Post #3912258
2026-07-18 12:40 UTC
@davidrevoy@framapiaf.org That's because AI was trained on your many great CC-licensed webcomics! Instagram doesn't know you're the source. 😉
Replies (5)
-
@proedie@mastodon.green 2026-07-18 12:51
@Pezito@mamot.fr @davidrevoy@framapiaf.org This.
-
@DrJLecter@beige.party 2026-07-18 12:57
@Pezito@mamot.fr yepp, very very VERY likely 💀💀 @davidrevoy@framapiaf.org
-
@crimsonfall@mas.to 2026-07-18 13:22
@Pezito@mamot.fr when the ai be so incest it can’t do its own research :blobfacepalm:
-
@qwazix@bananachips.club 2026-07-18 16:49
@Pezito@mamot.fr @davidrevoy@framapiaf.org probably this is the reason, yes.
-
@csolisr@hub.azkware.net 2026-07-18 18:34
I'm surprised (not) to see that even the few models trained solely with copyleft-compatible data still don't bother tracking which tokens correspond to which files for attribution purposes - at most they're tacked in after the model is trained already.