BCCN3

View Original

AI: The End of Posthumous or a New Age of Music?

It’s 1722 and Johann Sebastian Bach has just published The Well-Tempered Clavier, an innovative new collection of preludes and fugues endorsing the Equal-Temperament Tuning System, a groundbreaking development in piano construction that would change Western music forever. 

Two and a half centuries later in 1976, Stevie Wonder would use motifs from Bach’s Prelude No. 2 in C Minor while experimenting with Yamaha’s latest synthesizer the GX-1 in Pastime Paradise during recording sessions from his hit album Songs in the Key of Life. It was the first time a synthesizer had been used to mimic the sounds of a string section. 

Despite not being the album’s most popular track next to the album’s hit singles I Wish and Sir Duke, the song would take on new fame when it was used by Coolio in his #1 hit Gangsta’s Paradise, taking a tougher perspective on life in the inner-city. 

Years later, the track has become the latest meme from generative AI in a video posted by Lace Editing depicting a rendition of Gangsta’s Paradise being performed by Frank Sinatra, showing how deep learning is creating a new dimension of art and the meaning of post-houmous content.

Bach’s Endorsement of Equal-Temperament

Johann Sebastian Bach, a German composer born in 1685, lived in a time when pianos and other keyboard instruments were tuned using meantone, a crafting method that tuned instruments to their most perfect pitch. 

While this made certain notes like C sound as best as possible, it caused other notes to lose their tuning the further away they were from the root note. Consequently, songs from certain keys were simply unplayable on most pianos. 

In order to solve this problem, a new crafting technique called Equal-Temperament was developed which sacrificed tuning quality for a more mathematical design that could play every note in a chromatic scale at ‘near-perfect’ pitch, opening the doors to vast amounts of musical exploration now that all keys could be represented accurately by keyboard instruments.

Bach recognized this innovation in technology and published The Well-Tempered Clavier, a collection of preludes and fugues in all 24 major and minor keys, to showcase the range of possibilities created by Equal-Temperament, cementing his legacy as one of history’s greatest composers. 

Less than a decade after Bach’s death, Mozart - a musical prodigy - was born in 1756 and took classical music to new heights, applying Bach’s vision of key modulation by incorporating complex key changes and harmonies that would be impossible to play on purely intonated instruments. 

15 years later, another musical genius - Beethoven - was born in Germany, utilizing the same Well-Tempered system popularized by Bach, highlighting its musical limitlessness with his 5th Symphony (which is also in C minor) and Ode to Joy.

The GX-1 ‘Dream Machine’ Creates a New Sound

By the 2nd half of the 20th century, popular music had taken on an entirely new sound following the advent of the radio and recording technology. Suddenly, music could be heard from anywhere, with musicians playing in everyone’s homes through speakers and vinyl records. 

Innovation hadn’t stopped either. Pianos and other instruments like the guitar were being introduced to a new electrical world, allowing them to plug into amplifiers and other manipulators to alter their sounds in unique ways. 

In 1973, Herbie Hancock became a major influence, pioneering the use of synthesizers in his album Head Hunters, a defining intersection between jazz and funk. The success of the album helped introduce electronic sounds to a wider audience by fuzing with recognizable R&B structures. 

A few years later in 1976, Stevie Wonder would explore this new realm of music too after receiving an exclusive GX-1 synthesizer from Yamaha. The instrument was so complex for the time that only a few had been produced with models going to other popular musical groups like Led Zeppelin. 

The GX-1 was able to manipulate sound, mimicking other instruments. Wonder was fascinated by the new possibilities and began experimenting with using the synthesizer to recreate a full-string section, shaping Bach’s Prelude No. 2 in C Minor into Pastime Paradise. 

Gary Olazabal, the audio engineer on Songs in the Key of Life said “[Wonder] and Greg Phillinganes were both just fooling around, playing a lot of classical chord progressions, and the melody of 'Pastime Paradise' was based on the string sounds that he was getting. In fact, we recorded the Dream Machine as if it was an actual string section, with Stevie playing the cello part, the violas and so on.”

Coolio Introduces a New Culture

In 1995, Pastime Paradise found new popularity after Coolio’s reimagined version Gangsta’s Paradise hit the airwaves and remained at #1 on the Billboard Charts in the US, the UK, and a dozen other countries. 

