Good music can brighten our day and lift our spirits. Great music has a power that borders on the supernatural. 

The writer Hunter S. Thompson captures this potency when he said, “On some nights I still believe that a car with the gas needle on empty can run about fifty more miles if you have the right music very loud on the radio.”

Put simply, great music is magic. It's because of this feeling of otherworldly connection, that the incipient wave of Artificial Intelligence feels so personal. As the headlines constantly remind us, Generative AI is here, and it’s poised to change everything itself - music very much included. 

Artificial Intelligence has long found its way into musical production, enabling artists to better execute on their creative vision. In recent years though, accelerated by the advent of Generative AI, technology has steadily crept into the creative process itself. Musical composition - a process that feels so distinctly human that it would forever be protected from the slow march of algorithmic progress, now finds itself with the enemy at the gates.

Many see these innovations as driving music into a very dark place. As the thinking goes, it’ll be a sad day when music doesn’t come from this vaunted place called human creativity, where one bears their soul and pours it into their art, but instead - from an algorithm. As Living Colour guitarist Vernon Reid told Rolling Stone, “Music, as made by humans driven by extraordinary circumstances … those who have suffered and struggled to advance their craft, will now have to contend with the wholesale automation of the very dear-bought art they have fought to achieve”. 

The concerns are real. But what would actually happen if this future were to transpire? What would happen if, against the intuitions of music lovers everywhere, people actually enjoy this new wave of music? And what if this AI-music isn’t merely catchy, and poppy and saccharine, and isn’t merely the “good” music that lifts our mood. What if AI music is truly great? 

What Happens When Artificial Intelligence Creates Great Music?

Consider the following thesis: At some point in your lifetime, truly great music will be created and produced almost entirely by Artificial Intelligence, with little to no human involvement. Not just “good” music, but great music - the quality that could power a car on empty another 50 miles. 

If you doubt this, you’re far from alone. But consider the following thought experiment: Bring to mind a song that you hold dear; a piece of music that you would consider “truly great”. 

Now imagine a new song popping up that you’ve never heard before. It’s different from your original song, but it produces in you the same personal connection, and conjures up a similarly rich tapestry of emotion. Only this song, in every dimension that matters, is better: a closer connection, deeper enjoyment, and a richer emotional texture. 

It’s the best sound to ever enter your ears. 

And then imagine being told that it wasn’t created by a human. It was AI-generated. How would you reconcile this? What would you think about the song? 

It would be hard to deny that the song produced this special emotional response, since you’ve already had this experience. Denying it would simply mean lying. If you’re objecting to this scenario then, you would need to reject that this kind of experience would ever be possible. You’d have to say that the experience of truly enjoying AI-generated music is impossible. This would entail at least one of two beliefs:

1) Only a human being can make music which is truly great 

2) Music can only be great because it is made by a human 

In other words, either AI won’t make great music, or AI can’t make great music. Let’s examine each of these claims in turn. 

Defining The Market’s Response to Great, Generative-AI Music

Let’s start with the first objection: Only a human being could make music which is truly great. This is a solid argument but it’s ultimately an empirical claim, which rests on the AI-music’s reception. It’s the audience that ultimately determines whether a piece of music is great. 

How do we adequately define the market’s response? How could we know if a piece of AI-music has achieved “greatness”? 

Of course, each person is different and musical taste differs widely. Any given person can find any given piece of music to be “truly great”. Of the 8 billion people in the world, any given song will be enjoyed by someone out there. How many people need to consider a song to be great? And how great do they need to feel that it is? If we’re getting into the nitty gritty, defining the market’s response is challenging to fully operationalize. 

Clearly, it’s an unachievable bar for an artist to be considered universally great. No artist - not even Taylor Swift, can stake that claim. 

For the sake of argument, think of the level of greatness that a random, moderately successful musician has achieved. It’s a level of musical greatness which goes beyond the individual and the idiosyncratic, and has crossed a much more robust cultural threshold. It’s music that has received accolades, has a notable fanbase, but wouldn’t be considered an all-time great. If you’re not a fan, you could at least hold a conversation with someone who is.

