Transcripts
1. Intro: Oh hello. Are you here to learn the advanced GPT prompting
techniques and patterns? And I'm going to take all your
results to the next level. Do you know what the lift
interaction pattern, the question refinement pattern, and the cognitive
verifier patterns are? No, that's great,
I got to show you. Come on in. Rafting
effective prompts involves both the process of
engineering and experimenting. In my day job, I've been
testing our different apps, different approaches,
different methods, and different patterns to
find the ones that will guarantee the best results to speed up our daily processes. Today, I love to
share with you some of my interesting takeaways. And share the most
effective methods that we're a total game
changer when it comes to transforming general prompt to personalized ones that
meet all my needs. This course will be laser focus on these
prompting patterns. These methods are extremely effective and they really
deserve a separate course. Here we are today. Also,
I didn't talk about those patterns in any
of my other AI courses. You're the lucky one watching
this one today, okay. What will you need
the first thing, access to chat GPT, it can be a free version. And the second most
important thing, the willingness to
try something new and learn more about PT and
other AI powered models. Some of the prompts will
be complex and long, so we also prepare a
downloadable PDF so we can copy paste the prompts and test them right
away along with me. And you might want to ask, do you need to be a chip pro to benefit from using those
patterns, those techniques? And the answer is absolutely no. Plus I'm sure it will help you optimize your tasks no matter what niche or what industry
you are in. Sounds good. Needless to say, I'm
really excited to show it all to you because
there's a lot to uncover. And I think many, many people using chap, don't know these
patterns, these methods. So you'll be really ahead. Let's go.
2. Our Class Project: Class project. So
your only homework, I mean our class
project for today, is to take a screenshot of your favorite result
you got from GPT. Take a screenshot of
your conversation with Cha GPT or any other IA model if you prefer the different one. And share your best
result you've got from using the techniques we will
be talking about today. Share your best result. Take a screenshot
experiment with Chachi Pt and show us the results so we can
be all proud of you. I've always posted right here in the class
project section. And if you need my
guidance or my feedback, let me know below
your screenshot. Of course, I will gladly help you out or answer
your questions, so don't hesitate
to let me know.
3. Why Advanced Prompting is Worth it - Prompt Engineering Advantages: Why advanced prompting
is worth it? Do you think unlocking AI's full potential is all
about better algorithms? In my opinion, the true secret lies in mastering the art of questions and giving the model more advanced, better prompts. That's what we are here today. And if you want to ask
what the prompt is, I don't know this term. The prompt is simply the input that we feed the AI
power tool with, and it's a response based on it. So the prompt is basically what we type in into the chat bot. Those can be questions,
those can be tasks. It all depends on what we
need from the chat bot. And more advanced prompts don't necessarily mean longer prompts. It's just all about
asking the right things. That's also what prompt
engineering is all about. And today we will uncover a few really cool
prompting methods. These methods are
quite advanced. I will do my best to
explain them simply so everyone can understand and use them right after the course. We won't use any technical
jargon or any difficult stuff. I will do my best to make
everything simple and logical. Yeah, take my word for it. My prompting journey and testing different approaches
was, and still is, exciting, intellectually
stimulating, and well, many times really
frustrating experience. But actually during
the few last months, my entire conception of
writing a better prompt, my entire conception of
prompt engineering and what is possible to achieve with the help
of AI changed so much. I deeply hope I will be able to surprise you
the same way I was surprised with all
the techniques and all the improvements we can implement into our
conversation with. Because let's be honest,
most of the time, basic simple prompts will only get you very generic answers. Although the model is
getting better and better, we can communicate with
it as we do with humans. So we can just type in the simple sentence
and it will answer, it doesn't that well. And the more advanced, the more polish the prompt is, better results we will get. That's what many
people forget about. And they are complaining that
the results from GPT aren't good enough or they are generic or they aren't
meeting their needs. You know, they forget about polishing the
prompt to get what they need because we really can get exactly what
we want from GPT, just just a matter
of communication. Ai isn't human and we have to communicate with
it a bit differently than the way we communicate with our assistants or our friends
or just people in general. So I often hear people around
me on line and off line, complain all the time about GPT. The poor results in that I doesn't help and doesn't
change anything for them. And I think so many
of these cases and these complaints are
because their problems were too basic and they just
can't see what GPT and other high power tools can do and how much they can
actually help them. Because look, when we
talk with a friend, our art of communication is all about knowing
what to say, when to say things
you want to say, and how to say the
tone, the timing. It all matters. Same way it
applies to talking with AI, it's also a collaboration. The output depends on the input. Actually, as I've
already told you, AI is even more
sensitive than humans. And our word choice is more important as every word matters, every word changes the prompt
and can make a difference. We can get a totally
different outputs based on the way we
type in the prompt. So it's really important,
and it's also really important to remember that
AI doesn't read our mind. Yeah, it doesn't work like that. And with friends, the situation is a bit different because some good friends can do
that, at least best friends. Right. And I can. So we need to tell it in
detail what we need and that's what we're
going to learn today. Why improving our
prompt is the key. Today I will show you
that better prompts can give you more fun and interesting
conversation with AI, so you can easily discover
its capabilities. You can explore the
full potential of GPT and see what it
can unlock for you. What it can do for you, can generate unique
and creative content. You will be surprised to what extent the list is
actually really long. And I could talk about that
for hours, believe me. But I know you're here
for the practical part. And that's where
we are heading to.
