A fictional persona representing a specific group or expertise. In most cases, the chosen persona is an expert in a specific field, but personas can also be customer personas, stakeholders like employees or shareholders, and more.
How to create an effective persona:
- Define the persona's characteristics & personality
- Create a detailed profile
- Use real data
- Test your persona
- Update your persona
A well-defined task description helps the model to understand what it needs to do improving the accuracy and effectiveness of the model's output.
How to describe a task effectively?
- Be specific
- Provide examples
- Use clear and simple language
- Define the scope of the task
- Define the objective of the task
3. Task Execution
The task description should be accompanied by detailed step-by-step instructions on how to complete the task. This helps keep the model's response organized and tailored toward your goal
How to describe task execution:
- Be specific
- Describe each step in detail
- instruct chatgpt to show interim results or suppress them
- use clear and simple language
4. Context & Prompt Injection
Providing sufficient context is a great way to improve the accuracy of the LLM's output. If you are not sure what context to provide use prompt injection.
- Again, be specific
- Use real data when possible
- Test different contexts and prompt injections
- Keep it simple
- only provide context that aligns with the task
5. Constraints:
Constraints are an essential aspect of the ChatGPT prompt construction process. They help to define the boundaries for the model's output, ensuring that it stays within the parameters of the task at hand.
How to set constraints effectively?
- Be specific (yes, again)
- Test different constraints
- Align with the task
- Be flexible
- Communicate clearly
Examples include limiting output length, setting tone and style, restricting certain words or setting formality
6. Define Goals
Goal setting helps ensure that the model's output is aligned with the task at hand. By setting a clear goal, you can ensure that the model's responses are appropriate for the project and that they meet the specific needs of the task.
How to set goals effectively?
- Be specific (yes, you guessed it)
- Align with the task
- Set measurable goals
- Be realistic
- Communicate clearly
7. Output Format
The output format is the final aspect to consider when working with ChatGPT. It is essential to ensure that the model's output is in the right format for the task at hand and that it is easily readable and understandable for the intended audience and/or use.
How to format output effectively?
- Align with the task
- Keep it simple
- Use visual elements (emojis, tables, ect)
- Test different formats
- Communicate clearly
Examples:
- Q&A format
- Table
- XML
- Conversational
Putting it all together
We have created a 26 page example with different advanced marketing-related prompts following this guide but the prompts are too long for tweets. So here are the examples:
The big breakthrough chatgpt brought to the market is its ability to respond to natural language prompts to generate amazing results. Yet, our first instinct is to formalize and standardize prompting. Is that the right approach?
Don't get me wrong the prompt construction suggestion from @thatroblennon is pretty solid. The only thing I would add is prompt injection as explained here:
@thatroblennon But the fundamental question remains: Is it really necessary to formalize/standardize prompting language to achieve better results? Does it really matter?
I will be conducting some research on this and share the results.
Maybe we can crowdsource something better. Let's do a contest!
Paying $150 for the best background image and profile picture!
Rules 👇
1. Background Image & profile picture must be created in midjourney 2. You must post the full-resolution images as a response to this thread with the prompt used 3. the background image should inspire "10x your productivity" 4. Winner will be chosen on January 25th EoD
Tagging a few midjourney giga chads and artists we like for reach.
🤯 100 years from now: “Imagine a futuristic world like Warhammer where gremlins meets Starship Enterprise. Jumping Mount Doom on a light bike with family by your side.”
🚀 The real holodeck is not just a virtual world, it's a world built on prompts and on-demand! Brace yourself for new products and possibilities.
💡 AI generated ideas will revolutionize many markets. But it all starts with how we describe what we want. We will be able to conjure worlds merely with our words.