Building AI Applications with Microsoft Semantic Kernel

2

Creating Better Prompts

As a developer, you can request that an LLM completes a task by submitting a prompt to it. In the previous chapter, we saw some examples of prompts, such as “Tell me a knock-knock joke” and “What is the flight duration between New York City and Rio de Janeiro?” As LLMs became more powerful, the tasks that they could accomplish became more complex.

Researchers discovered that using different techniques to build prompts yielded vastly different results. The process of crafting prompts that improve the likelihood of getting the desired answer is called prompt engineering, and the value of creating better prompts gave birth to a new profession: prompt engineer. This is someone who doesn’t need to know how to code in any ...