Since ChatGPT was launched on 30th November 2022, there has been a massive shift in the technology world especially in the department of AI. Today we are using most of these AI tools by prompting them with every weird stuff that’s possible. But you must have noticed that most of us fail to get the expected result with these AI tools, at least by a decent margin in a few initial attempts. In this article, we will be talking about how can we improve our prompting with AI but before that let’s discuss a few basic stuff for all the newbies.
What is AI Prompting?
As of now, the only way to communicate with AI tools such as ChatGPT, CoPilot, Gemini is by prompting. Prompting is nothing but explaining or questioning the AI tool in the form of text that you are looking for.
Here are a few examples of an AI Prompt
- Give me the list of 10 healthy fruits
- Tell me a song by Mohit Chauhan
Yes, it’s similar to a search query that we put in Google Search or Bing but there are a few differences.
Google Search works by giving you the search results in the form of a list of results from different sources. But AI tools like ChatGPT understand the intent of the prompt and directly give us solutions that are customized to our needs.
How to Prompt AI Tools like an Expert?
When you give a prompt such as ‘list 10 healthy fruits’, you are expecting AI to come up with 10 healthy fruits so this will meet your expectations. But when you ask AI to write a resume for you, it will ask you to fill in various fields such as your name, Address, Phone number, experience, Education and Projects. Even if you fill in those details, the Resume that the AI tool will give you back will not be impressive and will not land you any jobs in today’s time.
So what’s missing in Prompt? Or the AI tool is not that great?
You need to understand a very simple thing about all of these generative AI tools which is ‘AI tools are like a child that knows everything still you have to tell him everything’.
Why do you need to tell it everything? Why don’t they understand what Aman wants?
That’s because they can’t read my mind yet. The only way for me to explain what’s in my mind is through Prompt. So my prompt should reflect my expectations and my thinking.
What make a Good AI Prompt?
There are no specific rules around a good prompt or a bad prompt. In general, a Good Prompt will be something that gives you more than 50% of your expected result, while a bad prompt will give less than 50% of your expected result.
A Good prompt should have the following
Context (Who are you?)
Context resembles Who are you, which means we tell AI tools who they are, without the context it will always give generic outcomes. It plays a very crucial role if you want the results to be more relevant to what you have in your mind. You can give context such as
“Write a resume for a software engineer”
“Write an article as a professional writer of Washington Post”
“Think it like a person who is been to Japan, and suggest me…”
All of these examples tell AI tools to establish a mindset before generating a result.
End Result (What needs to be done?)
We typically specify the end result when using AI tools; otherwise, they might not generate any output. It is the most important part of a prompt. We can enhance our results by providing more detailed instructions.
“For scene, write an engaging script…”
“Recommend red-colored vegetables.”
“Provide facts about vegetables that are red.”
Conditions and Constraints (What to take care of?)
Conditions and constraints in AI prompts are like rules or limitations that guide how you ask questions or give instructions to the AI. They help ensure that the AI responds in a way that meets specific requirements or stays within certain boundaries. It filters out the output more towards the expected result.
Imagine you’re asking a friend to do something for you, but you have to give them specific instructions. For example, if you ask them to bake a cake, you might give them conditions like “use only chocolate as the flavour” or constraints like “finish it within an hour.” These conditions and constraints help your friend understand exactly what you want and how to do it.
Similarly, when you’re interacting with AI, you might set conditions like “generate a story about unicorns” or constraints like “keep it under 500 words.” These guidelines help the AI know what you’re looking for and how to tailor its response accordingly.
Hope you liked this article on Prompting Generative AI. Checkout Tech Space for more such informative article.
Also read