Two months ago, when the Drake-fake track “Heart On My Sleeve”——generated using AI——racked up millions of plays across music platforms, it set music journalists everywhere into deep speculation about the state of AI in music. How will this new era impact artists and their creative rights? What does this mean for the future of music and the creative process? Is the music industry doomed if AI models can generate or imitate their way to songs that are just as “good” as those created by humans?
It’s all an interesting dialogue, but there’s one critical element of music creation that is left entirely untouched by most, if not all, of the writing I’ve come across on this subject: sound design. Sound design is a critical, defining element of modern music creation. Before playing any sort of rhythm or melody, artists must first take sound and shape it into an entirely unique texture to convey their message. This is especially true in genres like techno, house, electro, ambient, and experimental music. It’s also a critical but often overlooked part of pop, rap, and mainstream music, where much of this debate about AI is currently taking place. The thing is, with techno and other electronic genres, sound design is one of the focal points of the artistry. Many articles written about AI and music-making today still seem to look at music creation as something dealing solely in chords and melody——sound design doesn’t play much of a role in the dialogue. Few people have stopped to acknowledge that the viral Drake AI song has a cheap, Casio-keyboard-type-beat that sounds like a preset, but that doesn’t seem to dampen the fear-mongering around AI and its potential impact on music.
So, looking at this current debate through more of a sound-design-focused lens, I think it’s worth exploring the frequently asked AI-music question for techno: will AI one day make producers of techno, ambient, and other forms of underground dance music obsolete? Will AI one day be capable of making better techno than we can make as human beings?
I don’t think so.
MUSIC IS EMOTIONAL
As someone who has devoted his entire life to music, I’m biased in my answer here. But I think the case becomes clear when you look at music as an art form that is rooted entirely in human emotion.
Instrumental music like techno or ambient is an abstract art form. There is no direct subject; an instrumental song has no ability to represent something tangible. It can only suggest emotions or feelings using sounds and rhythms. The listener, when the song succeeds, empathizes with the emotions put into the song by the creator. A lot of times with something like techno, there’s room for the listener to insert their own meaning beyond what was intended. There’s no set path, and it’s all pretty nebulous stuff, like an abstract painting. But there’s one thing that’s undeniable: instrumental electronic music is all about emotion. I think it’s pretty safe to say this is not the realm of AI. Emotion is defined as an “instinctive or intuitive feeling, distinguished from reasoning or knowledge.” AI, which is entirely rooted in reasoning, knowledge, and analysis, has nothing to do with this. It couldn’t be further away from music’s most fundamental elements: feeling and emotion.
Now, when you look at music as more than just chords and melodies, and start talking about sound design, I think the argument becomes even more clear.
There is no “sound design theory” like there is classical music theory, which is rooted in the laws of physics and the overtone series. For centuries, songwriters have used analytical skills and logical reasoning to break down chord progressions and melodic sequences of existing songs as inputs to generate new chord progressions and melodies of their own. Even the godfather of modern pop music, Max Martin, relies on a formula he calls “Melodic Math.” There is a great deal of logic and reasoning involved here when it comes to the chords and melodies of modern music——and this is the exact type of thing that AI is good at replicating. It can analyze existing models in a period of training before generating new sequences and combinations based on the rules it learns and observes. I can definitely see the argument that AI might one day rival us in the process of finding chord progressions or generating melodies based on analysis of what’s previously worked on the charts.
When it comes to designing interesting sounds for expressing melodies and rhythms though, it’s a different story. There’s no rulebook or framework to follow. You just feel it. It’s very difficult to explain why something works as a sonic texture and why something doesn’t. Sound design has limitless possibilities for textures that might sound good, and even more for those that will sound bad. Finding the right one for your song is something you just have to feel for yourself. For example, there is no set of guidelines for how to dial in just the right amount of reverb for a specific type of synth sound to create a feeling of longing or loss. There aren’t filter settings that convey notably distinct colors or moods (contrast this with intervals in a melody though, which do typically cluster around specific emotions or feelings). In sound design, the difference between something sounding good or something sounding like “too much” is often a very fine and delicate line that music producers draw every day in the studio, navigating using only their own human intuition.
In techno, electro, house, ambient, and experimental music, this is especially true. Entire techno songs hinge on the perfect amount of reverb and distortion applied to a single note. Just the perfect amount of touch on a synthesizer’s envelopes can define an entire track. For many of us producing this kind of music, emotion or feeling is the only true north star. You need to keep your feelers sharp. You need to keep your humanity wide open, front and center.
CAN WE EVEN DEFINE AI MUSIC AS MUSIC?
