Important stats on AI music from @bennjordan, adding to the growing evidence that AI music is competing with the human music it's trained on. - Of 560 top songs & staff picks on Suno, 549 (98%) are being monetized on streaming services - These are often on Spotify account posing as verified human artists - One track Benn shows has 688,000 plays. Just that one song has taken ~$2,000 from the Spotify royalty pool away from human artists. - As of June 2024 there was a 15% increase in songs uploaded to Spotify vs. the previous year. The increase is coming mostly from 'independent musicians'. Sadly, lots of this is probably AI music. There is an epidemic of AI slop on music streaming services, taking money from the artists whose work was used to train the AI models that generated it. Once again - generative AI competes the work it's trained on, and the people behind that work. Strongly encourage watching the whole thing: youtu.be/QVXfcIb3OKo?si=6mxW… h/t @jesseengel

This is a good example of why generative AI opt-outs don’t work: this is my music, but I have no way of opting it out of AI training. I composed this piece, but I didn’t make the recording and I don’t control the website where it’s hosted. So I can’t update the robots.txt file to opt out of training, I can’t add any metadata to the recording, and I can’t add the composition to a registry that will stop it being trained on. Even if AI music companies respected opt-outs (which the biggest ones don’t), I couldn’t opt this piece out of training. Whenever a choir makes a new recording, that new recording will not be opted out. And this is just one example from one creator. Imagine how many creative works will be opted into AI training against their owners’ will if governments make opt-out the norm for generative AI. #OptOutsDontWork boosey.com/audio-clip/I-hear…