Zero-shot Learning

Zero-shot learning is when an AI system is able to perform a task without having seen any direct examples of it during training. Instead, it uses general knowledge from related tasks to figure out how to handle the new one. It is like giving someone a riddle they've never heard before, but they solve it using logic and experience from similar puzzles.

This approach is useful when labeled training data is hard to find or constantly changing. For example, a zero-shot learning model might be able to recognize a new animal species just by reading a description of it, even if it has never seen a photo of that species before. It helps AI become more flexible and adaptive.