Zero-Shot Learning

Using AI

When an AI handles a task it was never explicitly trained for, working purely from instructions and general knowledge.

Zero-shot means no examples given. You ask a model to classify support tickets by urgency, or translate into a style it never saw labeled data for, and it does a reasonable job purely from understanding the instruction.Large language models made this normal - describing the task in plain words is often enough. Related tricks: one-shot and few-shot, where you include one or a handful of examples in the prompt to sharpen results without any retraining.