The Set Up
Today, I’m offering some results of working with Learning AI Prompt Engineering. Last week, I mentioned that I was going to start taking some time each week to learn a new aspect of AI. I’m starting with Prompt Engineering simply because it was the first item on my list. I took this short course through a free trial on Coursera titled: “Start Writing Prompts Like A Pro.” Since I only had seven days for my trial, it added an aspect of urgency to complete it.
There are four courses required to get a certificate in this. I’ll be discussing the first course: Learning Prompt Engineering. It’s broken into four main sections.
Start writing AI Prompts like a pro
The primary purpose of this course was to emphasize a five-step formula for writing prompts for Generative AI purposes. It uses the acronym: T. C. R. E. I. It uses two ways to define the acronym.
| Task | Thoughtfully |
|---|---|
| Context | Create |
| Reference | Really |
| Evaluate | Excellent |
| Iteration | Inputs |
The course works best by using a Generative AI platform such as ChatGPT, Genesis, or Claude. There are many others, but these are generally free, free for a trial period, or low-cost.
Task/Thoughtfully: This is the first part of the process, and defines what you want Generative AI to do. It could be to create an outline, organize data, create a timeline, create a schedule, or even a graphic.
Context/Create: Explain the situation. Do you need to create a timeline for employees to complete their tasks? If so, what are the roles of the employees, their experience, what tasks are needed, and when should they be done by?
Reference/Really: This refers to any information you can provide within your AI Prompt to help develop the solution. It could be articles, videos, previous examples, resumes of employees, websites, articles, or other information similar to what you want produced.
Check with your employer about putting personal information into a generative AI system.
Submitting Your AI Prompt
At this point in the AI prompt engineering process, you submit it for the first time. Keep in mind that the response you receive may not come back exactly as you’d like. The point of Learning Prompt Engineering is to take your prompts through all the steps. It can take several attempts to get the information you want in the way you want it.
Evaluate/Excellent: You have your response from AI. More than likely, it won’t be exactly what you’re looking for. Take the time to evaluate what’s been provided. Ask yourself what’s missing. One of the great things about using AI; it remembers the last thing you were doing. It’s not necessary type in the entire prompt and add additional information. You can continue from where you left off (assuming you haven’t cleared the AI settings).
Iteration/Inputs: Iteration is the process of evaluating the last AI prompt response, determining what else is missing, and adding additional information to AI for another response.
AI Prompt Engineering Watch Outs
Have you ever had hallucinations? Believe it or not, AI can, at times, have hallucinations. What?
Yep. If your engineered prompt is too vague, AI can produce information that doesn’t make sense. It’s important to verify any information AI provides, for accuracy and relevance. If you find that the answers you’re getting are inaccurate or irrelevant to what you’re seeking, try clearing out the AI cache or closing it out and starting again.
Summary
The course did a good job of defining terms, providing examples, and offering exercises. Additionally, there’s a quiz after each section, confirming the knowledge you’ve learned. I recommend the course from Coursiva for learning this process.
If you or your organization need help figuring out how to create prompts for your specific needs, reach out. I’m able to help get you started.