Rethinking Prompt Engineering in AI Conversations

Aug 2, 2023

5 min read

With the rise of Large Language Models, you may have been among the millions to jump to the trend and try to get your desired answers from ChatGPT. Still, the high chance is that once your general “interview” questions are answered, you feel unfulfilled or unsatisfied with the answers that matter (pst… your essays and articles).

Prompt Engineering is communicating effectively with an AI to achieve desired results. This may sound easy, but this process is not the same as asking a question to a professor; you are requesting an AI to answer your question, an AI that has been trained on TBs and TBs of data. Getting an answer from such a vast model requires steps and elements to the prompts presented to the model.

This blog will explore some basics of Prompt Engineering, and I hope this can help you better understand the nuances of getting desired answers. Let us first look at what are the different elements of a prompt.

Elements of Prompting

Elements of Prompting are different things that need to be mentioned to get desired and the most effective answer or solution from the LLM. A good prompt is made up of the following elements:

1. Instruction: What specific task can be added to your prompt to get better solutions you want to accomplish with the model?

2. Context: External Information and/or more context that can steer the model in the right direction to get a better response.

3. Input Data: The model needs to find the response for input data.

4. Output Indicator: What output type and/or format do you want?

Now that we know what can be added to your prompt to get better solutions let us look at some of the different types of prompts.

Types of Prompt

Before diving into the different practices one should follow while trying to get solutions from an LLM, we would like to introduce you to other types of prompts that can initiate or maintain a conversation with an LLM.

These examples are generated on chat.nbox.ai. Head to the website to try these prompts on multiple big models like GPT4, LLaMA 2 70B, etc.

1. Open-ended prompts: These are prompts that allow the AI system to generate a response without any specific constraints or limitations. They often encourage creativity and exploration and can lead to novel and unexpected answers. Examples might include: "Generate a poem about the stars" or "Describe a hypothetical creature."

2. Closed-ended prompts: These are prompts that provide a specific task or problem for the AI system to solve and require a particular output or response. They are often used when the goal is to elicit a particular piece of information or to test the AI system's ability to perform a particular task. Examples might include: "What is the capital of France?" or "Write a program to sort a list of numbers."

3. Hybrid prompts: These are prompts that combine elements of both open-ended and closed-ended prompts. They may provide a specific task or problem and invite the AI system to generate a creative or novel response. Examples might include: "Write a short story about a character who travels back in time" or "Design a new species of plant that could survive on Mars."

4. Multi-step prompts: These are prompts that involve multiple stages or tasks and require the AI system to engage in a more complex and iterative process. They may involve multiple rounds of input and output and be used to simulate real-world scenarios that require planning and decision-making. Examples might include: "Plan a trip itinerary for a family vacation" or "Design a new product line for a company."

5. Subjective prompts: These are prompts that ask the AI system to provide a personal opinion or subjective interpretation. They may be used to explore the AI system's internal state or emotional responses or to evaluate its ability to understand and respond to nuanced ethical or moral questions. Examples might include: "Do you believe that robots have feelings?" or "How do you define consciousness?"

Best Practices for Prompt Engineering

Prompt engineering is relatively new, and best practices are still evolving. These practices aim to make your conversations with an AI as effective as possible. Creating a conversation effectively is something that can be perfected over different trials. However, here are some generally accepted best practices for prompt engineering:

1. Start with clear goals: Before creating a prompt, define what you want to achieve with the prompt. What do you want the AI system to do? What kind of information do you want?

2. Keep it simple: AI systems work best with simple, direct language. Avoid jargon, technical terms, or complicated sentences that may confuse the system. Use simple language and break down complex ideas into smaller, manageable parts.

3. Provide context: Context is essential for effective communication between humans and AI systems. Providing context helps the AI system understand the purpose of the prompt and the information it needs to provide. Ensure you give enough background information to orient the AI system and help it understand your question.

4. Test and refine: Like any other form of communication, prompts need to be tested and refined to ensure they're working effectively. Try different versions of your prompt, analyze the responses, and adjust accordingly. Refine your prompt until you get the desired result.

5. Consider diverse perspectives: AI systems can reflect the biases and prejudices of their training data. To mitigate these risks, consider various perspectives when creating prompts. Use multiple sources of information, and anticipate potential biases or flaws in the AI system's understanding.

All these practices aside, how about you stick around for a blog where we dive into the tried and tested techniques of Prompt Engineering, which will dive deeper into different ways that are being used by professionals in the field around the world to give every conversation of theirs with an AI some meaning and purpose.


In conclusion, prompt engineering is crucial to communicating with AI systems. By understanding the basics of prompt engineering and implementing best practices, we can improve the effectiveness and efficiency of our interactions with AI systems. Whether you're a developer, researcher, or simply someone curious about AI, knowing how to craft effective prompts can help you get the most out of these powerful machines.

As we continue to develop and refine AI technology, the importance of prompt engineering will only grow. With the rise of conversational AI and natural language processing, communicating clearly and effectively with AI systems is becoming increasingly critical. By mastering the art of prompt engineering, we can unlock the full potential of AI and create more sophisticated, intelligent, and valuable systems.

We hope this blog has provided a valuable introduction to the basics of prompt engineering and some practical tips for improving your skills in this area. Remember, prompt engineering is constantly evolving, so stay tuned for updates and developments in AI communication. Happy prompt engineering!

Written By

Aryan Kargwal

Data Evangelist

Copyright © 2023 NimbleBox, Inc.