AI Songwriting: “We are about to witness the next generation of songwriters who will inevitably use AI as a creative tool”

As a member of Hermitage Green, I have been writing and releasing songs for over a decade. I also lecture at Trinity Laban Conservatoire of Music & Dance and the Institute of Contemporary Music Performance in London. In virtually every area of ​​the music industry, the prospect of computer-generated songs has become the prevailing reality. Some see AI as an evolution of existing technologies, while others foresee a dystopia of generic, soulless art and unemployed artists.

To find out how creative these systems are and what challenges they might pose for artists, I spent the first few months of 2024 researching AI-based songwriting software.

The method

I have made a plan:

First, I would create a portfolio of songs using AI. Then I would submit them to a first-year songwriting assessment at the Institute of Contemporary Music Performance, where they would be reviewed by an assessment panel who would award a grade and feedback based on the melodic, harmonic and lyrical components of the pieces, as well as other factors such as stylistic context and considerations of technique and process. This is how music is usually assessed, at the third level, and it is probably the fairest way to approach the somewhat arbitrary task of assessing a song.

I spent weeks experimenting with different software like Loudly, Chat GPT, Ryter, Mubert, Melobytes, etc. Most of these platforms are free to use and easy to navigate.

However, it was difficult to generate chords, lyrics and melody from a single prompt, 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” (no wonder Nick Cave called AI songs “the apocalypse”).

To get around these limitations, I created backing tracks using programs like Loudly and Mubert, and used Chat GPT or Ryter for the lyrics before creating my own melody to make the words fit the music. This could be considered a 'contaminating element' as it made me a creative contributor, but at the time it was the most effective way to create a fully fleshed out track (it should be noted that systems have become much more advanced since then).

How does this technology work?

AI songwriting algorithms mimic neural networks. When a user gives a prompt, the AI ​​searches its ever-growing memory for references and then combines elements from stored music of that genre to create a new “original” piece.

Everything we write is also shaped by the music we've heard and been inspired by throughout our lives. What makes a song unique and original is personality and idiosyncrasy – a new, authentic element that we recognize, remember and draw on. In this respect, I found the AI ​​songwriting programs disappointing. At the heart of every good song is a lived, emotional experience. I felt that the lyrics and music generated by the AI ​​systems were generic and cliched.

There was also a strong tendency towards precise phonetic rhymes. Is that bad? Well, in a not so benevolent survey, NME Readers once chose the lyrics of Des'ree “I don't want to see a ghost, that's the sight I fear most, I'd rather have a piece of toast”, as the worst line ever written. Rhyming for the sake of rhyming seems mechanical, even when humans do it. For any decent songwriter, meaning takes precedence over the need to rhyme. It doesn't matter if the rhyme scheme isn't perfect, and sometimes the lack of an expected rhyme can create a dramatic effect. AI systems have a long way to go if humans are to respond to emotional songs in the future.

Another problem I encountered was censorship. I found the Ryter program better than Chat GPT in this regard, but it was still prudish. For example, it refused to write song lyrics if the prompts contained words like wanker And penisor names like Muammar Gaddafi. Not very rock 'n' roll if you ask me.

The results

After a few weeks, I had three AI-generated tracks. One of my talented students lent me her vocals, bringing the melodies to life and reminding me of the immediate impact of a human voice. After the songs were finished, I used Chat GPT to create a detailed essay on how they were written, complete with a very well-cited bibliography, all based on nothing. This took about 20 seconds.

I submitted the project to ICMP for our end-of-year songwriting assessments, where it was included in a group of 70 other student portfolios. To avoid bias, the assessment panel was not informed of the use of AI – to them, it was just a normal submission.

A few weeks later, the results came back. The songs received 72% (an honor) and the Chat GPT essay, in which I discussed my “writing process,” received 68% (a second). Most surprising of all, the project had surprised no one.

The reviewers liked the melodies (which I wrote) and the consistent use of perfect rhymes didn't seem to bother them too much. Some reviewers had issues with the structure, as all three songs contained an identical pattern of verse, chorus, verse, chorus and ending, with no bridge sections, which is common with AI-generated tracks.

Although I don’t know for sure, it could be that the algorithm is trying to replicate the fact that many number one songs from the 21^st Century typically have neither a bridge nor a middle 8. The late Ralph Murphy speculated that this trend reflects our dwindling attention spans.

What now?

So what about this dystopian future, the so-called AI “apocalypse” that Nick Cave talks about?

Will sophisticated algorithms produce songs for every occasion? Will we flock to music festivals and venues with our robot friends to see our favorite artists before heading home to put our little robot-human hybrid babies to bed?

It's all happening so fast that it feels slow, but I don't see things in such a dire light. I certainly don't see AI songwriting as a death knell for artists.

At the end of the 19th century, mathematician Ada Lovelace predicted that one day we would use computers to compose songs. A century later, David Bowie used a computer-controlled random number generator to get ideas for song lyrics. AI is the next step in this development. Songwriting processes have been accelerated by machines for decades, and listeners have always been more interested in the people who operate them than in the devices themselves.

Songs are strange little vessels of emotion. They are entities; more than the sum of their parts. Much more than just a pile of notes and words.

Take Eric Clapton's “Tears in Heaven,” for example. It's an objectively emotional piece of music, both lyrically and harmonically, and when we learn that it's about the death of his four-year-old son, it takes on a whole new level of heartbreaking context.

Humans are tribal creatures. We connect with one another through stories, lived experiences, and shared emotions. Our innate need for connection cannot be replaced by algorithms that mimic human creativity. Instead, I believe we will soon witness the next generation of songwriters who will inevitably embrace AI as a creative tool.

Even though the songs I wrote were soulless and, let's face it, pretty crappy, I was struck by a strong sense of collaboration between the AI ​​systems and I. It was like we were co-writers, and I wasn't expecting that.

Good art often pushes boundaries. It challenges convention by breaking the rules. The songwriters of the future will lean on AI as an incentive to create authentic and original art, to express themselves and connect with audiences in new ways. It will be interesting to see how these authentic-artificial collaborations subvert the norm.