Prompt Engineering – Get better results with ChatGPT

Over the last couple of weeks, I’ve been studying and trying to level up my own skills in using the latest AI tools and especially ChatGPT. That’s when I learnt deeper about “Prompt Engineering” and how important it is to use appropriate prompts to tweak the results of large AI language models.

A lot of people look at ChatGPT as like it is a kind of magical box which gives us fascinating human-like output. But in a way you should look at it like a calculator for words. This is a tool and the output that you get, or the quality of that output is very dependent on the input that you key in.

You know, it looks like prompting was not a feature designed by AI experts. Rather, it was an emergent feature out of these huge ML language models. Prompting became the way to have the ML models execute the input.

By learning how to prompt the machine with better input and tell the system what you want and how you want it, you can get better output which is more aligned with what you’re looking for. When used right it can be like having an extra employee in your office. ChatGPT can do a multitude of tasks better than humans.

The goal of prompt engineering is to guide the model to generate text that is relevant, coherent, and consistent with the intended purpose of the output, such as a news article, creative fiction, or a technical document. Let’s start by looking at a very simple example and slowly tune the input to achieve better results. Let me ask ChatGPT to write as a sales letter with the following prompt “Write a sales letter for a SAAS product that edits video using AI”.

Now you shall get a response that is fairly decent. It’s a good cohesive letter you could use straight out. But let’s see how you can make it better.

So, the first thing I want to do is add a topic and give a Persona to ChatGPT. Like “You are a professional copywriter” and then I’m asking exactly what I expect it to write.

Create one page sales letter about the topic above using these strategies:

Use strong persuasive language.
Use short sentences and simple terms, make it easy to read.
Ask questions to transition between paragraphs.
Back up key points with figures, evidences and examples
Speak directly to the reader, make it personal.

Now these strategies are fairly standard copywriting practices. The final thing, I wanted to add at the end is a call to action. I want ChatGPT to know the purpose of this sales letter and what we’re trying to do. So, I write:

“The goal and call to action for this content is to: Sign up for the newsletter”.

Now, once you provide this content to ChatGPT, see how it compares with the previous result. You would be able to see, by just giving ChatGPT more information about how we wanted it to write the content, the length of the content and what we’re trying to do with – the piece of output we’re getting is much more compelling.

It is important to get yourself better in prompt engineering to use such Large Language Models. To do so, you can experiment yourself with different prompts and observe the model’s output. This will give you a sense of the model’s capabilities and limitations, and help you understand how to craft prompts that lead to the desired output. It requires you to constantly experiment, test and refine your input to improve yourself. Nevertheless, these AI models are constantly evolving, with new methods and techniques being developed all the time.

Thanks for Reading. Stay Tuned!