AI Blog: AI Prompt Engineering Part 1

Welcome back to our AI blog. Today, we will talk about prompt engineering, which can help you effectively interact with AI tools and get the expected results.

What is prompting?

If you ask ChatGPT, it will give you the answer as follows:

The reason why effective prompt engineering is important is that AI language models, including ChatGPT, Gemini, Claude and so on, generate responses based upon the input prompts they received.  The more accurate, detailed and well-structured prompts are, the more precise and relevant responses will be.

But how can we identify if a prompt is effective?  How can we improve our input to get the expected output?  First of all, we need to understand what kind of elements make a good prompt.

Five-Step Framework

Task: Clearly define what you want AI to do

For example, you can ask ChatGPT: “Give me suggestions to improve my oral English”. 

You will get a long response that provides practical suggestions on daily speaking, listening and repeating, expanding vocabulary, etc.

The output looks acceptable, but it is still very general, so you will probably require a more specific, tailored response.  You may elevate the output by using two [2] more elements.  The first of these is a persona, which means assigning a role to the AI model to instruct it to behave as a particular character or expert.

For example, you can edit the previous prompt to act as an English teacher to make an oral English improvement plan for you.

We can see that the results are much more specific than before, with finer granularity and a clearer breakdown into sections.

The second element to improve the response is to identify the format of the output.  The default format is a list and bullet points: you can ask ChatGPT to give you a scheduled timetable so you can arrange the learning tasks day by day, e.g.

Now you have a more structured response that is easier to follow, you can start the English learning journey right now.