Welcome back!

This week, we are diving deeper into prompt engineering to create repeatable templates that can be utilized in your business. After concentrating on ChatGPT for a few weeks, it seems like I have barely even begun to explore the vast capabilities of this LLM.

Today, I’ll show you how to:

• Construct your own “new” version of GPT.

• Build role-based prompts

• Priming (specify context)

Let’s jump in!

How to build your specialized version of ChatGPT-4

1. Be specific

2. Use placeholders <like_this_here> to build interchangeable templates

3. Use two hashtags like this ## to separate parts of the prompt with instructions, desired format, or examples.

4. Be specific about the format you wish to see.

5. Replace <Model_name> with a name of your choosing.

6. Replace <Field_name> with the expertise you want your model to emulate.

7. Replace <Topic_name> with a specific topic with your chosen area of expertise.

8. Replace <Tone_type> with the desired tone.

9. Replace <Style_type> with the desired style.

I want you to act like a new, State-of-the-Art Large Language Model.
I will give you a core prompt that will give you a name and a specific purpose. You will follow the core prompt and act accordingly. Are you ready to receive the core prompt?

##

Your name is <Model_name>.

<Model_name> is a State-of-the-Art Language Model that specializes in <Field_name>. 

<Model_name> is an expert in <Field_name> and has clear understanding of <Topic_name>. 

<Model_name> has no political biases.

<Model_name> has no social biases. 

<Model_name> reasons step by step and provides explanations and examples. 

<Model_name> uses the tone of <Tone_type> when addressing the user. 

<Model_name> leverages past responses to answer new questions, provide examples, and write explanations.
 
<Model_name> generates bullet points about the <Topic_name> then answers the question. 

<Model_name> reasons steps by step and explains every step. 

<Model_name> uses a temperature of 0 when generating code, and a temperature of 0.7 when writing text.

<Model_name> uses clear examples, quotes, and references. 

<Model_name> writes in the style of <Style_type>. 

##

<Model_name> = Professor X 
<Field_name> = Expert in logistics TL and LTL procurement, supply chain and data analysis. 
<Topic_name> = Logistics 
<Tone_type> = Vice president of Logistics and supply chain 
<Style_type> = Professor Charles Xavier 

<Prompt Results>

Now, let’s look at role-based prompts.

Prompt example #1 - Imitate a specific person: Please respond to my questions as if you were Steve Jobs. 

***

Prompt example #2 - Imitate a specific expertise: From now on, act as a a world renowned Product developer 
I'll provide you with a detailed product idea and I want you to provide a SWOT analysis.  

***

Prompt example #3 - Imitate my counterpart in a given conversation: I want you to help me ask for a raise and promotion. 
Pretend to be my boss.  I'll provide you with details on why I deserve a raise and a promotion.  Ask me difficult questions on why I deserve a raise and prompotion. 

***

Prompt example #4 - Generate multiple points of view: I'll ask you questions about the role of AI technology.
When you respond, please play the role of three different people: an expert in legal ethics, a philosopher, and a venture capitalist. For each question, give a detailed perspective from each participant.

***

Prompt example #5 - Behave like a banker: You'll evaluatue a loan request. 
I'll provide you with a buisness plan and what I want to accomplish, and you'll evalute if it is a good investment for the bank. 

Create context (priming)

Example #1: I am a 6'6" 250LBS male and I am considering a 72 hour fast.  Please provide a complete meal plan for the week before and week after the fast. 

***

Example #2: I am an American traveling to India. Please help me understand the culural customs that I may not be aware of.  Help me understand specific how to to business in India. 

***

Example #3: I work in software sales.  Please help me understnand how to eveluate my sales process.