When I wrote “Ljova’s Hot Takes on AI Music” in September 2024, I had never considered that someone would be quoting me, almost verbatim, in a casual conversation many months later –  the part that stuck with my reader the most is this:

“Writing music is a calling, a form of expression, and those who *need* to do it will continue to persist so long as there’s oxygen in the atmosphere by any means necessary.  AI music – at this point, anyway – is a utility for those who seek a shortcut to the finished product.   For many of us creatives, it’s not the product we’re after – it’s the process, the struggle, the personal journey of it. Sure – you can have a journey with a prompt – but it’s a different kind of struggle.”

How valiant of me to write this just a short while ago… “Nowithstanding the foregoing”, as lawyers say, I’d like to report that, since that time, I have:

– played on a recording session for a singer-songwriter, where, afterwards, I was gleefully informed that the music was composed with the help of AI.

– listened to some incredibly realistic AI realizations of backup vocals for a pop song.

– heard some very creative orchestral arrangements created with AI

—been asked to transcribe arrangements created with AI (my job was to transcribe audio to musical notation)

Everywhere I hear something to the tune of — “Adapt now. Get AI to do your work. Try to learn. Don’t get left behind.” Some of my colleagues have taken this advice/warning to heart and created songs, arrangements, new recordings — and I’ve been, at times, the cog in the wheel that made their AI-assisted project appear as a human-made reality.  

– I’ve heard from some media composers that AI has broadly killed music for advertising. (It certainly doesn’t help that most ads now are about every brand’s new AI-enabled features.)

– I’ve heard from some film composers that “my generation is probably going to be ok, but the next generation of film composers is (expletive for doomed)”

— Recently, while watching a documentary I found that the score sounded generic – “is it AI?”, I kept wondering. The film was released in 2025 — why would you write generic music in 2025?

– I noted that the great Italian composer-cellist Giovanni Sollima has a new piece called “I asked ChatGPT to write a piece in my style. I got scared”.  (Not sure if there’s a recording, but the sheet music is published.)

– I’ve heard a DJ retell a conversation where a listener said their favorite songs came from“this playlist on Spotify” but couldn’t name any individual artists.  In a similar vein, I recently shopped at a store that was playing lovely music, and when I asked the clerk whether they picked the music, they said “Yes” — and when I asked what it was, they clarified that they merely picked “this playlist on Spotify.“

I’ve been thinking about these issues over the past few days while experiencing some great art — the film “The Turin Horse” by Béla Tarr;  the intense musical worlds of Edgard Varese’s “Ameriques” and Igor Stravinsky’s “The Rite of Spring”… an early motet by Jean Mouton… Could AI make this? Sure, it could — but would I care to see it? Will art continue to be a communal experience or will it proceed with the personalization trend, where AI not only curates the playlist, but composes all the tunes as well?

My sense is that a lot more music creation now contains AI-composed or AI-crafted components than people are letting on…. (Over 50,000 AI-generated tracks are uploaded to music streaming resources on a daily basis, according to Deezer.)

How is it? AI music is neither good nor bad -– musical taste is not a monolith, and there is plenty of human-made music that I consider foundational that others would reject as unbearable.    Were I to be given an A/B listening test of sleuthing out which piece is AI or human in any particular genre, I would probably stumble at least 50% of the time. Mostly, it’s not the music that gives away the AI component – it’s the sound quality. (For now.)

While quality is always subjective, what is true is that AI music is “economical” – cheaper to implement. Infinitely faster to iterate. Cheaper, faster, for now — the environmental impact remains to be seen.  At this stage, AI music is “good enough” for 95% of use cases where quality is secondary to cost.  

Unlike a Juilliard-trained composer with an advanced degree, an AI graduated from every college,  toured with every band,  devoured every publicly available doctoral treatise, heard every record, internalized every book on orchestration, read every good and bad review, crunched the attendance numbers, found its niche….   A living composer needs to eat, sleep, fall in love, feed the family, find a place to live, pay taxes, nourish the soul… AI never gets tired or unwilling to please….   Game over?

Yes, AI Can “do my job” – but that’s not my job.

