Introduction to Week 3
This week we’ll begin prompt engineering – the process of writing, testing, and refining prompts in order to elicit the answers or behavior you desire from a generative AI model.
Weekly Activity: Part 1
This week you’ll spend some time reading and then a lot of time playing. Begin by reading about the basics of writing effective prompts.
Selections from Prompt Engineering Guide by Democratizing Artificial Intelligence Research, Education, and Technologies (DAIR.AI):
Teo, S. (2023). How I Won Singapore’s GPT-4 Prompt Engineering Competition.
OpenAI. (2023). Guide to Prompt Engineering.
Weekly Activity: Part 2
For this week’s activity you’re going to write, test, and refine three prompts. These prompts can do anything you like, as long as it’s non-trivial (e.g., not “Write a haiku”). They can be on three different topics or they can all be on a similar topic. You can create prompts related to a hobby, an interest, your dissertation topic, your job, or anything else.
Have fun with this assignment. Create prompts that will delight and surprise your friends or relatives. Write prompts that generate outputs that will get them to say, “I didn’t know computers could do that!”
Employ strategies from this week’s readings in your prompts. As I read your prompts, there should be clear evidence that you’re employing techniques covered in this week’s readings.
Keep track of all your experimentation with each prompt, from your first attempt to your last. You won’t get what you’re hoping for with the first version of your prompt. Iterate, experiment, and try different techniques to get the results you’re hoping for. Keep each version of the prompts you write and the output you receive for each. Combine all the experimentation from the three prompts into one document, and submit it to Assignment 3: LLMs in Canvas. Alternately, if you’re using ChatGPT Plus, submit links to the three conversations in which you developed your prompts. This assignment is due by 11:59pm Mountain on Thursday, Jan 25.