1. Negotiating with an FBI agent
Welcome back!
2. Navigating negotiations with an FBI agent
Prompt engineering is the practice of optimizing a prompt to get the best response from language models like ChatGPT.
Today, we’ll begin our journey of mastering prompt engineering with the help of an ex-FBI negotiator combined with ChatGPT in a real-time interactive scenario so that the next time you negotiate a raise, you’re armed with the tools to succeed.
3. Negotiation simulation
When I think of negotiation, one person comes to mind: Chris Voss. He’s not your average negotiator. Drawing from his rich experience as an FBI hostage negotiator, he wrote "Never Split the Difference," a field manual for high-stakes negotiations. The techniques inside the book, crafted for critical scenarios, are invaluable for everyday situations.
We’re taking a unique approach by simulating a negotiation with ChatGPT. For each step, you’ll use Voss's techniques to conquer the negotiation for your next pay rise.
Let’s see what this prompt structure looks like.
4. Prompt
Wow, that’s a big prompt!
A prompt this size can seem overwhelming at first if it lacks the appropriate structure, which is exactly what’s missing here.
5. Prompt with XML tags
We can solve this by using XML tags to mark different parts of our prompt.
In our example, we use XML tags to indicate the beginning and end of text that ChatGPT should identify.
There are three key benefits of using XML tags in your prompts.
6. XML tags - Structured input and output
Firstly, structured Input and output. Extensible Markup Language, or XML, is a markup language designed for storing and transporting data. When used in prompts for LLMs, XML tags can help define a clear, structured format for both input and output. This structure can enable more precise and systematic data extraction, interpretation, and generation.
7. XML tags - Improved processing
XML tags can help delineate different sections or types of data within the prompt more clearly. This helps the model process the information more effectively, potentially leading to more accurate and contextually appropriate responses.
8. XML tags - Facilitating complex interactions
In scenarios where complex data interactions are necessary (such as in our negotiation example, where we work through multi-step reasoning), tags can mark different steps, making it easier for the model to follow the required logic and sequence.
9. The next level
Let’s take this one step further.
We’ll provide examples of achievements of what we’ve accomplished to reinforce our negotiating position.
Here’s what the new prompt looks like.
10. Augmented intelligence
Instead of us prompting ChatGPT, the AI can now prompt us during the negotiation - acting as Chris Voss - leveraging these examples to our advantage.
11. Augmented intelligence
Chris Voss acts as our negotiating copilot. Not only do we get a tip and a framework to inspire us, but we also get a potential reply and highlight from our list of achievements to reinforce our position.
12. Let's practice!
It’s time to turn our attention to the exercises, where you can test your newfound negotiation skills.