Intro to Prompting for Beginner's and Business
Donovan Rittenbach
Website Copywriter for Small Businesses and Entrepeneurs ? WordPress Marketing Strategist ? Founder Rocketbooks.biz
Part 1: Introduction
“Any sufficiently evolved technology is indistinguishable from magic.”
-Arthur C. Clarke, Sci-Fi Author
This course will teach you how to structure and build prompts to unlock the VAST power of ChatGPT. Sounds scary? Don’t worry. It’s not.
If you can write in plain English, and use a word processor, you can use ChatGPT.
This course assumes a basic familiarity with ChatGPT, a GenAI-powered chatbot from OpenAI. Before we get started, there’s an important point to make about ChatGPT.
OpenAI makes ChatGPT. Microsoft is an OpenAI investor. Microsoft has been trying to take search market share from Google for decades but hasn’t been able to budge it. ChatGPT was launched to take a massive bite out of Google’s search business.
ChatGPT is first and foremost a search tool. It’s a better, more powerful search tool than Google in many ways, but it also has shortcomings we’ll get into.
Now that you have this context, let’s talk about some course requirements.
To do the coursework, you’ll need ChatGPT 3.5. It’s the free version and gives good results, as well as working with most of the prompts I teach in this course. If you don’t have an account, I’ll walk you through how to set one up.
However, when we get into the more advanced sections, I recommend you upgrade to ChatGPT 4. At only $20 a month, it’s a steal, as it gives you access to the most advanced technology on the planet. In addition to providing superior results, it will unlock essential tools like custom instructions, plugins, and other advanced functionality.
Course Structure
This course is broken into two sections, for now. The Basics and Prompt Building Blocks, are available now.
In “The Basics” I’ll give you the big picture view to help you understand what ChatGPT can or can’t do, and how it does it. This context will help you understand how to get better results. Once you have this foundation, you’ll be able to get the best results from your AI conversations.
In “Prompt Building Blocks” you’ll learn how to provide context that gets great results. You’ll learn how to structure a prompt, starting by assigning a role for ChatGPT to play. Then you’ll learn how to output your results in 100 different formats including anything else you can imagine. You’ll also learn utility prompts you’ll use often and how to configure custom instructions, so you can get consistent high-quality results and save time.
Next in the pipeline? Copywriting and Marketing sections are in the works and will hopefully come online soon.
But before we dive in….
What’s a?Prompt?
Prompts are like wishes. That’s good because life just gave you a wish-fulfilling genie, called ChatGPT.
But we all know what happens in the stories when you make a wish…
The genie ignores the spirit of the wish and interprets it literally.
The result is that the wish has unintended consequences because it wasn’t clear and specific.
“But that’s just a fairy tale,” you might say. “There’s no such thing as genies.”
Partly true. Partly not.
Genies are metaphors for powerful forces that can completely change your life and reshape your reality.
And those definitely exist.
Two examples of “genie-level” technologies include the atomic bomb and AI.
Starting to make more sense now?
“Okay but who puts these wild, uncontrollable forces back in the bottle?”
Wizards, or “the wise.” They’re powerful people who think strategically. Rather than doing something impulsive and foolish, they consider the consequences of their actions.
Maybe you’re thinking, “I wish there were more people like that in the world. Okay, since I’ve got this genie, how do I make good wishes?”
Great question. Here are the rules:
1) Always be very literal.
2) Always provide context.
3) Provide it with detailed, specific instructions, preferably with examples.
For instance, tell ChatGPT, “Write me a 500-word blog on coffee.” You can expect poor results because you didn’t use the right format.
This would get better results: “You are a leading authority on coffee and beverages and a top copywriter. Write me a 500-word blog on the pros and cons of coftea in a clear and concise format.”
And that’s prompting, in a nutshell.
Now that you know the basics, are you excited to make some wishes?
I hope you’re as excited as I am to teach you.
Buckle up, because it’s about to become very hard to tell the difference between magic and advanced technology.
?—?— —
Part 2: How to use this?course
First I’ll review the basic outline of this course. As you read through it, see what catches your eye, so you can read it later.
