In plain English
Zero-shot learning is when a model performs a task it was not explicitly trained on, using only a natural language description of what is required and no worked examples. It relies on the broad knowledge the model picked up during pre-training. It is the simplest way to use a capable model and the right starting point for clear, well-defined tasks.
Why it matters
Zero-shot prompting is fast and effortless, and on modern models it handles a surprising range of tasks, so it is worth trying before reaching for examples or fine-tuning.
A worked example
Without any examples, you ask 'classify this review as positive, negative, or neutral', and the model does it straight from the instruction.
Common confusion
Zero-shot means no examples in the prompt; few-shot means a handful are included. For tricky or very specific formats, a few examples usually beat none.