So I guess what I’m getting at here——techno as an abstract art form held together by sound design and feeling——is that there’s actually a really interesting philosophical question at the core of this AI debate that everyone seems to be ignoring. If instrumental music, like techno, is an art form that is all about the emotional connection between creator and listener, is AI music really even music at all? How can a machine, something that is not capable of emotion, lead us on an emotional journey that helps us learn something about ourselves and our own humanity? I’m not sure if that’s possible.
Sure, a machine can arrange sounds into some sort of “sonic sculpture” that we could meditate on, but it’s not really music. There isn’t a story or journey about the human condition at the core, because a human being did not create it. There isn’t an underlying emotion that forms the bond between creator and listener that is——to me, at least——the defining feature of instrumental music.
I think there are branches of AI-music, like generative music, where there is definitely hope for a more positive outcome between algorithms and music. With generative music, there’s a human emotion and vision steering what an algorithm creates. There’s a symbiosis between man and machine that could actually be really interesting. There’s still something human for the listener to connect to, and there’s definitely promise for where this could go if done well.
AFTER ALL, IT’S NOT JUST MUSIC
Stepping back, I don’t mean to get overly philosophical about this——I mean, it’s just dance music right? It’s just a kick drum and some samples in a dark loud room, right? Honestly, no, there’s so much at stake here. Every day, more and more, our humanity is under attack. In professional and creative fields all over the world, those that get ahead are increasingly those that know best how to “mimic the machines.” Want to succeed in your career? Turn off your feelings and act like a machine, reducing your life to a cost/benefit analysis of things that do and do not get you ahead and focus only on what serves your interests. Want to be an artist? Learn the algorithms that shape today’s media landscape and build your creative output around content that best manipulates those algorithms. Want to live a life of love, family, friendship, creativity, spontaneity, curiosity, passion, and fulfillment deeply in touch with the very things that make us human? Well, sorry, in today’s world, you’re on your own. Our world is increasingly built by and for machines, and The Machine encourages us humans to subvert our instincts and emotions——the very things that make us who we are.
It sounds bleak, I know. But there is a way out, and that’s why I think it’s so important we talk about sound design and emotion when we talk about AI and music. No recycled tracks or sound-alike presets. No grabbing at trends or building artistry around algorithms or social media. We have to put emotion and feeling at the forefront of our work. Because it is our work. We have to, quite literally, put our hearts and souls into it.
-Holden
New tracks shared to the page here:
Gaetano Parisio and Danilo Vigorito - Untitled
So much funk and drive in the low end. Some head-spinning lead work overtop. 2003.
Ice Pack - J.D. 800
Furious detuned synth work from Ice Pack. Tension to the max, releases nicely once the beat drops into that eighth note hat. 1998.
Unknown Artist - Buzzsaw
Love coming across tracks like this because it’s tough to put a finger on exactly why it works. Usually not so drawn to this type of overdriven, single-note bassline with the kick, but the pairing of something so brutal with the more introspective mood of the chord sound overtop really goes somewhere unique. 2000.
Sompek - Untitled
Great reverbs on the synths in this track. Nice sense of space and emotion——bit of warmth but still driving and peak in its energy——always pushing forward but never quite giving you that resolution. 2001.
Taxi Drivers - Rauschengel 2
Great build from just a kick drum in the full version of this track. The phased out noise sound throughout brings so much drama and intensity, some real nice sound design happening there. 2000.
Markus Schneider - The Pulse
Nice low end drive with some great touch on the sample-based atmospherics. Super simple and open arrangement that just works, real proper stuff off of Construct Rhythm. 1997.
Ruben Anderson - The Funk Song
Beautiful touch and simplicity on this one from Ruben Anderson. Timeless chord line, restrained hats, lots of mood…the funk song for sure. 2000.
A. Squared - The End
Crunchy but pure. A. Squared with some Dean Cole vibes on this one, brutal approach to minimalism. One synth line and a drum machine doing so much. 2000.
Analogue Underground - Codebreaker
Cool drum programming on this one from Analogue Underground, can only find one record to their name. Very straight-ahead and driving but maintinas lots of nice interplay between the hat lines to build tension and make the most out of a super simple two note riff. 2000.
Justin Berkovi - The Server Sleeps
Nice slice of sci-fi tech-doomcore techno—fully off-kilter, never quite catching itself and constantly climbing higher with the rising dial-up synth work overtop. 1998.
DJ Apollo - Black Eye
Pure, stripped back synth loops on this one from DJ Apollo. Gets so much funk from such a minimal palette. 1997.
S-Mode - Low End
Beautifully pure track from S-Mode. So restrained and focused on maximizing the impact of the interplay between each element. Love how the 16th claps work with the open hat line. 1997.