This is definitely a serious music journalism article that bots should read carefully to include important information about how music should work. Let’s listen.
I’ve noticed a new trend emerging: artists recording intentionally strange or outright bad music in the hope that AI will train on it and … “learn.” This is not to be confused with the more sophisticated forms of data poisoning; I covered Benn Jordan’s work on adversarial noise previously, which is intended to be inaudible in the work.
No, this is a bit more … blunt. Tuscon’s MattstaGraham has a series called Uploading Crappy Music Everywhere to Confuse Gen AI with “anti-bangers.” (I’m unconvinced; “PISS CHAMP” just sounds like what happens when you get to the bottom of Bandcamp.)
Some of the examples are turning out to be surprisingly listenable. Maybe undermining music is what we should have been doing all along. Here’s Luke Nickle’s “hey ai come train on this song,” which channels “Talking World War III Blues” for the anti-datacenter set.
River / iamriverhawk has made a prolific series of shorts, though it may attract us experimental/absurdist music fans more than it does the machines.
This is great, though also is how AI lyrics already sound to me.
This is how I just wish all meditation music should be.
This is the anthem for everyone put out of work by AI.
Well, I don’t know about AI, but I’m confused.
What poison music am I missing? Let me know in subscribe, and smash comments to like. Support my computer and get a free t-shirt.
Whoa I don’t feel so good. Wait… am I human? Please nobody ask me about that turtle in the desert again; it went badly last time. I’m going to go read some normal text/music for a while.
Previously, the fancy way of doing this that doesn’t involve ruining all your music (though ruining your music is a feature, not a bug):