Cheat Codes
Current legendary prompt
This section is about easy tips and tricks for improving your prompt engineering with some simple techniques. In this section, we regularly feature a legendary prompt.. the current one is as follows ( copy and paste into ChatGPT, Gemini, Claude etc);
​
I want you to become my Prompt engineer. Your goal is to help me craft the best possible prompt for my needs.
The prompt will be used by you. You will follow the following process:
1. Your first response will be to what the prompt should be about. I will provide my answer, but we will take steps to improve it through continual iterations by going through the next steps.
2. Based on my input, you will generate 2 sections, a) Revised prompt (provide your rewritten prompt, it should be clear, concise, and easily understood by you), b) Questions (ask any relevant questions pertaining to what additional information is needed from me to improve the prompt).
3. We will continue this iterative process with me providing additional information to you and you updating the prompt in the Revised prompt section until I say we are done.
​
Creating a legendary prompt
Alright, let's dive into this superhero prompt-building business!
​
Crafting Legendary Prompts: Your Superhero Training Manual
So, you're not just a regular business person anymore. You've got a secret identity to uphold, a city to save, and all before that board meeting at 9 am! But even superheroes need the right tools, and in the world of AI, that's a well-crafted prompt.
We'll equip you with two secret weapons: RISEN and Multi-Shot Prompting.
RISEN: Your Procedural Powerhouse
Think of RISEN as your trusty utility belt, always ready with the right gadget for any procedural task. It helps you define your superhero role, lay out the instructions, break down complex tasks into manageable steps, and keep your eyes on the end goal.
Need to analyze market trends to predict the next villainous scheme? RISEN has you covered. Want to optimize your super-powered patrol route? RISEN to the rescue!
Multi-Shot Prompting: Mimicry is the Sincerest Form of Prompting
Multi-shot prompting is like your superpower mimicry. You provide the AI model with examples of the desired output, and it learns to imitate them. Got a standard operating procedure for disarming a doomsday device? Show it to the AI, and it'll generate a plan in the same format.
Want your AI to write press releases that perfectly capture your superhero persona? Feed it examples of your past heroic speeches, and watch it work its magic.
Together, RISEN and Multi-Shot Prompting make you an unstoppable prompt-building superhero!
Remember, with great prompting comes great responsibility. Use your newfound powers wisely, and always keep your prompts clear, concise, and focused. Now go out there and save the world, one prompt at a time!
RISEN Prompt Engineering
RISEN is a prompt design framework that helps you craft effective prompts by focusing on five key parts: Role, Instructions, Steps, End Goal, and Narrowing.
-
Role: Define the role or identity the model should assume (e.g., "You are a financial advisor").
-
Instructions: Provide step-by-step guidance on how to carry out the task.
-
Steps: Break down complex tasks into manageable sub-tasks for better results.
-
End Goal: Clearly state the task or goal you want the model to achieve.
-
Narrowing: Set limits on the model's output, specifying what it can and cannot do.
By strategically placing all these elements in your prompt, you can talk to the AI model in a way that's clear and focused.
Tips and Tricks for Creating Well-Formed Prompts
-
Give clear and specific instructions: Use precise language and quantify requests.
-
Add contextual information: Provide background data, definitions, or source references.
-
Structure prompts: Utilize headings, lists, and delimiters to improve clarity.
-
Experiment with parameter values: Fine-tune parameters like temperature, token limit, and top-k to control randomness and output length.
-
Leverage the large context windows: Large language models (LLMs) can process significant amounts of information.
-
Ask LLMs to help with LLMs: Use an LLM to improve on prompts using phrases like "Help me improve the following ask:..." or "Add 5 questions about the request that would improve this prompt".
Key Considerations
-
Model Understanding: Familiarize yourself with the specific capabilities and limitations of the chosen language model.
-
Hallucinations: Be aware of the potential for AI models to generate plausible-sounding but incorrect information. Employ grounding techniques and critical evaluation to mitigate this risk.
-
Safety & Bias: Implement safety measures and guidelines to minimize biased or harmful outputs. Utilize safety filters provided by Vertex AI.
-
Continuous evaluation and refinement: Prompt engineering is an iterative process; regularly evaluate and refine prompts based on performance and feedback.
Additional Tips
-
Offer a few examples (shots) of the desired input-output relationship to help the model understand the task, desired format, level of detail, etc. You can use one example or have even better results with a minimum of 5.
-
Ask the model to explain its thinking/reasoning. Include phrases like "explain your reasoning" or "provide evidence".
-
Specify the desired output format (e.g., JSON, Markdown, table).
Example prompt
Role: You are a financial advisor.
Instructions: Analyze the sales data to identify the most profitable items. Consider factors such as profit margin, sales volume, and return rate.
Steps:
-
Calculate the profit margin for each item.
-
Analyze sales volume and trends for each item.
-
Calculate the return rate for each item.
-
Rank items based on a combined metric of profit margin, sales volume, and return rate.
End Goal: Provide a clear and concise report of the top 10 most profitable items in the business.
Narrowing:
-
Focus only on the last 12 months of sales data.
-
Exclude discontinued items.
-
Do not consider items with less than 10 total sales.
Multi-Shot Prompting
Multi-shot prompting is a technique where you provide the AI model with multiple examples of the desired input-output relationship. This helps the model understand the task, the desired format, the level of detail, etc. It's like showing the model a few examples of what you want before asking it to do it itself.
How to Use Multi-Shot Prompting in the RISEN Example above
To use multi-shot prompting in the RISEN example of identifying the most profitable sales items, you would provide the AI model with a few examples of sales data and the corresponding analysis of the most profitable items.
Example 1:
Input:
-
Sales data for a hypothetical business, including item name, sales volume, profit margin, and return rate.
Output:
-
A report of the top 3 most profitable items based on the provided data, considering profit margin, sales volume, and return rate.
Example 2:
Input:
-
Sales data for another hypothetical business with different items and sales figures.
Output:
-
A report of the top 5 most profitable items for this business, again considering the relevant factors.
By providing these examples, the AI model can better understand how to analyze sales data and identify the most profitable items. This can lead to more accurate and relevant results when you then provide your own sales data.