The topic of AI in music could be an endless conversation without a definitive conclusion.
 

What we know so far is that AI uses algorithms to analyse already existing music, composed by human beings, and generates other material based on that.

One could argue it’s a similar approach to the idea of a collective consciousness that humans tap into for creative endeavors, but I believe that AI lacks the emotional and unconventional behaviour that humans have because of our unique or subjective views of reality influenced by emotions.

These 10 pieces I’ve selected that I believe AI couldn’t reproduce range from classical and jazz to rock and electronic music.

They all have some form of influence from previously composed works or unique human experiences that can spark creativity and lead to beautiful works of art, or just pure exploration for the sake of it.
 

At the time of their release, they were considered to be genre-breaking and are still relevant and loved to this day.

Rachmaninoff - Rhapsody on a Theme of Paganini.

The famous 18th Variation, one of the most moving pieces of classical music, was inspired by Paganini’s 24th violin caprice; examples of both pieces are below.

​Explanation on how the theme was used

Miles Davis - Sketches of Spain.

Great artists usually seek inspiration in different styles and cultures, and Miles Davis was no exception. In the beautiful album entitled Sketches of Spain composed together with Gil Evans, he used the Concerto de Aranjuez for guitar as his main source of inspiration, perfectly bridging the gap between classical orchestration and arrangement and jazz improvisation and unpredictability.
 

Queen – Bohemian Rhapsody.

Arguably the most iconic rock song of all time; at the time of its release, nobody understood its meaning, but they almost all agreed it was a masterpiece, besides the label, who didn’t want to release it. To me it’s a perfect example of unconventional human thinking, blending multiple genres, introducing an opera section in a rock song, breaking all rules of song structure (most songs have choruses, whereas this one doesn’t), and underlining the idea of unique human creativity.

Stevie Wonder - Overjoyed.

This one is, to me, another great example of how the unique human mind of a musical genius can introduce complex harmonic structures in a seamless way and include nature-recorded sounds like the raindrop sample used instead of a clap.

Pink Floyd - On the Run.

Even before the seminal Autobahn album by Kraftwerk, Pink Floyd used programmed synthesizers on The Dark Side of the Moon. We might consider this normal today, but at the time of its release in 1973, it was as uncommon as you could imagine, yet somehow it still feels very human and unique.

Aphex Twin - ∆Mᵢ⁻¹=−α Σ Dᵢ[η][ Σ Fjᵢ[η−1]+Fextᵢ [η⁻¹]] (also known as Formula)

This is a track most would think was written by a machine rather than a human, but it’s one of the best examples of human experimentation and forward thinking for its time. Released in 1999 on the groundbreaking Windowlicker album, it uses a spectrogram to generate sounds on top of a programmed drum groove, and he famously used the spectral analysis of his face on the outro of the track.
 

Boards of Canada - Aquarius.

One of my personal favorite downtempo tracks was released in 1998, when samples were a big part of electronic music production (also a form of collective consciousness, one could argue) – the unique use of sampling ranging from Sesame Street (a kids' educational TV program) to country and rock.

Susumu Yokota - Blue Sky and Yellow Sunflower.

Another good example of humans reinterpreting genres, taking classical music themes and rearranging them into new formats.

Floating Points - Silhouettes (I, II & III)

Another great example of human exploration and complexity, particularly coming from an electronic music producer, in this case Sam Shepherd (Floating Points), which explored genres ranging from disco and techno to jazz and progressive rock.

Four Tet - Skater

To wrap up the idea of tracks that AI couldn’t reproduce certain songs, I would probably argue that what AI can’t reproduce is human emotion, because it uses algorithms based on logic, whereas humans have emotions. In this song, Kieran (Four Tet) declared in an interview that he produced the song after recording his daughter ice skating on a frozen lake during the pandemic, a powerful emotion that sparked the idea behind a work of art.