|
I have a new rule. Universally true: All AI outputs prompted by other people is subjectively mid. But all AI content you spawn yourself is objectively great, and interesting.
Full Show Notes: https://zexulo.xyz/2025/06/01/2512-overdosed-on-ai-music/
Experience.Computer: https://experience.computer/
Worldrunning.guide: https://worldrunning.guide/
Subscriber Zine support the show! https://startselectreset.com/Permanently moved is a personal podcast 301 seconds in length, written and recorded by @thejaymo
Subscribe to the Podcast: https://permanentlymoved.online/
Quarterly zine; my gift to you ✉️
Overdosed on AI Music
Sloppy Intro:
This week, I stood at the Slop Machine, experiencing not the future, but the extreme present.
All day, every day, I’ve been listening to AI-Generated music. And this experiment has provoked some unexpectedly strong reactions in me.
The first is a reactionary whiplash, but not in the way you might be expecting.
Music is a deeply personal pleasure in my life. Half of my undergraduate was music. I play instruments, I’m in a band, I seek out new music constantly, and try to see it live whenever possible.
For all my life, when I’ve said to people that I love music, I’ve meant two things: Live performance, the direct communication from musician to audience, and secondly, my relationship with recorded sound: CDs, records, mp3s, streaming, etc.
This past week has troubled my relationship with the second definition.
Many, myself included a week ago, would expect this experiment to have driven me towards romanticising ‘human’ authenticity or creativity. Real recorded music, recorded by real people. But that isn’t what’s happened.
Instead, something more unsettling occurred. The very seamlessness of the AI-generated material, the knowledge that its sounds were entirely synthetic, ‘engineered’, and it has cast a shadow over my perception of all recorded music. A kind of unease about the truthfulness (or untruthfulness) of any media that wasn’t live and unmediated.
I’ve developed a pervasive negative reaction to all of what I’m going to call “engineered sound.”
This umbrella covers everything: studio recordings, electronic music, and AI tracks. AI-generated or human-made, all recorded music now has this weird uncanny quality to me. It’s all just engineered sound and it makes me uneasy, ontologically. This unreality, this sense that the material is synthetic, has now tainted my experience of all recorded music.
Like I’m a modern day John S. Wilson calling the entire enterprise of recorded music suspect.
As a music lover, I’m concerned. Will this uneasiness dissipate if I stop listening to AI? I suspect I will need a break from “all music” – or rather, engineered sound – altogether.
My second realisation concerns my taste.
Listening to AI playlists all week, particularly from newer models like Suno 4.5, the experience is often seamless; an AI track can be largely indistinguishable from an unfamiliar one on Spotify—frankly, often just as ‘mid.’ You genuinely might have no idea. And as I’ve said, if it’s new to your ear, it’s simply new.
Now, hours of current AI music can be tiring due to hit-and-miss fidelity, causing ear fatigue, much like a badly recorded podcast or static-laden radio. But that’s a quality issue, not content-related, and it’s rapidly being solved. With SUNO’s fancier v4.5 model, or what’s next, audio fidelity is basically solved. It’s going to sound just like, quote, ‘real life.’ And I think this is bad.
I’ve long thought that AI-spawned music is best when it leans into itself as a medium. This week confirmed that: I don’t want endless, perfect pop simulations with hyper-realistic vocals. Just as the most interesting AI images are puppy-slugs, not photorealism, the same holds true for music synthesis.
I want machine-dreamed music, not music dreamed up by a machine. I want alien music.
In his superlecture ‘Nobody listened to music anymore’ Marek Poliks argues music is now a web of relations, a database taxonomy; Spotify is a platform of metadata, not music.
So new combinations of genres, tags, sounds – things that might previously have counted as new genres or innovation now already exist within the logic of the database. All potential creative acts now become recombinations of existing metadata, latent possibilities within both databases and AI models, waiting to be actualised. By humans or AI.
Given my growing negative feelings towards any engineered sound, human or not, I don’t care if a human or an AI made that ‘classical/emo-trap/ambient/jazz’ track.
Which brings me to my last realisation.
I have a new rule, universally true: All AI text, video, and engineered sound prompted by other people is subjectively mid. But all AI content you spawn yourself is objectively great, and interesting.
All week I was prompting the Slop Machine to spawn songs like the intro, in the style of a manic pixie dream girl in a yoghurt advert. But based on the ancient Odes to Solomon. Unsurprisingly, no one around me is clamouring for New Age AI slop with 2000 year old religious lyrics. But I am, in a “This is my hole! It was made for me” kind of way.
Giovanni della Casa, in his 1558 etiquette manual Galateo, wrote, “One should not annoy others with such stuff as dreams, especially since most dreams are by and large idiotic.”
Joan Didion also remarked that “Nobody wants to hear about someone else’s dream… nobody wants to walk around with it.” There’s little more boring than hearing someone’s rambling dream, and this is true of AI slop shared on social media.
In this period of Information Age iconoclasm, machine dreams are now endless, infinite, and can inspire. But raw and unfiltered, they’re sloppy.
A new social etiquette for sharing machine dreams is required. I’ll happily listen to your dreams over a beer, especially if they involve me. But in general, in polite company? You don’t share dreams. And perhaps people shouldn’t be so quick to share the raw dreams of the machine either.
Subscribing to SSRZ supports my online work and creative projects.
As a thank you, I send you my zine four times a year, just like it’s 1994.
No spam. No email. Cancel at any time.
Leave a Reply