I came across this video from content creator Rick Beato this week. In it, he plays some clips of AI generated songs, where the AI had been tasked to copy and recreate the style of a famous artist.
Most of these are from modern singers/rappers like Drake, but some of these include an AI generated version of Kurt Cobain singing Chris Cornell’s ‘Black Hole Sun’. Beato closes the video by sharing his thoughts on the commercial direction AI generated music is going to take the music industry. Have a listen and see what you think:
What do you think?
I’m still distilling my own thoughts on AI in music, but one thing does spring to mind with the modern music examples. Namely, that the fact the AI could copy the modern music so closely but be less convincing with the other examples. To me, this touches on the the formulaic, often indistinct and (frankly) dehumanised nature of a lot of modern music.
Consider this: in music of years gone by – even as recent as the 2000s – rhythm and tempo would vary. Chords would be voiced differently or with different intensity as the song progressed. Sometimes notes would be played sung ever so slightly out of perfect tuning, but would sound consonant with the piece as a whole. These things occur because musicians are only human.
In contrast, modern music is increasingly ‘locked to the grid’, i.e. the tempo is mathematically locked in and perfect. It never wavers. The drum beats land in exactly the same place every time. The tone and timbre never fluctuates. Every note is autotuned/heavily tuned to be perfectly pitched. Nothing pitched will lie off the piano keyboard. Vocals all echo what the last number one hit did, rather than tread new ground. Artists are turning increasingly to computers and software to either generate these notes, or iron out the portions that were generated by real musicians.
Caveat: To be clear, I am not trying to say modern music is worse than older music – you can make that aesthetic judgment for yourself, and if you like a lot of that music, feel free to enjoy it.
What I AM saying, is that modern music and the way it is produced, has ironed out so much of the human elements that make music actual art (making it distinct from just being a commodity/product) and we are already turning to computers to make our current crop of number one tracks, that it stands to reason an AI would be able to synthetically generate a convincing facsimile of such dehumanised art. Virtually none of the components that make art a human endeavour seem to remain in modern commercial music… so why would you need a human to do it? Other to maybe give the AI some modest direction?
Truthfully, I don’t know where it’s all going
All of the above are just my musings on the human elements (that in my opinion cannot be put ‘in the box’) vs the dehumanised elements that now populate increasing amounts of commercially produced music. Maybe, the more convincingly an AI can generate a piece of music that harkens to a particular artist, the less human/artful that original artist’s style actually is in the first place.
Food for thought, and I’d welcome yours.