What is Prompt Engineering?

Manuar Alam • 2 min read • 10 views

Prompt engineering is the art of crafting instructions to guide AI models, particularly LLMs(Large Language Models), like OpenAI's GPT4, Google Gemini, Anthropic Claude, etc, towards generating the desired outputs.

In simple words, asking questions or giving better instructions to AI to get the better results.

Types of Prompt Engineering :

Zero-Shot Prompting: Providing AI models with clear direct instructions without any example.

Example:

Input: Describe cloud computing in simple terms.

Output: Cloud computing is the on-demand delivery of IT resources over the internet with pay-as-you-go pricing.

One-Shot Prompting: Provided instructions with a single example to perform its task.

Example:

Input: Linux command for system info

Output: uname

Multi-Shot Prompting(Few-Shot Prompting): Providing an AI model with a few examples of a task to guide its response.

Example:

Input: Dog -> Pet, Lion -> Wild, Cat->

Output: Pet

Chain-of -Thought Prompting: This technique enhances the output of LLMs, particularly for complex tasks involving step-by-step reasoning before giving the final outcome.

Example:

Prompt: I have $20, spend $5 on food, and find $10 on the street. How much do I have? Let's think step-by-step.

Tree of Thoughts: This technique makes the AI consider multiple solution paths, compare them, and select the best one.

Example:

Prompt: Come up with three possible dinner recipes, evaluate the prep time for each, and write a grocery list for the quickest one.

Prompt Chaining: It is a technique where multiple prompts are linked together so the output of one prompt becomes the input for the next, allowing the AI to complete complex tasks step-by-step.

Example:

Task: Analyzing disk usage

Step 1: “Explain the purpose of the du command in Linux.”
Step 2: “Write a command to find the largest directories in /home.”
Step 3: “Create a small bash script that lists the top 5 largest directories.”
Step 4: “Review the script and optimize it for better performance.”

Each step uses the output of the previous step to build a more complete solution