Advanced Prompt Engineering: 4 Mindset Shifts
Let’s delve into advanced ChatGPT prompt engineering.
There are four fundamental mindset shifts that will allow you to unlock Chat GPT’s full potential.
ChatGPT has the potential to improve every aspect of your life if you know how to wield it.
Because ChatGPT is like a pair of nunchucks.
Examples of Potential Unlocking
But to explain why, I need to show you a few examples.
If you put in the prompt “Write me a book about productivity,” you get something that looks a little bit like this.
It’s got a title, it’s got a page of contents, and it’s got some short chapters.
But this is not a full book.
What are you going to do with this?
Nobody’s going to buy this.
It’s just a bunch of regurgitated cliches.
Many people give up at this point, and that’s where the first mistake is made.
Advanced ChatGPT Prompt Engineering
Let’s compare this output to the output of using a more comprehensive planning prompt.
So now I’ve asked ChatGPT to suggest a detailed non-fiction book writing process map at each stage and the exact outputs needed to move on to the next step.
What you get here is the entire process.
And then what you must do is use each of these points as prompts individually.
You then go through the entire process step by step.
Chain of Thought Prompting
So for example, if I were to focus just on titling right now, I would take this point: “Create a working title” and use this as a basis for a prompt.
I would input exactly this.
So now I’d go in and say “Create 10 options for working title on a book on productivity.”
Come up with creative angles on creativity for digital entrepreneurs.
And now we have 10 ideas for titles, everything from “Digital Dynamo” to “Cultivating Creativity.”
By breaking down the process, we can improve our outputs.
Now here’s what we call more advanced ChatGPT prompt engineering.
Now you can get a little bit more advanced and connect each of these prompts you develop and you link them together into a super prompt.
This is known as Chain of Thought prompting, and that is a key technique to understand to elaborate and improve your processes inside of Chat GPT.
Thinking in Tree of Thoughts
Okay, so now you can understand the importance of linking prompts together like the chain in your nunchuck.
But to take this one step further, instead of thinking of a chain, you can think of a tree.
Let me explain. If you are putting in an input and based on a certain result, you can ask Chat GPT to do other actions.
So if it comes up with, for example, an idea that leads it down a different path, you want it to take a different set of processes.
And this is known as a tree of thoughts.
Collaborative Expert Thinking
So let me show you an example of this in practice.
Imagine three different experts answering this question.
All experts will write down one step of their thinking, and then share it with the group.
Then all experts will go on to the next step.
If any expert realizes they’re wrong at any point, then they leave.
The question is, and from there, you can add your own question.
The reason why this process is so useful is that it enables large language models like Chat GPT to self-evaluate the progress in intermediate steps too.
This helps them move towards solving a problem through a deliberate reasoning process.
So now we’ve got prompt chaining and trees of thought.
Understanding Chat GPT’s Role
So remember, ChatGPT is your assistant, not your slave.
ChatGPT is your guide, not a genie.
It will not perform miracles, but you can ask it to help you find the answers.
It is a multiplier of your own creativity, your own perception, and your own understanding of processes.
You can imagine that you are the composer and ChatGPT is the orchestra that you’re orchestrating to play the music that you have brought.
Reasoning Over Fact Recall
Now the next move to add to your arsenal is to understand that ChatGPT is for reasoning, not for fact recall.
Many people go into ChatGPT expecting it to be a library of information, but that’s not the right approach.
What you want to do is to set some guardrails.
Ask ChatGPT to start thinking within those guardrails and see what it comes up with.
For example, if you ask it for famous quotes about a certain topic, let’s say inspiration, it’s very likely that ChatGPT will actually fabricate and invent quotes that do not exist.
So you might ask ChatGPT to write you a set of quotes in the style of the philosopher Alan Watts, and these will actually meet the requirements of your prompt.
This plays on the strength of ChatGPT to understand that it’s for reasoning, for creating within a set of parameters, not for regurgitating facts.
Gaining Probabilistic Insights
Now, another little mindset shift that I think is particularly potent is when you’re trying to get ChatGPT to have an opinion on something, it often refuses.
Take this example where I’ve asked ChatGPT its views on what will be the future of global warming.
But to get a better insight, you can actually ask ChatGPT for probabilities.
