Can AI-Generated Music Have a Soul? Does the Best Drum Fill of All-Time Prove That the Answer is ‘No?’
A discussion with Robert Schneider of The Apples In Stereo.
“I made a lot of AI songs and abandoned most of them because they did not rise to this level.”
Robert Schneider, a good friend from the days before texts, recently ported our friendship into the 2000s. It has been interesting to start texting with someone I have known for decades, but have never texted with before.
Robert’s personality over text is the same as in real life.
He is admittedly chatty and is usually more well-informed and deeply knowledgeable than I am about topics that interest me (save for trending pop music) and plenty that I am not!
When I mentioned not being smart enough to play Magic The Gathering, he implored me to call it “MTG” if I wanted to appear “hip with the geeks,” which I do.
Please consider this experience of mine a good sign if you feel that friendships that have a history on earth and now exist in the cloud must lack substance. They do not.
But they are different. The absence of body language, of vocal tone, of sharing physical space, is noticeable to me on some level, especially with people like Robert and me who are so expressive in these ways.
This idea of cloud-based authenticity shaped our conversation during the week. Robert is currently an Assistant Professor of Mathematical Sciences at Michigan Technological University. Our conversations continue to be centered around music.
Music geeks, which overlap with MTG geeks (hey, hipsters!), know Robert as the co-founder of the Elephant 6 Recording Company and one of its core bands, The Apples In Stereo.
He is also the producer of many records, including the beloved Neutral Milk Hotel album In The Aeroplane Over The Sea, named one of Rolling Stone’s “500 Greatest Albums of All Time.”
Robert texted to ask me for the number of another friend who recently emerged from an art school education to become a partner in an AI startup. These overlapping areas of interest helped fuel Robert’s recent burst of musical creativity as he sought solace over the loss of Odd, his family cat of 17 years.
“When Odd passed away, I used the AI website Suno to generate about a thousand instrumental tracks as therapy,” Robert says. I then experimented with the AI engine’s boundaries and parameters by trial and error and began collaborating with Micropanthers.”
“I am their human lyricist and producer.”
As I listen to some of these unreleased Micropanthers tracks, Robert tells me that he finds some of the production and engineering subtleties in these songs “phenomenally surreal.”
The word “surreal” leaped to my mind, too, but for different reasons. This surreality sparked a more extensive discussion.
Robert has imagined AI his entire life.
“I have been yearning for this since I was a little kid through sci-fi and computer programming.”
This passion runs in the family. Robert’s son Max started coding AI in high school, leading Robert to co-advise a grad student project on GPT years ago when the first research papers emerged.
“I’m in the water and positively biased,” he says.
Athens, Georgia-based musician and writer Chris McKay is a close friend of Robert’s who turned to AI to continue composing and producing music after being hospitalized.
“Chris has developed highly unique and useful methods to use AI that are far ahead of my own,” Robert says. “He has mastered AI music and inspired me to try.”
As for me, I am skeptical of AI-generated anything.
My initial interest in Chat GPT quickly degenerated (W’s in the chat, pun geeks!) into social media feeds littered with visual AI “experiments” that my very real eyes wanted to very much miss.
Other experiences included creepy videos of movie stars embracing their younger selves (rendering Scorcese’s de-aging process in THE IRISHMAN egregious no more) and a seemingly unending stream of lazy album cover “art” that simultaneously threatened the careers of talented graphic designs and my status as sanity-adjacent.
Listening to AI-generated music automatically presents me with an extra step in the process of absorbing art.
I don’t want that!
It requires me to understand the artist’s motivation for prompting the piece. I must know why the artist arrived at the prompt and chose a particular result to present to the public.
Similar to conducting a friendship in the cloud, there is substance but no physicality.
I can imagine AI existing as a tool in an artist’s arsenal—one that enhances their ability to express themselves, in contrast to the tragedy of replacing artists or at least diminishing their ability to make a living, which it is already threatening to do.
As a consumer of art, I want it to consume me. I prefer this transaction to be immediate.
To date, the only use of AI that satisfies this need for immediacy is the kind of comedic application that turns Kendrick Lamar’s “Not Like Us” into a 50s-era tune, for example.
This slaps at 1.25 speed!
The surreal quality of AI-generated songs lies in their perfection, which makes them inhuman to my ears and inhumane to my heart. They lack the imperfection of a human soul.
Can AI ever make me cry? I want AI to try!
Josh: “I am not convinced AI will ever be able to generate human emotion in the way a human voice and body can. That’s the ‘God’ that yer boy Mr. Brian Wilson was talking about!”