The song was an immediate success for Coolio, propelling his career and bringing Bach’s 270-year-old example of key modulation to a new, modern audience entertained by the growing genre of rap and hip-hop. 

Gangsta’s Paradise has since become an iconic hip-hop classic with the opening lines being some of the most recognizable in rap. Stevie Wonder even enjoyed the song, performing it with Coolio onstage at the 1995 Billboard Music Awards, adding the lyrics “Ain’t no gangstas living in paradise / Ain’t no racist living in paradise.”

Upon Coolio’s death in 2022, the hit single shot back up onto the charts, reaching as high as #5 on the UK Official Singles Sales Chart Top 100. The music video, which won Best Rap Video at the MTV Video Music Awards in 1996, has over a billion views on Youtube. 

The Fifth Element - AI

All of that is to say, the new video of Frank Sinatra singing Gangsta’s Paradise is pretty interesting. It’s a culmination of technological advances over the past 300 years to create a meme laughing at the juxtaposition between Frank Sinatra and gang culture, but it could use some improvement. 

The video is an example of AI using Natural Language Processing to recreate Frank Sinatra’s voice. The algorithm is fed a set of data containing Sinatra’s recordings to train itself to sound like the crooner.  

The recreation sounds good overall, capturing the quality of Sinatra’s voice and matching it to unfamiliar lyrics. However, there is a small hiccup in the track when the song reaches a climax at the 1:39 mark during the lines “I'm 23 now but will I live to see 24? / The way things is going I don't know.”

When you listen to the recording, it is obvious that the AI has miscalculated the final notes of the progression, trying to rhyme ‘24’ with ‘know.’ This can be contributed to the fact that the underlying track itself is not entirely AI-generated. 

Robyn Adele Anderson originally performed a jazz version of Gangsta’s Paradise in video that has garnered over 20 million views on Youtube since it was posted in 2015. The sound of the rhythm section is nearly identical to what was recreated in the Sinatra version similar to another AI-generated video of Elvis Presley singing Baby Got Back to the tune of Don’t Be Cruel

In Anderson’s version of Gangsta’s Paradise, she finishes the line “The way things is going I don't know” by drawing out the ‘know’ (unlike Coolio’s way of saying it quickly and without emphasis) which confuses the AI. 

The algorithm attempts to match the intended audio output by assuming that each line rhymes perfectly, but Coolio’s lyrics don’t follow a typical rhyming pattern with some lines ending with pairings like “facing / street” and “mind / eye.” This irregular pattern interrupts the AI, causing it to falsely predict the lyrics and pitch during a pivotal transition into the chorus.

Although we possess the ability to recreate posthumous voices and saran-wrap them around separate recordings, it’s akin to hearing instead of listening. All 4 versions (by Bach, Wonder, Coolio, and Anderson) contain nuances and intricacies that are difficult for machine learning to predict because of all the small variations that make each musician unique. 

However, what if predicting the correct pitch and rhymes for an AI voice to mimic is only the smaller picture? Songwriting and Musical performance are holistic experiences that draw from a pool of emotions and techniques; to recreate that artistic spark would be groundbreaking. 

A New Dimension of Posthumous Content 

We’re often told stories of “what could have been” when reminiscing about artists like Otis Redding who died days after recording (Sittin’ on) the Dock of the Bay, the biggest song of his career and a marked shift in his creative process, or Jim Morrison, who died weeks before the release of L.A. Woman, a career revitalizing album for The Doors. 

With generative AI, one has to wonder: can we go beyond “copying and pasting” the voices of past artists and recreate their artistic minds to discover “new” music by them? Paul McCartney seems to think so after using AI to finish the ‘last’ Beatles song with old recordings made by John Lennon. 

Despite the obvious legal and ethical issues involved with trying to “recreate the voices of the dead,” if we’ve learned anything from Johann Sebastian Bach and Prelude No. 2 in C Minor, it’s that technical innovation creates seismic impacts on our ability to create art and this next step using AI could rival Equal-Temperament (which led to the sudden emergence of two of music’s most brilliant minds - Mozart and Beethoven) by creating a new dimension of utility for all known musical recordings.