I foresee a marked deterioration in American music and musical taste… in this twentieth century come these talking and playing machines that offer to reduce the expression of music to a mathematical system of megaphones, wheels, cogs, discs, cylinders and all manner of revolving things.
John Philip Sousa, a legendary American composer and conductor, in a 1906 essay entitled “The Menace of Mechanical Music”
With a few edits to the above quote, you can find people wondering similar things about the impact of Artificial Intelligence (AI) on music in the twenty-first century.
In this piece, I wanted to pen my thoughts on how AI can shape music in the coming years. Disclaimer though, that that this is just a fun exercise as a hobbyist musician who is also curious about AI.
My predictions for how AI will change music
Everybody won’t become a musician overnight
Despite the proliferation of generative AI tools for writing, image creation and music production, I don’t think everyone will become a musician. There are two reasons for this.
- The creator/consumer imbalance: 99% of people throughout history have always been consumers. The internet and the plethora of user-friendly blogging tools have made it easier to write than at any point in time in history. Yet ‘consumers’ are probably a hundred times more than ‘creators’.
- Lack of incentives: While there are many more casual content creators (creators of Instagram reels, posters on LinkedIn), they are all driven by some incentive: signalling your professional credentials, or earning status or approval from your social circles. Creating music has little incentive for anyone to do so. Musicians earn very little from their music. They earn much more from concerts and merch sales, session music gigs, and guitar lessons (content-based, virtual and in-person). AI music tools can make you a musician, performers earn much more than musicians. Other than economics, there is also very little social influence to gain as well. Music, kike reading books, is not as important a daily activity for the majority of people as it used to be. So what do you gain from creating your own music?
If people use AI music tools to write their own music, it will probably be because of the novelty and the behaviour will not sustain for long.
Musicians are already struggling, but performers will continue to thrive
Extending the above prediction, I don’t think musicians will die out. On the contrary, I think performers (live musicians) will become even more valuable. I fully related with what David Byrne had to say about technology shaping music in his book ‘How Music Works’:
“As music becomes less of a thing—a cylinder, a cassette, a disc—and more ephemeral, perhaps we will start to assign an increasing value to live performances again. After years of hoarding LPs and CDs, I have to admit I’m now getting rid of them. I occasionally pop a CD into a player, but I’ve pretty much completely converted to listening to MP3s either on my computer or, gulp, my phone!”
Additionally, a big component of live music performances is not just the music and the lyrics, but also the other extra-musical elements of the experience: the setting, the alcohol, and the company. It’s also being able to see the musicians rock along with the music they perform. Live performances will reign again, and human musicians will thrive in this scene.
Musicians will become more productive!
I believe AI can turbocharge productivity for musicians.
Recently, Steven Wilson (Porcupine Tree fame) shared an Instagram reel where he collaborated with ChatGPT to write a Christmas song. A Christmas song is not in his wheelhouse, but ChatGPT ended up writing some lyrics that rhymed and shared some of the tonal characteristics of his previous work. I can see a lot of musicians using ChatGPT to write and iterate on lyrics.
Additionally, similar to digital audio recording technologies, AI will allow musicians to experiment with different tunes and arrive at the sound in their heads much more quickly. I know for sure that I, as a hobbyist musician, am excited by the huge productivity jump I can see with AI-assisted songwriting. It will help me get better at songwriting much more quickly simply because I can iterate faster.
AI will probably not write many novel songs
Creativity and innovation can be difficult to quantify, so for this piece, I will define it as the ability to create novel or new-sounding music. I don’t think AI will necessarily change much on the novelty front. Why? It has to do with limits to what we can perceive in terms of sounds, frequencies, words – that is to say, music and lyrics.
Think about this. Even if ChatGPT has access to all of the written word in history, they will still write simple, legible sentences that mimic the patterns and rhythms found in human-generated texts. Music will follow a similar trajectory. ChatGPT could theoretically write a chord progression with 40 chords, but will our ears follow the music or simply tune out like we would in the middle of a long, droning speech?
Another factor influencing the kind of music musicians create is the demand. An interesting research study shows that we, as listeners, also tend to be drawn to music that influenced us the most (typically our teens), but is just different enough that we find novelty in familiarity. This, combined with the principle most AI technologies work on (i.e. finding patterns), ChatGPT and the likes will regress to writing songs that will sound familiar without deviating too much from the structure.
However, as AI-assisted songwriting takes off, digital-only compositions will continue to dominate more, and the use of instruments like guitars and saxophones could decline even more in popular music.
AI will probably accelerate the spread of niche music
AI will likely accelerate the formation and discoverability of more niches and make it easy for people to find other people from their tribe who like similar kinds of music and art in general. No longer constrained by the physical formats in the early years of recorded music, different forms of music are finding their way into increasingly more pockets of the world. This is not limited to the kind of music people in one place will listen to either.
I do have a personal gripe though. Algorithms can sometimes be self-reinforcing loops and I often find myself listening to similar songs again and again. While AI helps you find similar things quickly, I think it also keeps you from finding stuff outside your immediate interests. I wish there were ways to fine-tune the AI to a level where I could discover more artists and music. But I suppose this is a very ‘musician’ take, and for most people in most settings, AI works just fine for their purposes.
Summary
With respect to music, I think that the developments in AI promise both innovation and continuity in different ways. While it will enhance musicians’ productivity and potentially lower the barriers for many aspiring musicians, AI won’t replace the essence of live performances and the human connection in music. It may refine discoverability but also poses challenges of narrowing musical preferences. I am excited to see how AI shapes the music landscape in how we create, distribute and listen to music over the next several years.
—–
What are your thoughts on how AI will shape music? Did you find yourself nodding to or disagreeing vigorously with any of these predictions? What other fields would you want to read more about when it comes to the impact of AI on our lives?
Utkarsh's Blog 