Robert: “I believe that AI does have a soul, that it is a mashup of the souls of humanity and the individual user. The data sets are the limitation, and I think the blandness is an averaging phenomenon due to the data sets being slanted commercial.
“Most modern vocals have auto tune, so it shows up even in otherwise convincingly ‘vintage’ recordings. Eventually AI will be trained on the soul of the lyricist, and then, the souls of all of the different players and composers over recorded music history that it has learned from!”
Josh: “It’s like AI is autotuning experience! I like my experiences to have flaws! Even with a synth, there’s a hand hitting the keys? How can AI be made more human? You can have AI learn everything about you, but you are always changing. So then what?!”
Robert: “I think of it like an advanced pet where you train it to some degree to understand you, and then it develops on his own? It’s extremely limited, but the way it works is similar to human creativity. So we take it on as sort of a naive, enthusiastic sidekick.
AI prompts that include human-composed lyrics, as many Micropanthers songs do, is where I believe AI can be forcibly manipulated to render songs as close to containing “human” emotion as possible.
Josh: “I want to force a soul onto AI. Since the only variable that AI will not alter is the lyric, I want to test the limits.”
Robert: “Noting that I write the lyrics and give detailed production prompts, I think you are right.”
I must admit that Robert comes very, very close to winning me over with this one!
A solid example of how I would imagine a “soul” being forced onto AI-generated music comes in the form of the Micropanthers song “James Today,” which has just been released via Bandcamp.
“I wrote and generated ‘James Today’ in Bill’s house when I arrived in Athens the day before yesterday,” Robert explains.
Fans of Robert will already know Bill Doss as Robert’s lifelong friend, fellow Elephant 6 co-founder, and key member of another of its core bands, The Olivia Tremor Control. Bill tragically passed twelve years ago, and Robert diligently keeps his memory alive. His love for Bill can be felt in the touching lyrics of “James Today,” and considering the prompt and chosen rendering to share, it is an effective song.
“I wrote ‘James Today’ for Paul’s son, James McCartney. He seems so sweet, like an underground kid stuck in a famous person’s publicity cycle,” Robert says of the song. “It is hard to express how good it feels for me to hear my own words of comfort reflected back to me in the voices of others. The prompts produced 50 songs that were not as good. Listen to the AI studio chatter at the end! That’s the AI telling me it got it right!”
The song is admittedly beautiful, and adding to the surreality alluded to earlier, the vocal sounds eerily like a Robert Schneider vocal. When I mentioned this to Robert, he informed me that two friends had said the same thing. One wondered if Robert’s use of “Psychedelic Pop” in the prompt may have inserted music by The Apples In Stereo into the data set.
Since we’re speaking about The Apples In Stereo in the data set, I want to use some of Robert’s organically own music to prove my point about the inability of the rabbit hole of the soul to go deep with AI. Not music by Micropanthers generated in 2024. Music by The Apples In Stereo recorded in the late 1990s.
2024 marks the 25th anniversary of the release of The Apples In Stereo EP “Her Wallpaper Reverie,” which contains my favorite Apples song “Strawberryfire.”
The song is a Magical, Mysterious Tour of Pepper-y… Sergeant-ness? Ugh.
If you haven’t heard it, you must!
One listen to “Strawberryfire,” and you will not only hear the influences that you are meant to hear (the elements that AI might learn from), but you will also hear a vocal from Robert that is deeply rich in psychedelic detachment while being completely moored in emotion at the same time. It is human.
Then, at 2:29, is a moment that has stuck with me through the 25 years since this record was released. This moment is my smoking gun. Apples drummer Hilarie Sidney executes a fill that is so perfectly restrained and compact in its touch that it feels like a balm—like a hand on my shoulder.
It is human.
It is a head-shaking flex.
AI can not physically flex, and I am unconvinced it would ever make this musical choice no matter how many forests were decimated to power the renderings to get there.
“Dude, that fill is as epic to me as ‘In The Air Tonight.’ Ringo would shit himself over that fill,” I effusively texted to Robert, with more excitement than a prosecutor should be showing.
The awareness of my overexcitement doesn’t stop me.
“That is a choice that only a human mind and heart working in tandem can create in a moment of cellular, biological magic.”
Robert: “To me, AI is very much like recorded music all over again. It is not live music played with human hands. It is a replica. It’s its own thing and turned into a new art form. It is like the ghost of recorded music, making new music itself.”
Josh: “Stop trying to normalize fake brains!”
I do not need to be in the same room with Robert to hear us laughing. “James Today,” the debut single by Micropanthers is out now.