So what is my job?

In short, my job is to live, and do whatever it is that I do. 

I will live however many years given, and create whatever music — compositions, experiences, collaborations, chance encounters — that I can. A life (in music) is process-oriented, reflection-based, and relationship-driven.  I have an identity fused with the places and people where I grew up and formed. 

AI will neither live nor die — AI can be unsubscribed from, downgraded to a previous version, censored, or upgraded.  AI can cosplay an identity.  AI will keep creating works inspired by the human canon and increasingly by the compost of its own making. Like a human, AI will also have assignments that are process-oriented, reflection-based and relationship-driven.  But, unlike a human, AI will not live in fear of making a mistake, of losing its house or a loved one.   Its biggest fear, perhaps, may be that of losing a paid subscriber.

Without question, I understand the appeal — the time one spent creating something can be slashed from weeks to hours or minutes.  If you have an idea, give it to AI and let it implement.  

Should I upload all of my released tracks into AI and ask it to create endless new versions and upload them to all of the streaming platforms, expand my discography and licensing library exponentially with just a few tweaks?  Automate posting of this new content to social media in 59-second bites?   It wouldn’t interest me — would it interest you? Certainly there’s one audience for all of this extra content — AI itself.

(Come back tomorrow for an opera based on this post. Come back the next day for a ten-part documentary series.)

For now,  AI-generated art plays on the sidelines. I’m grateful that we are still fascinated by human achievements — between the Olympics and the Super Bowl, we recognize and celebrate our limits, their triumphs and heartbreaks.   

Music — live music, concerts — still, largely, belongs to human creations. But the more we feed our work into AI for a helping hand, the more we lose what makes our work “ours”. 

Tumbao-la-laika.

To this date, I hadn’t tried using any of the AI music tools — but a few weeks ago I attended a concert for school students by the New York Philharmonic, as part of which “Tumbao” by Roberto Sierra was performed, a wonderful celebration of rhythm and culture. As I was listening and enjoying, I had a silly idea — to do a “Tumbao” from the Eastern European perspective. Then I had an idea for a pun — instead of “Tumbalaika” (a popular Jewish tune I knew from my childhood), to make a “Tumbao-la-laika!”. I jotted it down in my to-do list and made a tiny piano sketch the next day that I uploaded to Instagram.

This also felt like a perfect opportunity to try an AI music generator for the first time. Having uploaded my piano sketch to Suno, I asked it to generate a full symphony orchestra arrangement in a salsa style that would include an accordion solo, a modulation, and a percussion breakdown. Suno ignored my requests, creating, instead, a more conventional salsa band arrangement with a surprise alternative solo instead of accordion, transforming my 20-second piano sketch into a full length song. I give you both my original and the Suno-generated arrangement, below:

There were at least two good reasons for me to try this — for one, I already knew what I wanted; for another, I didn’t care to invest the time to create a full-on salsa band arrangement of this idea, and I was ambivalent about the end result. Creating a salsa arrangement of this tune manually would take considerable effort; recording it with a real band would take a healthy budget. This was not my intention — I just wanted to make a quick sketch to test the waters. Had I wanted to pursue it further, I could, of course, take the time and resources to put it out “for real”.

It would be quicker and much more fun for me to write a salsa arrangement without help of AI — despite the time it would take me to do this work “manually”, I would arrive exactly at the results I envisioned — taking the AI route is far more circuitous approach, relying on language-based prompting instead of musical, energy-based language. I imagine a painter would have a similar reaction – why would they spend time describing to AI what they had envisioned, when they could just paint it directly?

Having played around with AI generators for a minute, and having had all of the aforementioned experiences working with AI-assisted music as a performer/arranger, I come to the same conclusion — making music is a journey. It’s a mystery. It’s the process that I cherish, the silence, the insecurity, the inquiry, the unknowing.

My job is not to feel threatened by AI, just as I’m not threatened by the younger generation of players or composers, by each new name I hear, by every newborn child. There is space enough for everyone to do things.   

Thanks for reading this far – if you have thoughts to share of your own, feel free to comment below or pass this along to a friend.

-Ljova /
March 2026