The Hermitage Green singer discusses his distinctive AI songwriting experiment, and why the know-how changing into the norm in the music trade mightn't essentially be a foul factor.
As a member of Hermitage Green, I’ve been writing and releasing songs for properly over a decade. I additionally lecture at Trinity Laban Conservatoire of Music & Dance, and at the Institute of Contemporary Music Performance in London. In nearly all avenues of work in the music trade, the prospect of computer-generated songs has turn into a prevalent actuality. Some see AI as an evolution of present applied sciences, whereas others envisage a dystopia of generic, soulless artwork and unemployed artists.
To uncover how artistic these programs are and the challenges they might pose to artists, I spent the first few months of 2024 researching AI songwriting software program.
The Method
I set out a plan:
Firstly, I might use AI to generate a portfolio of songs. I might then submit them to a primary yr songwriting evaluation at the Institute of Contemporary Music Performance to be scrutinised by an evaluation panel, who would difficulty a grade and suggestions based mostly on the tracks' melodic, harmonic, and lyrical elements, as properly as different elements like stylistic context and reflections on approach and course of. This is how music is usually assessed at third stage, and might be the fairest option to go about the considerably arbitrary activity of grading a track.
I spent weeks experimenting with completely different software program like Loudly, Chat GPT, Ryter, Mubert, Melobytes, and so forth. Most of these platforms are free to make use of and straightforward to navigate.
That stated, it was troublesome to generate chords, lyrics and melody from a single immediate e.g., “write me a song that sounds like David Bowie, in the lyrical style of the Wolfe Tones, but sing it like Barbara Streisand” (it’s no marvel Nick Cave referred to AI songs as “the apocalypse”).
To work round these limitations, I generated backing tracks on programmes like Loudly and Mubert and used Chat GPT or Ryter for lyrics, earlier than fashioning my very own melody to make the phrases match the music. This could possibly be thought-about a “contaminating element”, as it made me a artistic contributor, however at the time it was the only manner of producing a totally fashioned observe (it needs to be famous that the programs have since turn into much more superior).
How Does This Technology Work?
AI songwriting algorithms mimic neural networks. When a person offers a immediate, the AI searches its ever-expanding reminiscence for references, then combines parts from saved music in that style to create a brand new “original” piece.
Everything we write can be colored by music we’ve heard and been impressed by all through our lives. What makes a track distinctive and unique is persona and idiosyncrasy – a brand new, genuine component that we recognise, bear in mind, and return to. In this regard, I discovered the AI songwriting programmes to be underwhelming. At the coronary heart of each good track is a lived, emotional expertise. I felt the lyrics and music generated by the AI programs had been generic and cliched.
There was additionally a robust tendency towards exact phonetic rhyming. Is that unhealthy? Well, in a not-so-charitable ballot, NME readers as soon as voted the Des’ree lyric “I don’t want to see a ghost, it’s the sight that I fear most, I’d rather have a piece of toast,” as the worst line ever written. Rhyming simply for the sake of it comes off as mechanical, even when human beings do it. For any first rate songwriter, that means takes precedent over the want for rhyming. It doesn’t matter if the rhyming scheme isn’t excellent and typically, the lack of an anticipated rhyme can create a dramatic impact. AI programs have a protracted option to go if emotionally-driven songs are what folks reply to in the future.
Another difficulty I got here throughout was censorship. I discovered the Ryter programme to be higher than Chat GPT on this respect, nevertheless it was prudish nonetheless. For instance, it refused to write down lyrics if the prompts contained phrases like wanker and penis, or names like as Muammar Gaddafi. Not very rock ‘n' roll should you ask me.
The Results
After a pair of weeks I had three AI-generated tracks. One of my gifted college students lent me her vocals, bringing the tunes to life and reminding me of the instant affect of a human voice. Once the songs had been completed, I used Chat GPT to generate an intensive essay on how they had been written, full with a really well-cited bibliography, all based mostly on nothing. This took about 20 seconds.
I submitted the mission for our end-of-year songwriting assessments at ICMP, the place they had been included in a batch of 70 different scholar portfolios. To keep away from bias, the evaluation panel weren't knowledgeable of the use of AI – as far as they had been involved, it was only a regular submission.
A pair of weeks later, the outcomes got here again. The songs acquired 72% (a distinction) and the Chat GPT essay discussing my “writing process” acquired 68% (a second). Most surprisingly, the mission hadn’t raised a single eyebrow.
The markers favored the melodies (which I wrote) and the relentless use of excellent rhyme didn’t appear to trouble them an excessive amount of. Some marks had been misplaced on construction, as all three songs contained an an identical sample of verse, refrain, verse, refrain, finish, with no bridge sections, a standard prevalence in AI-generated tracks.
While I do not know for sure, it could possibly be that the algorithm is making an attempt to copy the indisputable fact that many #1 songs from the 21st century have a tendency to not have a bridge or center 8. The late Ralph Murphy speculated that this development displays our dwindling consideration spans.
What Now?
So, what of this dystopian future; the so known as AI “apocalypse” which Nick Cave speaks of?
Are superior algorithms going to be churning out songs for any events? Will we be flocking to music festivals and venues with our robotic pals to see our favorite robotic artists, earlier than going house to place our little robot-human-hybrid infants to mattress?
It’s all taking place so quick that it makes you're feeling gradual, however I don’t see issues in such a dark gentle. I definitely don’t see AI songwriting as a demise knell for the artists.
In the late 1800’s, mathematician Ada Lovelace predicted that there'd be a day after we would use computer systems to compose songs. A century later, David Bowie was using a computer-driven random phrase generator to encourage lyrical concepts. AI is the subsequent step in that evolution. Songwriting processes have been accelerated by machines for many years, and the listener has all the time been extra considering the people working them than the units themselves.
Songs are unusual little vessels of emotion. They're gestalt; higher than the sum of their components. Far greater than only a bunch of notes and phrases.
Take ‘Tears in Heaven' by Eric Clapton. It's an objectively heartfelt piece of music; lyrically and harmonically, and after we be taught that it’s written about the demise of his four-year-old son, it takes on a complete different layer of coronary heart breaking context.
Human beings are tribal. We join with one another by way of tales, lived experiences and shared feelings. Our inherent want for connection can't be changed by algorithms mimicing human creativity. Instead, I consider we're about to witness the subsequent era of songwriters who will inevitably use AI as a artistic software.
Although the songs that I generated had been soulless and, let's face it, fairly shite, I used to be struck however a robust sense of collaboration between the AI programs and myself. It was like we had been had been co-writers, and I didn’t anticipate that.
Good artwork is usually transgressive. It challenges conventions by breaking the guidelines. The songwriters of the future will lean on AI as a stimulus to generate genuine and unique artwork, to specific themselves and to attach with audiences in new methods. It will probably be attention-grabbing to see how these authentic-artificial collaborations subvert the norm.
