In this week’s edition of ‘Down the AI Rabbit Hole,’ we explore structured prompting—a technique to enhance productivity with chatbot prompts. Additionally, I’ll share details about my new project this semester: the creation of an AI literacy overlay course.
This week’s tip:
One of the key distinctions between a generative AI chatbot, such as ChatGPT or Bard, and a Google search is the chatbot’s ability to remember your conversation in real time. This capability goes beyond a simple Google history, allowing you to engage in an actual “conversation” with the chatbot.
For instance, you might instruct it to “write an email to a student who was absent today.” After reviewing the response, you can request edits by adding details like “assume this student never misses class” or “note that this student often misses class.” The chatbot will then modify its response accordingly without needing a reminder that it’s drafting an email for an absent student.
This method is referred to as “unstructured or conversational prompting.” While effective, this approach can sometimes result in excessive or disorganized information. A more efficient method is “structured prompting,” where you provide the AI with carefully crafted and specific prompts to enhance the output’s quality. Dr. Phillipa Hardman discusses this technique in her blog post, Structured Prompting for Educators: Experiments in how to get the best out of your fave gen AI tools.’
Broader perspective: There and Back Again: A Mathematician’s Tale of AI Exploration
Like it or not, right now is the dumbest ChatGPT will be. That is, its ability to generate “correct or appropriate” responses will only get better. In late December 2022, ChatGPT reported that a spoon would win a race against a turtle. This error was rectified by January 2023, with the updated response acknowledging that a ‘spoon is an inanimate object’. To truly “inspire and educate our students to become autonomous thinkers, discerning moral agents and active citizens,” we must accompany them on their journey into an AI-driven world.
Last century, when I was a college student, we had a course called Computer Literacy where we learned which slot to insert a thin, 5 1/4 square piece of plastic, and how to retrieve a typed paper. In today’s AI-driven world, there’s a pressing need for an ‘AI Literacy’ course. This course should not only consider the uses and capabilities of generative AI but also explore its ethical, philosophical, and sociological implications. Key questions to address include: How do we mitigate its biases? What are its environmental impacts? And what are the implications for ownership and copyright?
Unfortunately, no such course exists. That is why this semester, I have decided to devote one day per week of my four-day-a-week essentials of calculus class to developing an AI literacy overlay. Together, my students and I will actively explore generative AI; we will discern its uses and implications. I will chronicle our journey in a column for the Mathematical Association of America: There and Back Again: A Mathematician’s Tale of AI Exploration. I invite you join me on this unexpected journey!