Module 4: Writing better prompts — AI is just a tool. I'll show you how to write prompts that actually get you what you want.
- Scenario: You hand the AI a task with no examples at all, forcing it to guess the format based on its base training.
- Zero-Shot — You are giving it zero previous references.
- Scenario: You literally show the AI three examples of how to format names before asking it to format a brand new name.
- Few-Shot — You are giving it a few specific examples to mimic.
- Scenario: You explicitly ask the AI to write out its logic sequentially to prevent math or reasoning errors.
- Chain of Thought — You are forcing it to show its train of reasoning.
- Scenario: You ask the AI to first write an outline, review it, and then in a separate step expand it into a full email.
- Chain of Draft — You are building the content in iterative steps, starting with a rough version.
- Scenario: The default way most people interact with ChatGPT, relying entirely on the model to fill in the blanks.
- Zero-Shot — If you don't use any advanced techniques, you are automatically doing this.
- Scenario: Best used when you need the AI output to rigidly mimic a specific style (like extracting CSV data) by giving it previous examples.
- Few-Shot — It's all about mimicking patterns you provide.
- Scenario: Adding the specific phrase "Let's think step by step" to the end of a complex logic problem.
- Chain of Thought — This forces the math model to generate intermediate steps.
- Scenario: Generating a rough bulleted list first, and then manually asking the AI to turn it into a full 1,000-word blog post.
- Chain of Draft — Breaking a massive writing task into smaller, reviewable chunks.
- Scenario: You perfectly demonstrate 'Input -> Expected Output' twice before giving the AI the real task.
- Few-Shot — You provided a specific number of examples to establish a pattern.