Adversarial examples

Adversarial examples are inputs deliberately crafted to fool AI. A tiny, invisible change to a photo can make AI misclassify it completely.

• Adversarial examples deliberately fool AI.
• Tiny input changes can cause misclassification.