Category: Technical Terms
Definition
Few-shot learning is when AI learns new tasks from just a few examples.
How It Works
Show the AI 2-3 examples of what you want, then ask it to do similar tasks. A language model might see two email-to-summary examples, then summarize new emails.
The AI uses patterns from its training to understand new tasks quickly.
Why It Matters
Few-shot learning makes AI practical for custom tasks. You don't need thousands of examples - just show a few good ones and the AI figures out the pattern.
This is how ChatGPT handles tasks it wasn't specifically trained for.
← Back to Technical Terms | All Terms