Then I’ll go through what the different color highlights mean so you can quickly scan course lessons if you want to. Finally, I’ll suggest some approaches you might take to learning so this stays fun, interesting, and productive.
Course Overview
This course contains two sections. In the first section, I provide important context such as:
In the second part of the course, I cover important structural and context-setting elements of prompts like:
Now let’s talk about your?mindset.
If you’ve never worked with an AI before, you’ll want to set aside your expectations and approach this relationship with a fresh perspective. Chances are good that you are bound by some old approaches to learning that no longer serve you.
Here are some things you can do to make sure you get the most out of this course.
Jump Around
This course is modular. It helps to have the initial context I give you, but I encourage you to read what interests you the most. Your learning journey doesn’t have to be linear, but it should be rewarding.
Understand Your Objectives
What do you want out of this course? Do you want to wander around the candy shop and just browse? Or do you want to build something?
If you choose the latter, then you have a quest and I’ll help you get there.
If you are just browsing, then have fun exploring and trying new things.
Set Your Imagination Free
Don’t limit yourself to being “practical.” Stretch your imagination.
The fact is, you now have a state-of-the-art, super-genius thought processor to help you solve any problem. Nobody knows what is or is not possible.
And even if it’s impossible today, it might not be tomorrow.
Listen, AI is evolving at an inconceivable rate. At the start of 2023, ChatGPT was “The Thing.”
Nine months later, there are hundreds of imitations, AND they are starting to work together as teams to form companies!
Heed my advice and you’ll make real magic.
Play and Make?Mistakes
“The expert is the person who made all the mistakes.”
We learn best when we play and when we play we make mistakes. You are in a safe place. Nobody is grading you so don’t be shy about trying new things.
Ask Questions
If you have a question the course doesn’t answer, just ask ChatGPT. It’s here to help you through complex issues and can even generate new questions to help deepen your understanding.
Build Your Prompt?Library
See a prompt you like or want to use later? Cut and paste the prompts that resonate with you, and stick them into a text file that’s easy to access. This will become your personalized library of prompts you can revisit any time.
Use Your Prompting Skills With?Humans
As you’ll learn, you get better results when you follow a pattern in your AI conversations. This pattern works with people too. Try it out as an experiment.
I asked ChatGPT “If I want to be a better prompt engineer, what should I do?”
ChatGPT replied:
To elevate your skills as a prompt engineer, here are some key steps you can follow:
Remember, becoming a better prompt engineer is a journey of continuous learning and practice. It’s about developing a deep understanding of the AI chatbot you’re using, honing your prompting skills, and consistently striving for better results. This course is designed to set you on that path.
Part 3: Common Questions
In this section, I’ll explain how ChatGPT works and how that affects you directly. I’ll give some basic background information on how the system works, then I’ll answer common questions like:
“GenAI is a magic trick. But it’s a really good one.”?—?Donovan Rittenbach
Let’s be frank. Artificial intelligence is marketing hype. This is all just machine learning and pattern prediction. There isn’t any actual intelligence here.
Strange as that may sound, especially when you have your first really good conversation with it, don’t be fooled. This is not a sentient intelligence.
Now, let’s look inside the black box of AI. I’ll demystify the magic, and talk about the system’s strengths and weaknesses, so you can get accurate, high-quality results.
How does ChatGPT learn “how to read and?write”?
Chatbots teach themselves to read large datasets. They focus their attention on learning tasks such as masking a single word in a data set and then trying to guess what it is. As they process hundreds of thousands of sentences, they formulate a mathematical probability that one word will follow another.
领英推荐
How does it decide what word comes?next?
If I were to ask you to fill in this blank “The sunset is _______” there is a high probability you would say yellow, red, gorgeous, beautiful, stunning. Chances are extremely low, almost zero, that you would say “delicious.”
As AI trains itself on a dataset, it learns which words will most probably be the next word. This enables it to fake thinking.
Why does the text appear to be typed out by?ChatGPT?
Generative AI builds sentences word by word, guessing which word will come next until it completes a sentence. One sentence leads to another and the result is a series of paragraphs.
