vochlea, the second AI startup taking part in Abbey Road Red, transforms vocal samples into drum samples, pseudo-guitar riffs, and brass sections based on the melody you sing to it, helping people capture their creative ideas in their heads.

While the debate over the potential threats of AI takes place in other fields, AI music companies are more optimistic about the future of AI music. Jon Eades, a runner of the Abbey Road Red incubator, feels that AI will be yet another double-edged sword that presents itself to the music industry. “I think there will be collateral damage, just like the internet,” he told The Guardian. “It created huge opportunity, and completely adjusted the landscape. But depending on where you sat in the pre-internet ecosystem, you either called it an opportunity or a threat.”

AI may eventually become integrated into the music streaming giants such as Spotify and Apple Music. François Pachet previously worked with Sony CSL, which released two AI-composed songs in 2016, ‘Daddy’s Car’ and ‘The Ballad of Mr Shadow’. The tracks were created in the style of the Beatles and based on cues from famous American songwriters, respectively, and featured lyrics and production polish from humans. Now, Pachet is working with Spotify, and although the streaming company declines to say what he will be doing, AI could potentially compose music for Spotify’s mood-based playlists for activities such as falling asleep, relaxing, and focusing.

Despite the overall promise of AI music, there are complications that arise. Who will be the legal creators of AI-produced music? Can AI be sued for copyright infringement? If AI shape-changes a song to the point where the original becomes indistinguishable, does it become its own original work? These questions are being currently being debated as new technologies are being developed. Ultimately, time will tell how these complications will pan out and what will happen to resolve them. The AI startups hope that AI will serve to inspire human musicians rather than threaten them. AI music could essentially become a new toolset for musicians to work with, and rather than taking the place of human musicians, they can help them expand the limits of creativity and bring music production to new unforeseen places.

h/t The Guardian