When Machines Sing: Can AI Imitate the Human Soul?

By Kavisha Pillay

Humanity has produced many great wonders – architecture that defies centuries, literature that continues to illuminate our inner worlds, philosophies that challenge how we live, and art that holds a mirror to time itself. But perhaps nothing captures our collective spirit as universally, as music. 

Music is one of humanity’s oldest companions. Long before we built parliaments or penned constitutions, our ancestors sat around fires, creating rhythm from stones, stretching animal hide into drums, or coaxing melodies from bone flutes. Those early sounds weren't entertainment in the modern sense – they were communication, ritual, prayer, and identity. In every culture, music has carried memory and meaning.

Across history, music has healed, confronted, and transformed. That is why Bob Marley summoned us to unite under One Love, while John Lennon dared us to Imagine peace beyond borders (and religions, too). Miriam Makeba’s voice soared above the cruelty of apartheid, becoming a beacon of freedom, dignity, and hope, while Carlos Santana’s guitar continues to bridge cultures and continents, bringing people together through shared rhythm. Back home in South Africa, local favourites like Amapiano and Gqom are fulfilling Madiba’s rainbow dream, serving as multicultural unifiers at the weekend grooves.  

That’s the extraordinary power of music. It flows from authenticity and raw, lived experiences like love, joy, anger, defiance, sadness, grace, and spirituality.   

But, what becomes of this magic when the music that we adore – rich with stories of love, loss, courage, and vulnerability – ceases to come from a human heart, but is instead conjured up from the sterile algorithms of artificial intelligence?  

Rage, against the machine. 

In July 2025, a band called The Velvet Sundown captured audiences on Spotify, racking up over one million streams with indie/folk style songs that had aching lyrics, poetic melodies, and harmonies filled with the promise of connection. 

But The Velvet Sundown’s did not exist – well, at least, not in the physical or human sense. 

The “band” itself was not locked in a garage for months, jamming and learning chords, and no writer was scribbling lyrics on scraps of paper as a result of heartbreak, anger at the state of society, or wonder about the beauty of existence.

The Velvet Sundown was a fabrication. The songs were entirely AI-generated, with data-sculpted voices, and melodies stitched together from probability models trained on decades of indie/folk music.

It’s one thing when AI helps a producer fine-tune beats or assists with arrangements. But when audiences are swept away by the emotions of a band that does not exist, we must pause. 

This phenomenon raises profound questions about the future of creativity, authenticity, and ownership in an era where AI can mimic the human soul, with unsettling precision.  It forces us to confront what authenticity means when a machine can fake it? Who owns the cultural stories being recycled into data? And, most critically, what is lost when voices are simulated rather than lived?

The rise of AI-generated music mirrors a broader crisis in the digital era: deepfakes. These technologies produce hyper-realistic replicas of faces, voices, and even entire personas, often without consent.

Think of the ache in Unchained Melody, or the grit in Kendrick Lamar’s bars, or the meditative improvisations of Abdullah Ibrahim’s piano, or the stirring resonance of Anoushka Shankar’s sitar – these are not mere sounds. They are living testaments of history, memory, and truth. To transpose these into algorithmic cut-and-paste operations risks hollowing out their meaning, reducing profound human testimony into empty aesthetics.

The threat extends beyond music. AI’s capacity to generate synthetic voices and faces threatens journalism, education, and even democracy. Deepfakes already pose risks of identity theft and political manipulation. Now, when entire artists can be fabricated and popularised, we stand at the edge of a cultural deepfake: a reality where our most trusted forms of expression, art and music, can be co-opted by machines.

The difference with music is emotional scale. Songs shape identity far more intimately than news articles ever will. They soundtrack weddings, revolutions, funerals, and heartbreaks. When these intimate anchors cease to belong to real people, what happens to the way that we trust art?

Imagine future generations having their formative coming-of-age moments tied to songs created by code, songs with no human story, and no ancestral or cultural thread. What kind of cultural memory do we build in that scenario?

Safeguarding humanity in art 

The problem is not technology itself. Tools have always extended human artistry. The piano was once a new technology, as was the synthesizer. But, this is not like when Bob Dylan went “electric” in 1965. 

If we allow AI to transform music into an ocean of synthetic voices, we risk silencing the singular, unique resonance of artists themselves. When art ceases to be a grappling of our humanity and an expression of our lives in all of its messy and beautiful forms, we lose not just authenticity but empathy, which is the very core of why art matters.

And so, to preserve music as a sacred space for human connection, we must; 

• Demand transparency and insist that streaming platforms and other providers clearly label AI-generated music. Audiences have the right to know whether they are being moved by human experience or machine mimicry; and,

• Advocate that artists have legal agency over whether their voice, style, or catalogue can be used to train AI models. Consent must be non-negotiable. Where consent is provided, compensation must be fair and just. 

In the end, music is not just sound. It is testimony. It is evidence that we have felt, endured, and loved. When machines replicate it without the origin of that feeling, it becomes the appearance of humanity without humanity itself. 

Music is our heartbeat, our memory, and our prophecy. Let us not surrender it lightly to the algorithms.