Raw LLM Responses

@MrToastyYT Let’s clarify a few things, because your response misses the forest for the trees: Reverse image search ≠ machine learning. Reverse image search is based on perceptual hashing or pixel comparison — not neural networks. So bringing it up here is irrelevant. The issue isn’t whether a human can manually find the image, it’s that machine learning systems — including copyright detection and moderation tools — do rely on the exact features Nightshade distorts. If your goal is to protect artists from theft, degrading the very systems designed to detect it is… counterproductive. “Medical imaging? Really?” — Yes, and here’s why: You’re thinking too narrowly. AI models are converging. The same transformer-based architecture used in Stable Diffusion or Midjourney is increasingly used in multimodal systems — models that handle art, photos, video, and even medical scans in the same backbone. If we normalize poisoning vision data at scale, we risk undermining shared infrastructure used in safety-critical domains. Not today maybe, but soon — and preemptively polluting the ecosystem is reckless. You say it’s not vandalism. It absolutely is. If an artist injected artificial noise into a museum’s collection database so that no digital preservation system could correctly catalogue or detect their work — would that be “self-defense,” or sabotage? Nightshade creates false visual cues that can mislead AI into thinking a real artwork is synthetic. That means it’ll likely trigger false positives in AI-detection systems. Ironically, it makes it harder to prove your art isn’t AI. That’s self-defeating. Marginalized creators need automated protection most. The “16-year-old in Liberia” is a perfect example — someone without legal backing, trying to make it online. If their work gets scraped, reposted, or stolen, and copyright bots or authenticity detection fail because someone “poisoned” the ecosystem to stop AI, how do they defend themselves? These tools aren’t perfect — but sabotaging them guarantees that the most vulnerable creators will be left completely exposed.