For the sake of argument, let’s call this “moderate popularity” and say that this is level that Generative-AI musical creation needs to reach. With this definition in mind, AI has been involved in musical composition for nearly 75 years. Has the market’s response reached this point? 

A Quick History of AI’s Role in Musical Composition

While excitement and concern around these innovations has increased in recent years, AI-created music has been around for a long time. Alan Turing, one of the pioneers of machine learning, was one of the first to apply artificial intelligence to music back in the 1950s. In 1957, two professors utilized a computer algorithm to compose Illiac Suite (later retitled String Quartet No. 4) at The University of Illinois at Urbana–Champaign. The piece, which consists of four movements, is considered to be the first musical score to be composed by artificial intelligence. 

Since these early origins, AI has continued to play an important role in music, though much more with production than with musical composition itself. In the minds of most music lovers, this is a key distinction: technology as a tool to express human creativity is more or less fine. Technology in lieu of human creativity is the line in the sand. 

When it comes to crossing this line, AI musical composition made slow and unimpressive progress through most of the 20th century. One of the first breakthroughs took place in 2016, when Warner Music Group signed the first algorithm to a record real. The algorithmic code, engineered by an audio start up called Endel, was contracted to create 15-20 albums that year. 

How did this new signing do? Meh. Beyond the excitement of a few tech utopians excited by the historical nature of the event, the music itself was underwhelming. A few years later, Warner tried again, this time signing an AI-artist in the form of “Noonori” - a rendered, twelve-year old girl with punk blue hair. Both artist and music were not a hit with the general public. So far, nothing approaching moderately popular music.

Generative AI Music in 2024: From Heart on My Sleeve to Suno AI

As of 2023, musical composition felt sufficiently safe from the onset of AI. Even amidst the early excitement of ChatGPT, which was launched to the public in late 2022, music - especially compared to mediums like text, images, and movies, was largely unaffected. 

This all changed in September 2023, when a song by the name of Heart on My Sleeve was uploaded to YouTube. If you heard it for the first time without knowing otherwise, you would think it was a genuine collaboration between Canadian pop-stars Drake and The Weeknd. The pulsating piano melody and clever lyricism (“I came in with my ex like Selena to flex / Bumpin' Justin Bieber the fever ain't left”) would not be out of place in a nightclub.

In fact, if it was played back to back with a new song that Drake actually created, fans wouldn’t be able to tell you which one was Drake and which one was AI - and they may actually prefer the former. The song was even submitted to the Grammys for best rap song and song of the year (though later ruled ineligible). But of course, this wasn’t a collaboration between Drake and The Weeknd. Heart on My Sleeve was the creation of an internet creator known as Ghostwriter, who produced the song from scratch using Generative AI. 

The song quickly made the rounds, going viral on social media. Part of the excitement was the sheer jaw-dropping novelty. After all, this was the “early days” of generative AI, and we were just coming to grips with its uncanny capabilities. The song’s AI vocals were indistinguishable from the voices of the real artists, giving the track’s poppy vibe an eerie, “post-human” feel which listeners had never experienced before. 

But beyond the novelty there was, above all, an undeniable enjoyment. People liked it. An important threshold in the bastian of human creativity had been breached.

Does Heart on My Sleeve trespass into great AI music? Does its existence prove that AI actually can create truly great music? And how does this stand up to more recent innovations in Generative AI music, such as Suno AI?

Photo by Andrew Itaga via UnSplash

This is part 1 of a multi-part series on the Psychology of Generative AI Music. Part 2 addresses the above questions, while also exploring the second objection to this thought experiment: Great music is great because it is created by humans.

References for Generative AI Music and Human Creativity

Alexander, A. (2024). “Heart on My Sleeve”: An AI-Created Hit Song Mimicking Drake and The Weeknd Goes Viral. SAGE Publications: SAGE Business Cases Originals.

Hiatt, B. (March, 2024) A ChatGPT for Music Is Here. Inside Suno, the Startup Changing Everything, Rolling Stone

Verma, S (2021). "Artificial Intelligence and Music: History and the Future Perceptive". International Journal of Applied Research. 7 (2): 272–275