Asking it to think in probabilities gives you the ability to think in bets and not get ChatGPT to simply refuse to give its opinion on a situation.
So if you’ve got an important life decision to make, you could enter in exactly the components that make it up.
It could be about whether you should take a new job or whether you should start a new relationship.
You can put these in and develop probabilities coming out.
Guided Creativity and Constraints
Now let’s move on to one of my favorite concepts.
Leonardo da Vinci once said that art lives from constraints and dies from freedom.
Now, how does this relate to prompting with ChatGPT?
Well, it’s important to have guided creativity for creative tasks. Guide the AI by providing a clear direction and concept.
Remember, the AI is not inherently creative, but it can generate responses based on its trained data that meet certain parameters.
So being specific is key.
Asking exactly what you want and defining what is the outcome that you are looking for will give you greater potential.
For example, you might ask for a set of brand names for a certain company.
Now you might want it to rhyme, or you might want it to be alliterative.
What you’re going to do is start to put constraints inside of ChatGPT.
So for example, I might ask it to give me 15 creative names for a brand and the company creates AI responses for dating apps.
I want these names to be either alliterative or to rhyme, and I want at least 50 of them to start with the letter ‘S’.
Using Chat GPT for Role-Playing and Exploration
Next up is a method that will be particularly familiar to actors.
You can ask ChatGPT to take on a certain role.
This is important because you would not ask your psychotherapist to perform brain surgery, or your cook to give you a massage.
So, what we can do is inform ChatGPT to take on a certain role.
One prompt I play around with a lot is, “I want you to act as a psychotherapist, psychologist, and life coach.
Reflect back to me musings, insights, and questions on the following input.”
Now, once I get an output from this, I might ask you to reply to me in the style of Sigmund Freud.
Or mix it up and ask it to apply to me in the style of Gabor or Mateo, one of my favorite psychiatrists.
You can even ask it to be anyone, even Albert Einstein to discuss relativity.
By doing so, you can create engaging and insightful dialogues that open up new realms of understanding.
Remember, the only limit is your imagination.
Exploring New Realms and Engaging Dialogues
You can take this one step further and ask ChatGPT to imagine a dialogue between two people exploring the concept you’ve asked it to think about.
There’s a useful resource called “Prompts Chat,” which has a library of different role-playing prompts.
Feedback and Growth: Treating Chat GPT Like an Athlete
Next, it’s time to get a little bit strict, because feedback is the breakfast of champions.
Let’s go deeper into advanced ChatGPT prompt engineering.
Think of athletes: average athletes want to be left alone, good athletes want to be coached, but great athletes want to be told the truth.
Similarly, think of ChatGPT like an athlete you’re training and coaching.
A major mistake people make with ChatGPT is taking the first output they get.
It’s important to give feedback to improve its output.
You can ask ChatGPT to be more concise, formal, casual, insightful, arrogant, and proud, or to explain things in a more complicated way.
By doing this, you can fine-tune the tone, style, and output you get.
Evaluating and Engaging with Chat GPT’s Output
Take it a step further by asking ChatGPT to evaluate its own work.
For example, if you’re writing a blog post hook, you can ask Chat GPT how well the hook engages the reader, and then incorporate its feedback into its work.
Remember, ChatGPT’s effectiveness depends on your ability to prompt and provide feedback.
Think of it as a two-way street of constant dialogue for growth and learning.
Priming and Crafting with Context
We’re not only providing feedback after getting output.
We’re also looking at priming with purpose and crafting with context.
Benjamin Franklin said, “Tell me and I forget, teach me and I may remember, involve me and I learn.”
We’re involving ChatGPT at every step.
For example, you can provide a collection of emails you’ve written and ask Chat GPT to respond in your own voice.
You can also share journal entries for ChatGPT to reply as you, allowing for a back-and-forth conversation.
To maximize your interaction, break down your processes into singular moves and build up combinations into super prompts.
This helps you engage with ChatGPT in various roles and contexts.
If you’re interested in learning more about prompt engineering, there’s a course available on Udemy with a special discount for viewers of the channel.
I’m Samson Vowles, and on this channel, we discuss how to get the most out of AI.
Thanks for watching.
And be sure to check out the next video for more on learning how to wield AI.
If you liked this article about advanced ChatGPT prompt engineering, check our other articles about AI below.