Exercises to Implement Length and Format Control

Exercise 1: Explicit Format Control

Design a prompt that asks GPT to write a haiku about artificial intelligence. A haiku is a form of traditional Japanese poetry consisting of three lines with a 5-7-5 syllable pattern.

Exercise 2: Implicit Length Control

Create a prompt that encourages GPT to generate a succinct summary of the Turing Test, as if it was for a tweet (Twitter's character limit is 280 characters).

Exercise 3: Format Control Using Lists

Compose a prompt that instructs GPT to list and briefly explain the top three principles of sustainable development.

Exercise 4: Using Follow-up Prompts to Limit Length

Design a two-part prompt about global warming. The first should ask for a detailed explanation, and the second should ask for a simplified summary suitable for elementary school students.

Exercise 5: Leveraging System-Level Prompts

Create a system-level prompt that instructs GPT to respond in the style of a telegram message to explain the concept of machine learning.

Remember, the goal of these exercises is to help you better understand how to control the length and format of GPT outputs through the careful crafting of your prompts. It's normal to experience some trial and error before you get it right, so don't hesitate to refine your prompts as needed based on the outputs you observe.

Master Level Prompt

Complex Policy Analysis Simulation

This prompt is designed for users with a high level of experience in GPT-4 prompt engineering, who wish to use GPT-4's capabilities to simulate complex scenarios and generate detailed analyses.

GPT-4, let's perform a multi-layered scenario simulation. Imagine you are the {Secretary of Energy} in the year {2030}. There's an ongoing debate about transitioning to {100% renewable energy}. Your task is to:

1. Evaluate the current {energy} landscape,

2. Propose a comprehensive plan for this transition, considering economic, technological, and sociopolitical factors,

3. Anticipate potential obstacles and suggest ways to overcome them,

4. Discuss the impacts of your plan on the {U.S. economy}, the {global climate}, and {domestic and international politics}.

This prompt is rated as master level due to the following reasons:

Multi-Dimensional Analysis: The prompt asks GPT to consider multiple layers of analysis - economic, technological, and sociopolitical. Each of these areas requires a separate deep dive into specific aspects and an understanding of their interconnectedness.

Long-term Memory and Contextual Understanding: GPT must remember and utilize the input scenario's constraints throughout the response and build upon them to create a comprehensive policy plan. This requires leveraging GPT's ability to maintain context over longer discussions.

Scenario Planning: The prompt asks GPT to anticipate potential obstacles and suggest ways to overcome them. This requires GPT to simulate potential future scenarios, adding an additional layer of complexity to the prompt.

Complexity of Subject Matter: The subject of energy policy is a complex one that requires nuanced understanding and handling. The prompt demands an in-depth exploration of this topic.

Exercises to Manipulate Context in Prompt Design

Now that you've seen the power of context manipulation in action, it's your turn to practice. Here are a series of exercises designed to help you master this crucial aspect of prompt design.

Exercise 1: Topic Switching

Craft a sequence of prompts that guides GPT-4 to switch between unrelated topics. Start with discussing a novel, then shift to a cooking recipe, and finally to a space mission. Ensure each transition is smooth and doesn't seem abrupt.

Exercise 2: Context Retention

Design a conversation where you ask GPT-4 for a movie recommendation based on specific criteria (e.g., a comedy released in the last five years that stars a specific actor). Then, pivot the conversation to discuss the film's director and their other works. GPT-4 should remember the previous film recommendation and use it to inform its responses.

Exercise 3: Multiple User Inputs

Create a scenario where multiple user inputs are required to reach a specific outcome. For example, troubleshooting a technical issue. The AI should keep track of the user's problem and responses, guiding them step by step to the solution.

Exercise 4: Ambiguity Resolution

Write a prompt where the initial user statement is ambiguous, requiring GPT-4 to ask clarifying questions. After the user provides more details, GPT-4 should provide a well-informed answer.

Remember, manipulating context in prompt design is an art. You need to guide the model while maintaining a natural and coherent conversation. The best way to perfect this skill is by experimenting with different scenarios and observing how the AI responds. Keep practicing and tweaking your prompts, and over time, you'll see significant improvements in the quality of your interactions with GPT.