Prompt Engineering

Over the past decade, our ability to search the internet effectively has greatly improved. In the early days of Google, we often relied on simple keywords, hoping to find relevant information. However, we’ve since learned to refine our questions to yield better results, whether we’re searching on Google or asking Siri for restaurant recommendations.

This evolution in searching parallels the concept of prompt engineering. Prompt engineering involves designing and optimizing inputs, or prompts, to guide AI models in generating the desired outputs. This practice enhances the accuracy and relevance of responses from generative AI models.

How to get better answers from Artificial Intelligence (AI)

Prompt engineers blend creativity with trial and error to craft prompts that provide AI systems with the necessary context, instructions, and examples. This approach helps the model grasp the user’s intent and respond meaningfully. Recently, I attended a lecture by my colleague Jaco Prinsloo , who explored this concept even further.

“...the words we use, the way we structure our sentences, the way we talk that might cause confusion and sometimes that needs clarification. Now in the case of prompt engineering and communicating with or chatting to LLMs or ChatGPT, that's really restructuring or finding a way to ask the questions so that you get a proper answer or maybe a better answer than you would have received otherwise. And typically this is just a trial and error process.

It's identifying or discovering patterns that work with a specific LLM. It's kind of asking a question, seeing what type of response you get, iterating on that, maybe providing more context, providing less context, giving examples, anything like that. I think for the attendees on the call, if you've ever asked a question of ChatGPT or a similar system and you weren't satisfied with the answer you received and then you went back and you kind of tweaked your question a bit, exactly what you did was prompt engineering.”

Mastering prompt engineering

I enjoyed this example from Prompt Engineering Examples, Techniques, and Practical Applications ?

To make the best use of Gen AI, you need to master prompt engineering.? You wouldn’t call a pizzeria and place an order for a “pizza.” To receive the dinner you desire, you’d specify your choice of crust, toppings, spices, accompanying drink, and anything you might be allergic to.? The Moral of that story is: The more detailed your instructions, the closer the pizza is to your preference.

Do you recall the days when you kept a cheatsheet on your desktop for commonly used shortcuts, like copy and paste (Ctrl C + Ctrl V) or Excel formulas? Today, my version of a cheatsheet consists of prompt shortcuts I utilize while working with AI. These are concise instructions I incorporate into each prompt to enhance the quality of the responses I receive. For example, I often include a specific phrase at the end of my prompts to achieve better results as shown below.

11 Prompt Engineering Ideas

The internet is full of great reference articles on how to craft better prompts including this list below from Unleash the Power of ChatGPT: 11 Epic Prompt Engineering Tips! .??

1. Be Clear and Specific
2. Provide Context
3. Phrase Your Questions Carefully
4. Anchor Your Prompts with Examples
5. Constrain ChatGPT
6. Break Things Down Step-by-Step
7. Encourage Multiple Alternatives
8. Experiment with Prompt Length
9. Iterative Prompting
10. Review and Edit
11. Test and Iterate

Give it a try and see which prompts work best for you. Also please feel free to comment with your favorite prompts that you use everyday.