What happens when AI composes songs?
There are countless ways AI can play the right, i.e., personalized correct song suggestions, thanks to recommender systems. Other vital introductions in the music industry have been achieved by establishing and evaluating streaming data to discover commercialization opportunities. An example of this is the investigation conducted by Ankit Desai: According to Iflexion, Desai looked at the streaming data of Swedish singer Tove Lo and found out that she has a considerable fanbase in the EDM scene. Through her collaboration with a well-known EDM artist, namely Alesso, she subsequently released a song that made platinum status.
There are other impacts of AI on the music industry, such as finding new talent with machine learning algorithms and optimizing the listening experience thanks to mastering products based on AI like LANDR. According to Bernard Marr, this service was used by 2 million musicians to master more than 10 million songs, providing access to a service many couldn't afford before. After reading Nic Fildes article When AI takes on Eurovision: can a computer write a hit song?, I was intrigued by this impact. AI can apparently compose its own songs, but how can algorithms piece together what musically works and what not?
Koalas, Kookaburras, and Tasmanian Devils
Much to my surprise, there is an Artificial Intelligence Song contest organized by Dutch broadcaster VPRO in cooperation with radio station NPO 3FM and NPO Innovation. Inspired by the Eurovision Song Contest, the AI version took place for the first time in 2020 with 13 entries (with the usual outcome for Switzerland - practically no points and last place). According to the experts' panel, there are many obstacles the teams faced when composing a song with AI. Describing the process of producing music with AI as a constant state of push and pull, "where you try your best to direct it in a certain direction with the data you have, priming it with your own tune, and then it steers you in another direction." Another task to tackle is listening to an unearthly amount of samples to find the gem pleasing to the ear while not losing sight of the ethical implications of using AI in songwriting.
A painful lesson a team of Dutch academics had to learn. They experimented in songwriting using AI algorithms, inspired by Reddit. Remember when after not even 24 hours of operation, Microsoft's AI chatbot was shut down because Twitter users had trained it to become an insulting nazi-lover? Well, the AI lyric generator urges listeners to "kill the government, kill the system." Another exciting mix was created by the German team called Dadabots x Portrait XO. Training nine neural networks, they mixed 1950s Acappella music babble for the lyrics with death metal and other strange sounds. Fildes describes these strange sounds as "AI-generated death-metal vocal styles and a chromatic bass line spat out of a neural network trained on Bach's canon"). The song is called "I'll Marry You, Punk Come" - how do you like it?
With Google's Magenta project, an open-source platform producing AI-written and performed music, and Sony releasing an AI-created song called "Daddy's Car," there is also resistance against this "Big Tech interference." There are concerns that AI may, in some respects, be deployed to replace human creativity or to undermine the artist-driven cultural economy. Only the future will show whether this really happens or whether AI is more conducive to creative work. Overall, this contest shows how humans and AI can co-create music and find new musical genres unheard of today.
Before I let you go, here is the winning song "Beautiful the World" by the Australian crew called Uncanny Valley. Did I mention they trained their machine model with audio samples of koalas, kookaburras, and Tasmanian devils? Enjoy!
As always - stay curious!