← Back to context

Comment by alentred

3 years ago

With this "solution" it looks like the world of art enters the cat-and-mouse game the ad blockers were playing for the last decade or two.

I just tested it with Azure AI image classification and it worked - so this cat is yet to adapt to the mouse’s latest idea.

I still feel it is absolutely wrong to roam around the internet and scrape images (without consent) in order to power one’s cash cow AI. I hope more methods to protect artworks (including audio and other formats) become more accessible.

  • Artists copy from each other all the time. Arguably, culture exists because of copying (folk stories by necessity); copyright makes culture top-down and stagnant, and you can't avoid it because they have the money to shove it right in your face. Who wants trickle-down culture?

I might be missing something because I don't know much about the architecture of either Nightshade or AI art generators, but I wonder if you could try to have a GAN-like architecture (an extra model trying to trick the model) for the part of the generator that labels images to build resistance to Nightshade-like filters.

  • It doesn't even have to be a full GAN, you only need to train the discriminator side to filter out the data. Clean reference images + Nightshade would be the generator side.