Prompt Engineering — Five Basic Principles

Lean how to interact with AI systems effectively

Note, full “Prompt Engineering” mind map is available at: “Prompt Engineering Mind Map”

In this article, I will talk about some simple but essential and effective principles when interacting with AI systems to create prompts.

1. Focus on Topic

First and foremost, it’s vital to comprehend the issue you’re seeking to address. Have a clear mind is very crucial when working with AI, you can just imagine that you are talking to a very smart but unexperienced person.

If your grasp of the problem is nebulous, anticipating that AI will provide a solution may be unrealistic. If you’re struggling to articulate or delineate the problem succinctly, it might be overly complex or sizable to handle. In such instances, breaking it down into manageable tasks could be beneficial.

At times, the outcomes produced by AI might deviate from your initial subject. It’s crucial to prevent such digressions and ensure the focus remains steadfast. Make sure you are as specific as possible when crafting your prompts, this will increase the possibility of getting better results.

Whenever you get stuck, one strategy to consider is soliciting suggestions from the AI system itself. The AI model might propose an approach that surpasses your current methodology.

2. Don’t Assume

Do not assume that AI knows the answer behind every of your question. Just because it has been trained on a huge amounts of data, does not necessarily mean that it knows everything.

Do not assume that AI understands the context automatically, sometimes it helps a lot to provide more context or even simple examples in your prompts. Keep in mind, the AI has limited context during an interaction and may not remember the previous interactions. Always check the returned output for validity, no matter how convincing it may look.

The problem with AI is that sometimes they give false results. This behavior is called “Hallucination”. Hallucination in the context of language models like ChatGPT refers to the generation of information that isn’t grounded in the provided…