/e/, only we have the taste for proper AI training.
If you're out of the loop, multiple projects have opened up showing that there are people using self-attention GANs instead of convolutional GANs to handle cartoons instead of real life pictures. This has created a new level of degeneracy science previously thought impossible:
https://www.gwern.net/CropsNow because I don't know if I'm allowed to share these kinds of sites, I won't. But what I can do is take the images here and upload them. AIs just learn from what you show them. That means we have an opportunity to inflate the amount of boobs we like in the dataset to bias it towards our preferences. I'll winzip up everything (I paid for the Winzip 24 Pro version) and go around to the couple of sites I know doing this and will deploy huge honkers right into those AIs - they won't know what hit them.
>TLDR If we upload only big boobs to AI datasets, they will bias towards drawing more big boobs. Share big boobs here and I'll upload them. 