Prompt engineering is a new and emerging field in the rapidly evolving Artificial Intelligence sector. A basic definition of Prompt Engineering is; a process that allows a user to guide a generative AI platform to complete desired tasks by asking questions, entered as Prompts using a platform prompting interface.

Although the generative platform can generate answers that sound as human as possible, it still requires well thought out lines of reasoning to generate valuable answer and results. So when entering Prompts using an interface, key phrases, appropriate formats and other guides are still needed to guide the system. Getting the exact desired results is still very much a trial and error that is driven by how close the entered questions (prompts) are to the desired results.

What is a Prompt?

In a nutshell, a Prompt is simply a natural language text that asks a generative AI platform, to create or perform a task. Prompts rely on LLMs (Large Language Models) that they have been trained on to answer and questions and perform task when asked by users.

Use Cases:

Subject matter Expertise.
Being that LLM’s are trained on vast amounts of Data, Prompt Engineering can be used to fetch and learn expertise on various subjects.

Critical Thinking and Creativity
Creativity and critical thinking is one area where Prompt engineering excels. This is so, because, LLMs, have access to huge amounts of already written text. Using a well structured prompt, a creative thinker can utilize a well crafted Prompt to formulate and grow creative ideas without plagirising the original content.

Important Prompt Engineering Techniques to Know:

Chain-of-thought:
Chain-of-Thought prompting is a prompting technique in which complex questions are broken down into smaller units.
As in humans, chain-of-thought is based on the knowledge that breaking tasks down into smaller units leads to better desired results. When using chain-of-thought method, it is advisable to rephrase the prompts in several forms and then choose the most reoccurring answer.
Tree-of-Thought:
Tree-of-thought is another prompting Techniques in which complex questions are asked of the LLM in what can be considered as, “leading” questions. The very first question is intentionally left open ended, and once an answer is given to it, the Prompt engineer then follows with additional questions. The given answers can then be "chained" or put together to get the desired solution. For example, a Tree-of-thought  prompt, could be, “What are the benefits of rainfall?” The answer would then be followed by, “list specific benefits to plant growing”?

The Basics of Writing Good Prompts.

Being that prompt engineering can easily be summed as “Asking questions and expecting clear answers from LLMs”, some guidelines are important to follow when phrasing and writing Prompts.
1. Unambiguous Prompts is to be avoided when writing effective prompts. Attempts should be made by the Prompt engineer to be as explicit as possible. For instance, a prompt that asks, “List the foods eaten in America”  is ambiguous. A more explicit Prompt could be, “a list of  twenty (replace the 20 with a number of your choice) eaten in North America.
2. Context: As in human interaction, context is important A prompt engineer should endevor to include as much context as possible when writing prompts. Using our food example again, adding the words “popular” adds context to what the expectations are. Foods, is too generalist, but, adding “popular” adds context to what is being sought.
3. Balance between simplicity and short sentences: When writing Prompts, endeavor to use short sentences that clearly zoom in on the points. Long winding sentences in most cases leads to hallucinations by AI LLM.

4. Finally, writing effective Prompts is a game of “trial and error” that in most instances requires many rewrites and experimentations to get the desired output. A prompt engineer should be prepared to write, rewrite and go through several iterations of the same Prompt till the desired results are achieved