Artificial intelligence looms over the creative industries, but The Cotton Modules show how the tech unlocks new opportunities for those willing to tinker. The pair sit down with Clovis McEvoy to discuss music technology, ethics, and creative sparks that come from working with an AI vocalist — and it's so much more than imitation.
“AI is just like an old analogue synthesiser – it has unique sonic properties,” says Robin Sloan, one half of music’s AI auteurs, The Cotton Modules. It might seem odd to use a word like ‘old’ to describe Jukebox, OpenAI’s generative audio tool, but Robin has been at this longer than most.
Formed with long-time collaborator Jesse Solomon Clark, The Cotton Modules’ debut album, Shadow Planet, was a landmark release in autumn 2021, demonstrating the musical potential of generative AI long before the technology exploded into the mainstream late last year. Their recently-released second album, The Greatest Remaining Hits, takes that foundational work and blasts it, quite literally, into the stratosphere.
The album tells the tale of an AI spacecraft entrusted with the sum total of humanity’s cultural output – its art, music, literature, and more. However, with its memory banks damaged and its mission imperilled, the algorithm sets about trying to recreate the human canon using incomplete and corrupted data. The results are what we hear across the album’s 16 tracks.
A highly conceptual piece of work, sure, but if there was ever a time to put AI at the heart of the narrative, it’s right now. “Artificial intelligence had to be part of the story,” Robin explains. “Its development is so rich and provocative that it seemed a shame not to address it directly.”
“The album’s concept involves piecing together the entire catalogue of human recorded music,” Jesse adds. “An AI glueing together these broken files into something new for future humanity.” It’s a narrative intimately tied to the very real challenges the pair faced in the process of writing the album.
Serving as the band’s lyricist, Robin uses his own customised version of Jukebox to ‘play’ the algorithm through what the pair describe as “audio seeds”. These text prompts help nudge the output in a certain musical direction – but even so, the process requires Robin to painstakingly comb through the generated audio in order to find moments of musical magic. “The process is slow,” Robin admits. “I proceed in 10-second segments, coaxing it along and rejecting any garbled misfires, of which there are many.”
These short vocal ‘performances’ are then passed to Jesse, who sets about weaving them into his original compositions. “I appreciate that technology is often viewed as something to make work easier, faster or more efficient,” he says. “But it can also make things weirder and, in this case, more difficult. Working with the AI vocals is much harder than working with a professional singer.”
Far from being a barrier, these apparent inefficiencies provide the creative friction Robin and Jesse rely on to spark ideas and shape their musical style. Similarly, the frequent imperfections produced by Jukebox – which they fondly depict as an “unreliable diva” – are a source of inspiration rather than one of frustration.
“We embrace the parts of it that feel broken,” Jesse says. “Odd little pops, warbles, and out-of-tune moments that no one in their right mind would ever consider keeping – but I love them. I think once AI vocal work gets ‘better’ I might no longer be interested; I want the dirt and the distortion that the current tech has.”
This algorithmic ‘weirdness’ is key to the otherworldly, genre-defying sound heard on The Greatest Remaining Hits. From the R&B melodies of Veronica to the country-inflected vocals heard on the beautiful Hoag’s Delta, each song conjures up a ghost from pop music’s past.
“The voices often sound familiar, yet you can’t quite put your finger on who or what they are,” Robin muses. “It’s a legitimately cool aesthetic effect, and one you probably couldn’t achieve any other way.”
That indefinable quality didn’t happen by accident – the duo went to great pains to ensure that their generated output didn’t copy the sound of any individual artist. “We very specifically did not instruct the AI model to imitate any particular singer,” Robin emphasises. “I added extra code to ensure it could only inhabit the fuzzy space between real voices in its training data. I think aiming for a particular performer's voice (with AI) is both aesthetically boring and morally scummy.”
“It’s an aesthetic you couldn’t achieve any other way.”
— Robin Sloan, The Cotton Modules
“It’s not about imitation,” Jesse confirms. “The whole point of this tech is that it can make sounds that have never existed before. The proliferation of ‘fake Drakes’ is the inevitable first stage of people discovering this tool, but I’d love to see things move past that quickly into finding new forms of music. I think this album is a great argument for what we can do here.”
The potential for new sounds, styles, and genres is an intoxicating prospect for the creatively curious, and it is for this reason that Jesse, Robin, and artists like them see the main disruptive potential of generative AI as empowering novel forms of human expression.
“Generative AI isn’t a solution to a problem,” Robin muses. “Instead, like most new technologies, it simply presents a whole swathe of new problems, challenges, puzzles, and opportunities. The question is: What new kinds of work might you generate? What things wouldn’t have been possible, or even imaginable, without such a model?”
In spinning a tantalising glimpse of music’s next thread, The Cotton Modules are offering the world answers to those questions.
“Working with the AI vocals is much harder than working with a professional singer.”
— Jesse Solomon Clark, The Cotton Modules
Clovis is a New Zealand born writer, journalist, and educator working at the meeting point between music and technological innovation. He is also an active composer and sound artist, and his virtual reality and live-electronic works have been shown in over fifteen countries around the world.
