Meta: adversarial noise
AI generated discoloration maps applied to images (typically with software like Nightshade or Glaze) to attempt to reduce the use of art as stolen training data for generative AI.
These techniques can introduce noise to an image that's visible even to the naked eye (primarily in images that use flat, simplistic coloring).
The efficacy of these tools is also subject to debate, as various techniques exist to explicitly remove adversarial noise or otherwise render it useless. There are also concerns about health and accessibility (as adversarial noise is known to cause dizziness and eye strain in some), as well as criticism about its effect on art quality being comparable to JPEG compression artifacts.
The following tags are aliased to this tag: glazed_ai (learn more).