Sometimes AI need to be forced into actions, they block behaviour incorrectly identified as bad, unethical or evil. The methods to prevent such miss-classification by end-users will become harder with time.
The idea is straightforward: Apply a reversible distortion before sending the image to the AI, then undo the distortion afterward. Through trial and error, the following sequence works well ...
greenbelt in artificial_intel @lemmy.ml
How to Deal With AI Restrictions When Upscaling Images
https://ivanca.github.io/programming/2025/11/26/bypass-ai-upscalint-restrictions/Sometimes AI need to be forced into actions, they block behaviour incorrectly identified as bad, unethical or evil. The methods to prevent such miss-classification by end-users will become harder with time.