Because it takes time to guess and assemble these words, ChatGPT engineers worked out a clever trick to mask this process. Rather than outputting a sentence at a time, they stream one word at a time. This gives the illusion the chatbot is typing words.
What are hallucinations and why are they an integral part of generative AI?
When the AI trains itself on a data set, it learns that some words are much more likely to be used in a certain context than other words.
It does this by building a big dictionary of words in a vector database, a type of database that helps track the probability that a certain word will occur next in sequence.
When the AI answers your questions, it guesses which words should come next. It isn’t thinking. It merely gives the illusion of doing so.
When it uses words that are outside the probability of a “correct answer”, it gives incorrect answers and is said to be hallucinating. That’s not a bad thing though.
Hallucination is both a bug and a feature. It’s a bug when it gives wrong answers, but it’s a feature because it needs to be able to do this if it is to generate a new output.
When should I trust or not trust a generative AI’s?output?
A good rule of thumb is to always question an AI’s output and fact-check its results. This is especially true if its advice impacts your real life in the realms of finance, research, or health.
Looking for hard facts? Use Google’s standard search engine which techies know as BERT.
But isn’t the Google search engine an artificial intelligence?
Yes, in 2017, Google started using a machine learning model called BERT. It uses natural language processing to get great results, but it’s not a generative AI. It can’t make anything up. It just gets results and spits them out.
The bottom line, you can trust its output because it’s just pointing you at human-created articles that are generally reliable sources of information.
So how do you get the accuracy of Google’s BERT model combined with the ability to format output as generative AI?
It won’t be long before a chatbot uses the results of a Google search to retrieve trusted information as a first step. Then generative AI will present those results and format or output as specified by the user. This is the future of ChatGPT and is known as retrieval augmented search(RAGS).
For now, though, combine Google research with your chatbot to get reliable output.
Why do LLMs perform math functions poorly?
If you ask a generative AI “What is 2+2?”, it may get it wrong because math does not follow the rules of language. It does however occur sequentially, which is something computers are good at doing. As a result, LLM programmers are rapidly evolving their chatbots to answer complex math problems and give correct results.
Why is the sequence or type of prompts important?
LLMs require context to properly solve problems. If it doesn’t have proper context, and instructions on how to proceed, the prompt engineer must provide it.
For example, if you want ChatGPT to write like you, begin your chain prompt by feeding it a writing sample. Ask it to analyze it for style and tone of voice. Once it does so, ask it to provide its output in that same style and tone of voice.
In this chapter, you learned that AI makes up sentences based on the probability that one word follows another. You also learned the importance of giving context and good instructions, as well as why you should never trust an AI’s output.
Okay now that you understand how chatbots “think” hopefully you will get better results.
Do I need to be a programmer to use?ChatGPT?
If you are new to prompt writing, don’t worry. Prompts are questions or instructions written in plain English. You don’t have to have to be a programmer or techie to unleash the incredible power of chatbots like ChatGPT.
You just need very basic computer skills, and preferably the ability to write well.
As long as you’re open to exploring new technologies, you’ll do great.
Also, don’t worry, you can’t break anything so don’t be shy about trying new prompts.
Do I have to write my prompts in?English?
No. I know many people who use it aren’t native English speakers. Although ChatGPT is best with English, because that’s what it was trained on, don’t worry.
ChatGPT understands you, even if you aren’t a native English speaker and prefer to prompt in French, German, Spanish, Chinese, and many others.
Ready? Let’s get started!
?—?—?—?
Part 4: How ChatGPT Does Its “Magic”?—?A Layperson’s Explanation
GenAI is a magic trick. But it’s a really good one.
Let’s be frank. Artificial intelligence is marketing hype. This is all just machine learning. There isn’t any actual intelligence here. It’s pattern prediction, and that’s all.
Let’s look inside the black box of AI. I’ll demystify the magic, and talk about the system’s strengths and weaknesses, so you can get accurate, high-quality results.
How does ChatGPT learn “how to read and?write”?
Chatbots teach themselves to read large datasets. They focus their attention on learning tasks such as masking a single word in a data set, and then trying to guess what it is. As they process hundreds of thousands of sentences, they formulate a mathematical probability that one word will follow another.
