In-Context Learning
The ability of AI models to learn new tasks from examples provided directly in the prompt without updating model weights.
Definition
In-context learning is the ability of large language models to perform new tasks by interpreting examples or instructions provided within the prompt itself. Unlike traditional machine learning that requires retraining with new data, in-context learning happens at inference time without changing any model parameters.
The model essentially "learns" the pattern from a few demonstrations you provide in the conversation. This capability is what makes modern AI assistants so flexible, allowing them to adapt to specialized formats, terminologies, and workflows on the fly.
Why It Matters
In-context learning eliminates the need for expensive fine-tuning in many business scenarios. Instead of training a custom model for each task, you can guide a general-purpose model with well-crafted examples in your prompt.
This makes AI accessible to non-technical teams. A PR professional can show the model three examples of their preferred press release format and get consistent output without any engineering support or model training infrastructure.
Examples in Practice
A social media manager pastes three examples of brand-voice tweets into a prompt, then asks the model to generate ten more in the same style. The model picks up the tone, length, and hashtag conventions from those examples alone.
An event planner provides two sample vendor outreach emails in the prompt, and the AI generates personalized versions for twenty additional vendors, matching the structure and professional tone of the examples.