News for hackers. Stuff that matters.
Re:Search 5.8B images used to train popular AI art models by educaysean on Wednesday September 14, 2022 @06:43PM (#32844461)
One thing that sticks out to me is how so much of the the images in the collection have really terrible labels. I uncovered a large collection of pieces by an illustrator who was unsearchable by name, only via image upload. The reason they were unsearchable: the majority of this particular artist's images had labels that were all in the format of:

{username}'s profile image

Re:Search 5.8B images used to train popular AI art models by megablast on Thursday September 15, 2022 @12:34AM (#32847094)
Welcome to the real world of data.
Re:Search 5.8B images used to train popular AI art models by simandl on Wednesday September 14, 2022 @07:33PM (#32844933)
I'm not 100% on this, but I think a big portion of the captions for the images come from the alt-text, and are probably auto-generated by the sites they were scraped from.