Listening to All the Music AI Is Trained on Would Take Decades

New from @alexreisner.bsky.social: Millions of songs—including hundreds by Taylor Swift and Paul McCartney, and "Penile Drip" by Gwar—have been sucked up by AI companies, which often claim to only use free music, which clearly produce plagiarized tunes, which are clearly doing whatever they want.

Listening to All the Music AI Is Trained on Would Take Decades We obtained four datasets that reveal an astonishing number of songs. 🎁-link

😠 www.theatlantic.com/technology/2...

Curious which Suede songs have been swallowed by the machines grinding out soulless AI pastiche-pop? The short answer is 'pretty much everything' but for the detail... www.theatlantic.com/technology/2...

AI simulates, it doesn't create. With language, we fall for the illusion. With music we don't, when we listen with our hearts. "Walk My Walk" charted anyway. Maybe the problem isn't AI. It's us. www.theatlantic.com/technology/2...

AI developers who knowingly stole artist’s music to train AI models should be charged with felonies, stripped of all of their profits, and have to pay restitution. Now that the cat is out of the bag, these artists will forever be harmed by this theft. www.theatlantic.com/technology/2...

Investigators found 12 million scraped tracks in datasets used to train AI music generators, as record labels sue Suno over copyright

"[F]our giant datasets of songs [...] are being shared within the AI-development community. One has 12 million tracks. Another has 9 million. The two smaller datasets each have more than 100,000" www.theatlantic.com/technology/2... via @theatlantic.com

www.theatlantic.com/technology/2...

www.theatlantic.com/technology/2...