Leveraging Prompts with Generative AI

What are prompts?

In the context of a generative AI system like ChatGPT, a “prompt” refers to the input or instruction given to the AI to generate a specific response or output. The prompt is essentially the text or command that sets the context and provides guidance to the AI model for generating text-based responses.

A prompt can take various forms, including a question, a sentence, a paragraph, or even a code snippet, depending on the application. The quality and specificity of the prompt are essential in influencing the nature and relevance of the AI-generated response. A well-crafted prompt helps steer the AI model toward providing useful and coherent information or generating content that aligns with the user’s intentions.

(Written by ChatGPT. Prompt: Can you provide me a definition of what a prompt for a generative AI system is?)

Types of prompts

The prompts provided in the attachment and the topics listed below were developed through collaboration between myself, ChatGPT (providing prompts), and Claude (proposing topics). I utilized two distinct models to demonstrate that the responses you receive are influenced by the prompts you provide, not solely by the system. By using appropriate prompts, these models can assist you in generating high-quality content. Please download the following prompt sheet.

You can download examples of the following prompt topics here:

Generation – These prompts help create new course content from scratch. For example, you could use a prompt to generate a draft syllabus outline for an Environmental Science class, including sections on course policies, schedule, assignments, and grading. Or you could ask for a list of thought-provoking discussion questions about Climate Change for a faculty forum.

Enhancement – These prompts strengthen existing content. You could provide a product description for wireless headphones and ask for suggestions to expand on details about sound quality, battery life, and compatibility. Or you could summarize a long research report on market trends into a one-page executive brief highlighting key takeaways.

Q&A – These prompts allow you to ask questions to gain new knowledge. You could ask for an overview of machine learning and how it applies in diverse fields like healthcare and finance. Or you could ask for insights on experiential learning strategies to incorporate into a History of Science curriculum. The model can research trends and recommend innovative technology tools for specific disciplines.

(Co-written by Claude [https://claude.ai/chats]. Prompt: The document [Faculty Prompt Sheet] I just uploaded is a collection of recommended prompts for faculty to use to best leverage the affordance of generative AI for course design. Using the recommended prompts under each topic, can you write a short description for each topic?)

Ways to improve prompts 123.

The previous text was a collaborative effort between myself, ChatGPT, and Claude AI. I shared my initial prompts and the resulting output is the outcome of our iterative process. To demonstrate how to iterate with a generative AI, I have included my conversation with Claude to further clarify the topics I obtained from ChatGPT.

Conversation with Claude transcript:

Remember, prompts are about establishing context, rules, and output expectations. Also, remember to be concise and clear. The prompts in the AI Prompt Sheet (see the download above) are meant to be starting points for you. Try them in ChatGPT, Claude AI, or Bard. Even if you like the output for the default prompts, consider changing it to see what it says.

For example, the first prompt in the sheet “Create an introductory paragraph for a course on ‘Modern Theories of Physics’ for faculty members.” could be changed to say Create a 5-sentence introduction paragraph for the course ‘Modern Theories of Physics’ and be sure to include the following…” then include your course outcomes and expectations. This will help tailor the output to your course further.

When altering your prompts, think about what the chat is doing that you want changed.

Is the output missing vital data? Reread your prompts for context and see if the chat has access to the correct prior knowledge. Explain some more back story. For reference check the transcript between myself and Claude. Look at how much I tell it in the first prompt.

Is your output in a structure you don’t want? Think about the expectations you set for the model. Let it know if you want numbered lists or not. Give it an article or paper you wrote so it can try to emulate your voice. Tell it to put your output into a table if you need a large body of text with numerical data.

Finally, consider changing your prompt. Instead of a command or a basic question, give it a task. Ask it to finish a sentence you’re struggling with. For example,

            “Please finish the following sentence, generative AI is best used to…”

You can view Claude’s response to this prompt here:

Since generative AI outputs are based on complex algorithms, changing the input slightly can provide you with new results.

It’s important to explore AI and see how it can benefit you. If you’d like more information on AI use at FHSU, please see the following resources:

Packback and AI: https://tigerlearn.fhsu.edu/packback-launches-teach-with-gpt/

Ten Commandments of Instructional Design According to ChatGPT: https://tigerlearn.fhsu.edu/ten-commandments-of-instructional-design-according-to-chatgpt/

Also, please visit Forsyth Library’s Generative AI Guide:  https://fhsuguides.fhsu.edu/ai

Resources:

  1. https://arxiv.org/abs/2304.14670 ↩︎
  2. https://www.zdnet.com/article/how-to-write-better-chatgpt-prompts/ ↩︎
  3. https://www.sciencedirect.com/science/article/abs/pii/S0099133323000599 ↩︎

Be the first to comment

Leave a Reply

Your email address will not be published.


*


This site uses Akismet to reduce spam. Learn how your comment data is processed.