Raw LLM Responses

"it doesnt work" - it does for now and when it doesnt anymore there will be a new Version that does. You can significantly reduce the quality of AI results that try to mirror your work for a long time with it and once they figured that out and manage to use your old stuff without concent they will probably still Not be able to get the New stuff without killing their model EDIT because i watched to 15:00 "they will just make it detect the poison" - I'm not sure this Person understands the process of Image/sound Generation by AI or the process of poisoning either pictures or music. The poisoning only works because of the core principle of how AI works. AI basically generates from random noise and predicts things from patterns it recognized in the Digital Version of an Image. This is true for Image AI as Well as music AI. the music stuff just does an extra step to transform Sound into what is, put simply, the Image of the wave form of that Sound and learn on that. It basically generated waveform-images and transforms Into what you hear in the end. So both AIs only do images in the end. The fun part: AI Sees _everything_ in the Image while we do not. If it is there, the AI will Pick it up. If it is a prominent pattern it will Pick that up even faster. Now Nightshade, poisonify and whatever else there is will not just put the same stuff in every Image so the AI can recognize that one pattern and get that filtered out. The "poison" is different each time based on the Image that is put in but still very prominent if you only see a Serie of 1s and 0s. The AI needs to Figure out what is unique in that Image to Replicate it but it Sees more than us and it will Spit that out aswell. Since it is in a form that seems very prominent for an AI it will Enhance the poison Part because in all other images those patterns would be the right patterns to replicate. In the end AI is not going the Route of Early machine learning where it was basically glorified if/else Statements and you can't just exclude something specific that easily. I do believe the images that have been poisoned with current technology can be used for AI in the future because that technology is obviously evolving fast but I am also 100% certain that the projects that try to counteract that are constantly evolving aswell. You won't get an AI that magically detects malicious parts any time soon.