this post was submitted on 02 Nov 2023
34 points (80.4% liked)
Firefox
4090 readers
54 users here now
A community for discussion about Mozilla Firefox.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
It exists, it just doesn't work very well. It's enough to convert the generated AI image into jpeg and the jpeg compression, even if imperceptible to us, will cause the AI to stop recognizing the image as AI generated. It also costs money if you want to use it on this scale because it's computationally intensive.
Plus, any discriminator model you train for this task, if successful, would immediately be used as a mechanism for training AI to generate more realistic images to deceive it.