Fuck AI

Here's my unmedicated and completely idiotic take.

I'm glad he's doing these tests, but it's pretty obvious in this case. The lyrics are very obviously AI generated, first of all. The overuse of certain words (smoke, boots, multiple song titles with the word dust, shakes, whispers, echos) are a strong indicator. Not every song using those is AI generated, obviously. But it's like they said "see what words an LLM overuses when trying to write songs and put as many of them in a verse as possible".

Ok, so maybe they just use it as a songwriting tool because they have no lyrics but they're great musicians and want to put something out. Even without splitting stems most of it currently feels unnatural. But he makes a great point. Just listening without splitting there's weird reverb. There are strange vocal artifacts. The guitar transients don't make sense (which is part of why his splitter tool couldn't split things properly). For a while everything sounded underwater and there are still traces of that. I see they moved past the country voice that was all over the place for a while. There's a "shimmer" (I wish I had a better word to describe it) that sounds wrong.

It sounds like they took a human track and put every free plug-in they could find on it and dialed it up past the point of good taste. There's no dynamic range. There's too much (or just flat out wrong) reverb. There's digital artifacts. A lot of that doesn't matter if you're listening on a Bluetooth speaker because you'll never hear it. Vehicle systems hype and mask certain frequencies and may not reveal it. But if you listen on any halfway decent system it's currently pretty obvious.

The real question: will enough people care? I do. Even if they could perfectly replicate session musicians and if they didn't sing about whispering to your boots and shadows echoing through the dust but were able to really write in the style of my favorite artists I would care. Because the people making new music, trying new things, and pushing boundaries wouldn't get the funding. So we would eventually end up with homogenized rehashes of what currently exists. If this had happened when all we had was orchestral music we'd all just be listening to AI rearranging that instead of having rap, rock, blues, metal, EDM, country, and all the other genres and their endless subgenres. We'll never get to moon stank or whatever the next big iteration of music is. And that's a loss for humanity because art is important.