What is a Prompt?

Think of a Large Language Model as a very powerful text generation engine. It knows a tremendous amount about language, grammar, facts, and even different writing styles, but it doesn't know what you want it to do until you tell it. That's where the prompt comes in.

A prompt is simply the text input you provide to an LLM to get it started. It's the instruction, the question, the piece of text to complete, or the topic you want the model to write about. It's your side of the conversation.

Consider these analogies:

When you use a search engine, you type in keywords or questions. That's your prompt.
When you use a calculator, you input numbers and operations ( $2 + 2$ ). That's your prompt.
When you talk to a digital assistant, the phrase you speak ("Hey Assistant, what's the weather?") is your prompt.

With an LLM, the prompt serves a similar function but is often more flexible and open ended. It guides the model's text generation process. The LLM takes your prompt, processes it based on its training data, and then generates a relevant continuation or answer.

Input (Prompt) -> LLM Processing -> Output (Response)

For example, if you provide the prompt:

Explain the main benefit of running an LLM locally.

The model understands you're asking for an explanation about a specific topic. It will then generate text based on its knowledge, aiming to answer your question directly.

The quality and clarity of your prompt significantly affect the quality and relevance of the LLM's response. A well-structured prompt clearly states the task or question, making it easier for the model to generate a useful output. Conversely, a vague or ambiguous prompt might lead to unexpected or unhelpful results.

Prompts can range from very simple to quite complex:

A question: What is the capital of France?
An instruction: Summarize the following text: followed by the text.
A sentence starter: Once upon a time in a land far, far away, there lived a...
A request for creative text: Write a short poem about a robot learning to paint.

In the context of this course, learning to write effective prompts is the primary way you will interact with and control the local LLMs you run on your computer. The upcoming sections will show you how to structure these prompts for different tasks.

Was this section helpful?

References

Introduction to prompt engineering, OpenAI, 2023 (OpenAI) - Official guide explaining prompt engineering concepts, starting with the basic definition of a prompt.
Guide to prompt design, Google for Developers, 2024 - Official documentation from Google providing an overview of prompt design, including the role of prompts in interacting with LLMs.
A Survey of Large Language Models, Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, Yifan Du, Chen Yang, Yushuo Chen, Zhipeng Chen, Jinhao Jiang, Ruiyang Ren, Yifan Li, Xinyu Tang, Zikang Liu, Peiyu Liu, Jian-Yun Nie, Ji-Rong Wen, 2023 arXiv preprint DOI: 10.48550/arXiv.2303.18223 - This comprehensive survey provides a broad overview of LLMs, including their interaction mechanisms like prompting and in-context learning. It establishes the context for how prompts are used.