Prompt Engineering 

Lately we have been hearing many artificial intelligence (large language models) names like ChatGPT, Gemini, Dal-E2, Midjourney, ArtBidder, StyleGen. Through these AI generators, we answer questions, search for information, draw pictures, create pictures of artificial faces, write applications or emails in formats we didn’t know before, etc.

Prompt design is a process by which we get the desired response or answer from large language models like ChatGPT. Writing effective prompts is essential to getting accurate, high-quality answers from any language model.

AI generators understand our input and create something close to the demand. But many times those answers do not fulfill our needs. Then the AI ​​generator has to ask more specific questions or prompts.

Prompt engineering works to determine how asking a question or prompt will get the closest answer. Prompt engineering is the process of experimenting with prompts.

You can start with an initial prompt in the “input field” or “command line” of ChatGPT or any generative AI, and then gradually change it until you get the output you like.

The matter may be clear from the following example.

An Example Of Making A Cake

Let’s say you want to make a chocolate cake for a loved one’s birthday. If you want, you can use ready-made cake-mix and add oil, eggs and milk to make a cake. Or you can make a cake by collecting all the ingredients yourself without any ready-made ingredients: flour, sugar, cocoa powder, baking powder, baking soda and salt.

Now, what if you add espresso powder or pink salt to the cake? Or if you use coconut milk or almond milk instead of cow’s milk? And if you put raspberry topping on the cake?

Each step will change the taste and characteristics of the cake, and you can get closer to the result you want.

This is how prompt engineering actually works.

You can start with an initial idea or prompt—like a ready-made cake-mix. You can then refine the concept, adding different ingredients or parameters—like changing the taste and properties of a cake.

Just like making a cake, you can repeat and refine the prompt until you get the results you want. Repeatedly changing the prompt to get the desired or required output is part of prompt engineering.

Generative AI

A branch of artificial intelligence is generative AI, also called artificial creative intelligence. 

Generative AI models are trained with large amounts of data, which teaches them to generate new data types similar to the training data. 

Let’s say a text-generating model is trained on the Books, Features and Websites dataset. Then when it answers the prompt, the new text it generates will be based on the dataset it received during training, albeit a completely new one.

Generative AI can create new data in the form of text, images, video, audio, and even code. It learns patterns in existing data using machine learning algorithms and then creates new data using those patterns, data that has never been seen before.

What Is Prompt Engineering?

Prompt engineering is the process of creating and improving prompts or instructions to get maximum performance from artificial intelligence (AI) and machine learning (ML) models. 

The main goal of prompt engineering is to present information or questions to the AI ​​model in such a way that the language model can provide accurate, relevant and actionable answers.

Prompt engineering is an important skill in communicating with artificial intelligence, helping to bring out the maximum potential of AI. 

Earlier, only programmers and AI experts wrote prompts using programming languages, but now anyone can easily create prompts using natural language (human language).

Examples Of Prompt Engineering

Let’s say you need to create a resume for a job. Due to not having a clear idea, you are seeking ChatGPT’s assistance in this regard. For this you can use a prompt like this in chatgpt’s input field:

“Sample Resume for a Marketing Analyst.”

ChatGPT will then present you with a sample titled “Marketing Analyst Resume Sample”.

You want the sample to be smaller. In that case you can ask chatgpt for feedback again. For example:

“Write the example in 100 words.”


“Make pattern language more formal.”

ChatGPT will then present you a sample marketing analyst resume tailored to your needs.

Similarly, if you want to create images, you can write prompts using Dal-e software. 

For example:

“Imagine working in the office of a professional marketing analyst.”

This way you can get the results you need by using specific and clear prompts. The most important thing in creating prompts is to minimize the distance between your question or request and the AI’s meaningful response. 

By crafting a perfect prompt, AI engineers can elicit responses from chatbots that are relevant to the user’s request and overall accurate. This reduces the need to redo the response or re-edit the image. 

Examples Of Adaptive Prompts

Engineers and researchers are now developing adaptive prompts that can adapt themselves according to the context of the prompt. 

An adaptive prompt is a prompt that initially begins with a simple instruction.  It is then gradually refined to be more detailed and specific to obtain specific results. It is basically a technique of prompt engineering that helps the user get the output they want.

Let’s say, you want to make a chocolate cake. Start with a simple prompt first and then gradually change and refine it. For example:

  1. Initial prompts: “Give me an easy chocolate cake recipe.”
  2. Modified prompt: “Give an easy chocolate cake recipe that uses ready-made cake-mix.”
  3. More specific prompts: “Here’s an easy chocolate cake recipe that uses ready-made cake-mix and adds espresso powder.”
  4. Specialized prompts: “Here’s an easy chocolate cake recipe that uses ready-made cake-mix and espresso powder, and uses coconut milk instead of cow’s milk.”
  5. Final prompt: “Here’s an easy chocolate cake recipe that uses ready-made cake-mix, espresso powder, coconut milk, and topped with a raspberry topping.” With adaptive prompts you start initially with a general idea and then continually modify and refine it to move toward specific and desired results. This process helps you step by step to get output according to your needs and preferences.

How Does Prompt Engineering Work?

