News for hackers. Stuff that matters.

A comment on Nightshade: A Defensive Tool for Artists Against AI Art Generators. Parent comment

Re:Nightshade: A Defensive Tool for Artists Against AI Art G... by visarga on Sunday January 21, 2024 @12:28PM (#39080480)
Modern generative image models are trained on curated data, not raw internet data. Sometimes the captions are regenerated to fit the image better. Only high quality images with high quality descriptions.
[ Parent | on HN ]
Re:Nightshade: A Defensive Tool for Artists Against AI Art G... by og_kalu on Sunday January 21, 2024 @11:19PM (#39086137)
I wouldn't call what Stable Diffusion et al are trained on "high quality". You need only look through the likes of LAION to see the kind of captions and images they get trained on.

It's not random but it's not particularly curated either. Most of the time, any curation is done afterwards.

[ Parent | on HN ]
Re:Nightshade: A Defensive Tool for Artists Against AI Art G... by visarga on Monday January 22, 2024 @11:28AM (#39091356)
Have you seen the BLIP paper? It's a bit old now, but it introduced a curation method.

https://arxiv.org/abs/2201.12086

[ Parent | on HN ]