How does it decide what word comes?next?
If I were to ask you to fill in this blank “The sunset is _______” there is a high probability you would say yellow, red, gorgeous, beautiful, stunning. Chances are extremely low, almost zero, that you would say “delicious.”
As AI trains itself on a dataset, it learns which words will most probably be the next word. This enables it to fake thinking.
Why does the text appear to be typed out by?ChatGPT?
Generative AI builds sentences word by word, guessing which word will come next until it completes a sentence. One sentence leads to another and the result is a series of paragraphs.
Because it takes time to guess and assemble these words, ChatGPT engineers worked out a clever trick to mask this process. Rather than outputting a sentence at a time, they stream one word at a time. This gives the illusion the chatbot is typing words.
What are hallucinations and why are they an integral part of generative AI?
When the AI trains itself on a data set, it learns that some words are much more likely to be used in a certain context than other words.
It does this by building a big dictionary of words in a vector database, a type of database that helps track the probability that a certain word will occur next in sequence.
When the AI answers your questions, it guesses which words should come next. It isn’t thinking. It merely gives the illusion of doing so.
When it uses words that are outside the probability of a “correct answer”, it gives incorrect answers and is said to be hallucinating. That’s not a bad thing though.
Hallucination is both a bug and a feature. It’s a bug when it gives wrong answers, but it’s a feature because it needs to be able to do this if it is to generate a new output.
When should I trust or not trust a generative AI’s?output?
A good rule of thumb is to always question an AI’s output and fact-check its results. This is especially true if its advice impacts your real life in the realms of finance, research, or health.
Looking for hard facts? Use Google’s standard search engine which techies know as BERT.
But isn’t the Google search engine an artificial intelligence?
Yes, in 2017, Google started using a machine learning model called BERT. It uses natural language processing to get great results, but it’s not a generative AI. It can’t make anything up. It just gets results and spits them out.
The bottom line, you can trust its output because it’s just pointing you at human-created articles that are generally reliable sources of information.
So how do you get the accuracy of Google’s BERT model combined with the ability to format output as generative AI?
It won’t be long before a chatbot uses the results of a Google search to retrieve trusted information as a first step. Then generative AI will present those results and format or output as specified by the user. This is the future of ChatGPT and is known as retrieval augmented search(RAGS).
For now, though, combine Google research with your chatbot to get reliable output.
Why do LLMs perform math functions poorly?
If you ask a generative AI “What is 2+2?”, it may get it wrong because math does not follow the rules of language. It does however occur sequentially, which is something computers are good at doing. As a result, LLM programmers are rapidly evolving their chatbots to answer complex math problems and give correct results.
Why is the sequence or type of prompts important?
LLMs require context to properly solve problems. If it doesn’t have proper context, and instructions on how to proceed, the prompt engineer must provide it.
For example, if you want ChatGPT to write like you, begin your chain prompt by feeding it a writing sample. Ask it to analyze it for style and tone of voice. Once it does so, ask it to provide its output in that same style and tone of voice.
In Summary
In this chapter, you learned that AI makes up sentences based on the probability that one word follows another. You also learned the importance of giving context and good instructions, as well as why you should never trust an AI’s output.
Okay now that you understand how chatbots “think” hopefully you will get better results.
About the Author
I’m Donovan Rittenbach. I use ChatGPT almost every day.
As a marketer, content creator, and copywriter, I’ve spent hundreds of hours collecting valuable prompts and using them.
This course is the culmination of my research and represents a fukton of work.
I did this project because I don’t have a job, and I need to build a network and a community.
I also love changing lives, helping people, and solving problems. One way to do this is by teaching this life-changing skill.
Contact Info
Follow me on LinkedIn: https://www.dhirubhai.net/in/drittenbach/
Full Course & Community Access
Check out my YouTube channel: https://www.youtube.com/channel/UCNYy22Jn7gI6upachL_uQOA
AI Meetup
I teach ChatGPT on Meetup.
Got feedback? I welcome your suggestions, feedback, and testimonials.