Generative AI models are built using the Transformer architecture. Transformer architecture is a neural network design in artificial intelligence, particularly effective in natural or human language processing (NLP) tasks. 

This sophisticated architecture helps AI understand the nuances of language and process the vast amounts of data stored in neural networks. As a result artificial intelligence is able to provide a relevant and meaningful output. 

Prompt engineering plays an important role in making this output of the model more refined and relevant. Different prompting techniques are used to ensure that generative AI models provide maximum performance. These techniques help to better understand the model and generate relevant and meaningful feedback.

Why Prompt Engineering Is Important?

Prompt engineering helps AI engineers handle complex tasks like customer service or legal contracts with chatbots. 

With generative AI services like ChatGPT, engineers need to code and train the AI ​​with rich and accurate data to produce output.

We have reached a big-data driven world where AI models can extract results from large amounts of data much more efficiently than humans. Proper prompt engineering can detect and mitigate prompt injection attacks (attempts to hack the logical structure of a chatbot or chatgpt to harm it). This enables organizations to provide more consistent and accurate service.

Different Techniques Of Prompt Engineering

  1. Make the question or inquiry as clear as possible  Generative AI is a deep learning model, trained with data from humans and machines. AI doesn’t have the ability to search and figure out what you mean. The AI ​​will work on whatever you say. So the best way to pose a question or request to an AI generator is to present the question precisely in simple language, without using any kind of unnecessary filters.  For example, rather than the prompt “Create a writing structure with a title and action steps,” write something like “Create an academic research proposal structure that includes the title, summary, and next steps.”
  2. Experiment for better results You can write the same request in different ways, such as creating an executive summary, creating a research proposal, or creating bullet points in a CV. For example, you can specify whether the writing language will be official or formal, or informal. You can also vary your prompts. A sample or example can be given along with the prompt, so that the answer can be modeled on it.
  3. Follow up with further questions or instructions Once you can get the output of the prompt into the correct format and language, you can then instruct the text to be generated within the specified word count. Or you can create two separate versions of the text. The AI ​​generator can perform these tasks based on previous instructions given to it. Remember, repetition is your friend when it comes to AI. So keep doing prompt engineering until you get the results you want.
  1. Practice a variety of prompting techniques Generative AI is still a very new technology, which is a work in progress. That doesn’t mean researchers haven’t yet figured out a strategy for creating successful prompts. The more you work with generative AI, the more you can experiment with different prompting techniques. This will make it easier to get the answer you want from the chatbot. Here are some effective ways:

• Zero shot prompting—This is the most direct and simple way of prompt engineering.  Here the generative AI is simply given direct instructions, or questions.  No additional information is provided.  This technique works better for simple tasks than complex ones.

• Few Shot Prompting—This approach provides some examples as a guide to help generative AI.  This method works better than zero shot prompting for complex tasks.

• Chain of Thought Prompting—This method improves the performance of large language models by simply explaining complex logic.  As a result the model can respond more accurately.

• Prompt Chaining—Prompt writers divide a complex task into smaller tasks.  It then uses generative AI to complete the entire task.  This method sometimes works well for maintaining the reliability and consistency of complex tasks.

Here are just a few prompt techniques you can experiment with.  Sometimes you can combine several techniques to create one technique for the desired result.

Examples Of Using Prompt Engineering 

As generative AI reaches more people, organizations are finding ways to solve their everyday problems using prompt engineering in new and innovative ways.

• Chatbots—One can talk to AI chatbots just like our daily life conversations and get relevant and relevant answers to various queries from the chatbots. Prompt engineering works as a very powerful tool in this task. In order for the AI ​​to understand the user’s question and give a meaningful response, chatbot developers ensure that through prompt engineering.

• Healthcare—Prompt engineers in the healthcare sector can take help in summarizing and documenting medical data by instructing AI systems. With prompts, the AI ​​model analyzes patient data and provides well-informed feedback and advice.

• Software development—Prompt engineering plays a major role during software development by creating code snippets and solving complex programming problems. This helps developers work and saves a lot of time.

• Software engineering—Since generative AI systems are built with different programming languages, prompt engineers can create all kinds of code snippets here and make difficult tasks simple. By automating coding, debugging or debugging, or building API integrations, laborious tasks can now be done more easily. And by creating API-based workflows, resource utilization and data flow can be better managed.

• Cyber ​​security and computer science—Prompt engineering is used to test and strengthen security systems. Researchers and cyber security experts can test the success of their security strategies by artificially creating cyber attack scenarios using generative AI. Also prompt engineering is used to find software vulnerabilities.

A Few Words About Prompt Engineer Profession

Prompt engineering as a profession has a bright future. Prompt engineers must be proficient in the fundamentals of Natural Language Processing (NLP). These topics include various libraries and frameworks, Python programming language, various generative AI models. Apart from this, you have to work on various open source projects.

However, prompt engineers in most cases require a bachelor’s degree in computer science or a similar subject. Also from other backgrounds (writers or journalists) come prompt engineers who have no formal knowledge of technology, but have gained experience working and experimenting with AI.


Prompt engineering will continue to change in the era of artificial intelligence and machine learning. In the near future there will be prompts that can use text, images, and code. And as AI grows, we’ll see more prompts where integrity and transparency are effectively protected.

Leave a Comment