We tried to get AI to make KONGOS music...

And the results will shock you

by Johnny Kongos

Usually when I was done working hard (or pretending to) in the studio at the end of a day, I’d veg out in front of Netflix etc. But over the last few months I find that more of my "off hours” are spent fooling around with generative AI, mostly making stupid memes for one of the countless group chats I’m on, but also trying to learn about it and find ways that it could actually be helpful to my work or life. It’s been fun, and I’ve learned things that have made significant improvements to certain aspects of the business side of our music career.

After you read this story, come back to this link for a live acoustic version of This Time I Won’t Forget from our album Lunatic:

But when I became aware of Suno, a generative AI music site, I signed up immediately... 50% thinking “let’s see just how bad this AI music is, there’s no way…” and 50% terrified that the time had come when even I would meet my end in the hostile digital takeover, me… an essential worker, an artist, the most important of all tradesmen. (I know sarcasm doesn’t read well online so for 100% clarity, I’m not being sarcastic 😜).

I generated a song or two…

“Holy shit this is incredible.” And it’s only going to get better. About 108 songs later, countless WTF moments and a bunch of lolz, I started to feel that there was a subtle but essential element missing from everything I was hearing. It’s impossible to describe but people have tried:

“After silence, that which comes nearest to expressing the inexpressible is music.” ~Aldous Huxley.

Whatever that inexpressible thing is that music tries to scratch at, most everything I was generating in AI felt like it was missing it.

Very occasionally when I’m writing a song I have a feeling for brief moments somewhere in my solar plexus that can only be poorly described as the friction between a deep “sadness” that contains no negativity and a profound “happiness” that is not based on the pleasing of the senses. Sad is the wrong word, so is happy… like I said, poorly described. But the point is, when I experience these moments, it’s markedly different than my feeling on the spectrum of happy or sad. I trust this feeling, because when I’ve felt it before while writing, the songs or at least sections of the song ended up being my better work. Songs that I inevitably listen to years later and think “I don’t remember how I wrote that”, and “I’m not sure I could do it again.”

This is a long way of saying that so far AI does a great job of generating emotionally vacant material that is almost indistinguishable from a lot of what is currently topping the charts, it excels particularly in pop, EDM, pop punk, even hardcore, but it’s still missing something important, I’m not getting anything resembling the feeling I just described.

Here are some examples. FYI all I did was type the prompt and hit enter. Everything… lyrics, music and the lyric video - was all generated… in about 22 seconds. I don’t like any of this music, but I don’t know if I’d know the difference between this and 90% of the stuff I hear daily on TikTok.

Prompt 1: “A song about the end of the world. Female vocal. 2020s minimalist pop. Sultry. Medium tempo”

Prompt 2: “A song about the end of the world. Male vocal. Raspy. 2020s trap. Very sparse. Haunting piano melody. 808s.”

Anyway you get the point. Is this creativity, consciousness? That’s a longer and probably impossible discussion, but in practical terms this is going to radically change the music business. Picture playlists that are entirely generated on topics, moods, genres all to your specific tastes. Plus the thing about this technology is we don’t seem to be able to accurately predict exactly how it’s going to change things.

So what about us? Can AI take our job as KONGOS? A band that ChatGPT describes like this:

“KONGOS blend rock, alternative, and electronic elements with a distinctive rhythmic drive rooted in their South African heritage, often featuring accordion and heavy percussion. Their sound is cinematic and anthemic, combining introspective lyrics with explosive, groove-driven arrangements. Across albums, they’ve explored themes ranging from existentialism to social commentary, all with a raw, genre-defying energy.”

Let’s try. I asked chatGPT to turn that description into a prompt that Suno will understand to try to replicate KONGOS’ sound. You can’t actually tell Suno to “make a song like KONGOS, or Queens of the Stone Age”, but in most cases you can approximate the description and get something pretty close to the band’s sound.

KONGOS Prompt 1: ”A song about the end of the world. Rock/alt with heavy drums, accordion, and electronic elements, cinematic and rhythmic, introspective yet anthemic.”

Hahahahaha. Let’s try again. I’ll alter the prompt to try to target Come With Me Now specifically. That song must have been ingested by AI at some point right? (allegedly). Here’s how ChatGPT describes the song:

“High-energy rock song with driving tribal drums, gritty accordion riffs, aggressive vocals, rebellious tone. Alt rock with kwaito-inspired rhythms, intense, anthemic”

Pretty accurate. Let’s give it a go. I included 2 results after generating a bunch of songs with the following prompt.

KONGOS Prompt 2: “A song about the end of the world. High-energy rock song with driving tribal drums, gritty accordion riffs, aggressive vocals, rebellious tone. Alt rock with kwaito-inspired rhythms, intense, anthemic”

Listen, I’m sure we could refine the prompt and get it closer, and I know that if Suno let you specifically target an artist it could get it much closer, but for now I’m calling it…

KONGOS is safe.

Especially if we continue to push ourselves to keep trying to find something new, something essential.

Let us know your thoughts, but please if you find a way to generate our sound in AI don’t let us know. Let us live in blissful ignorance so we can continue happily creating music with our heads in the sand.