Poetry

AI 101

Background

Prompt Engineering

  • We recall a few tips we learned to write better prompts.

Elements

  • Format
  • Context
  • Example
  • Tuning
  • Turns

Types

  • Direct
  • \(n\)-shot
  • Chain-of-thought
  • (Hybrid)

Strategies

  • Clear Goals
  • Context
  • Iterate
  • Avoid Pitfalls

Iterate

  • Experiment with different prompt variations
  • Observe the model’s responses
  • Refine your prompts based on the results.

Your Task

  • Compete with human authors to generate poetry.

The Poem

  • I wrote the following work as an example.
    • We needed something original.
    • We needed something written without assistance of AI.
      • I think I used a thesaurus for when writing line 3.
    • We needed something that we own the rights for.
      • I own the rights for this, since I wrote it.
Wait!

Do not provide this poem, in whole or in part, to any LLM, such as ChatGPT or Gemini.

It will ruin the experimental design of the assignment.

  • Once you have read the warning, feel free to expand below to see the poem.

Outcomes

  • Your responsibility is to use prompting to get Gemini to generate this poem.
    • Or something indistinguishable from this poem.
  • Consider aspects of the poem such as tone, content, theme, rhyme scheme, meter, and more.
    • If you are unfamiliar with these concepts, you can use the methods of choice, including LLMs, to study a similar poem.

Example

Final Product

  • You should have a Colab document which:
    • Contains the iterate process by which you refine your prompt.
    • Culminates in a generative prompt creating an indistinguishable poem.
    • Provides no part of my original poem, in whole or part, as a component of any prompt.
    • After the generated poem, answer in 2-3 sentences in their own text cell:
      • Which strategies you found helpful, or unhelpful.
      • How intelligent or original you found Gemini.
      • Whether you found the process or prompting easier or harder than writing your own originals.