Loading…
23
Pro tip: I looked up how many images AI art tools trained on and one dataset alone had 5.85 billion image-text pairs
1 comments
Log in to join the discussion
Log In1 Comment
valc911mo ago
Heard a podcast last week where a guy who works in AI said those datasets are so huge that nobody actually knows what's in them. Like companies can't even check every image because there are way too many. Some of them got taken down later after people found out they scraped stuff like medical photos and personal blogs without asking. That number you found, 5.85 billion, is just one dataset too, there are several out there. Makes you wonder how any artist could ever track their own work in a pile that big. The whole thing is kinda wild when you really sit with it.
5