Imagine you have a super-smart robot that is an expert at identifying pictures. If you show it a picture of a cat, it says "Cat!" every single time.
Here at AdversarialWonderland.com, we make "adversarial images"—which are basically "invisible costumes" for pictures. To you and me, the picture still looks exactly like a cat. But we add tiny, secret patterns to the image—sort of like a magic spell—that are invisible to humans but totally confuse the robot.
When the robot looks at our cat wearing its "magic costume," it might suddenly get confused and yell, "That's a toaster!" or "That's a school bus!"
We love them because they show us that even though AI is very fast, it doesn't "see" the world the same way we do. It's like finding a secret glitch in a video game that lets you walk through walls. It proves that human brains are still special, because we can't be easily tricked by a few tiny dots.
The art we make uses these exact same "magic costumes." By building our art out of patterns that AI can't understand, we are creating a "Shield."
When you support our art, you are helping to build a world where AI doesn't get to decide what everything is. It's our way of saying: "You might be a smart robot, but you can't see the whole picture like we can." We use the AI's own "glitches" to make something beautiful that the AI can't control or copy perfectly. To learn more about all the good things you're helping us build, check out our Terms.