this post was submitted on 17 Jun 2026
43 points (97.8% liked)

Technology

2594 readers
51 users here now

Tech related news and discussion. Link to anything, it doesn't need to be a news article.

Let's keep the politics and business side of things to a minimum.

Rules

No memes

founded 1 year ago
MODERATORS
 

Millions of copyrighted songs trained AI music generators, and new searchable databases from The Atlantic now confirm which tracks were used.

top 2 comments
sorted by: hot top controversial new old
[–] misk@piefed.social 12 points 4 days ago* (last edited 4 days ago)
[–] mindbleach@sh.itjust.works 2 points 4 days ago

Yes, that's how it works. And it's fine because training is transformative use. Sluicing millions of songs into a gigabyte-sized model means each one contributed about a kilobyte. Even for MP3s, a kilobyte is approximately zilch. 96kbps is 12 kilobytes... per second.

You are completely free to do math about copyrighted works. By all rights, you should be free to do a lot more, because copyright is only a monetary incentive for new art. Nothing past thirty years old should even be in question, because it belongs in the public domain. People have a right to culture, and anything you grew up with is yours to share and iterate upon. Don't make the mistake of getting so mad about spicy autocomplete that you demand more power for Disney and Sony.