4. How to Transform Generic ChatGPT Outputs into Quality Content: How to transform generic GPT outputs into
quality content. As you already
know, GPT is one of the most stranding Ls with
over 100 million active users. It can respond to every
question you ask, every prompt you give it, and has so many text
generation features. But if you have been using
Cha GPT for a while, you already know you start
noticing that the responses generated by the model have
a very similar pattern. They often have this flat
friendly and ultra polite tone and they just lack
emotional death. Gpt will quite heavily write a page of content
for your social media, for your blog, for actually any purpose you
need. So prompted. However, there are
some drawbacks to simply prompting
to, For example, write content on topic X for my blog if your
prompt is that simple, if your prompt is, for example, write a story about why journaling is beneficial
for your health. The AI might create
an interesting story. Something you will love, but it could go
in any direction. Basically, you don't know
what it will produce. So of course, there is
this element of surprise, but sometimes the result. Most of the times we need
results that meet our needs and are tailored to our business,
to our preferences. So we can't use such
generic prompts. If your prompt is, write an
emotional story about stress, 27 year old woman who
avoids cardiac issues. Thanks to journaling,
I need to make all the skeptical people
believe in the power of journaling and how it can
help you relax the AI will be better guided and
produce the story after. Do you see the
difference? Let's compare the results we got from the first prong and the results we will get from
the second prong. So we can see the difference
in put in practice, we will start with
the first prompt and we will see how it goes. Okay, pays the first prom. Write a story about why journaling is beneficial
for your health. Let's see what GPT
generates for us, Okay? The story is quite long, I think the length
of it is, okay. It's basically what
I was hoping for. But there are some parts of
it that I don't quite like. I think I wouldn't use
it for my business. For example, if I had
a journaling business. If I was selling journals, I don't think some of the
details of the story would be, would be okay for it. The story is okay. It actually surprises me in a positive way, but at the same time,
I feel like there are so many details
need to change. So let's see what GPT will give us when prompted
with the second prompt. Okay, I already see that the second prompt is
performing much better. That's not a surprise
because actually we knew that the more detailed
prompt will perform better, as you can see, because
we're doing it only for a short comparison
because that's not one of the techniques
we will talk about. We'll talk about totally
different techniques. So this is just a
comparison what PD, when it's prompted with a very simple, very generic prompt. And what it does when we
elaborate on our needs. So as you can see, the
second story is really good and all the details that I
needed are added to the story, say so the way she avoids cardiac issues,
thanks to journalism. And I think it's really good. Of course, I still would need to edit it on a bench to use it. But I just wanted
to show you how different the result can be when the prompt is more complex. We can also go a step
further and tell Chad GPT. As a skilled storyteller, your task is to create an original and captivating
story about here. Report our topic prima target audience aged he
repeat the age A ears. All your story should be details immersive
and filled with vivid descriptions that
engage the senses or readers. The plot should focus on a meaningful life lessons with relatable characters
that the reader can emphasize with
and learn from The story should be a
minimum of 600 words. Here we can change the length. Please ensure that the story unfolds in an exciting manner, building up the thrilling climax that captures the
reader's attention. Conclude the storytelling
with a satisfying ending that effectively conveys
the moral lessons learned through the man's
character experiences. Your story should be
creative and original, providing enough detail
to clarify the narrative, evoke emotions, and create a memorable experience
for all the readers. When crafting the story, keep the target age
group in mind and the moral lesson you
wish to convey them. Feel free to incorporate elements of romance
and adventure. And always consider
the importance of engaging story techniques that
appeal to this age group. This power prompt quite long and don't worry
if you want to use. It's also included in the PDF, although it's not using the
methods we'll talk about. You can download it in the resources section
and let's compare how this prompt will perform compared to our
previous responses. And we need to edit
with the topic and our target audience
details and we are pasting it instead of in
the place of our topic. Okay, so K story teller
task is to create alternate cap story
about the stress by seven years old woman who
avoids cardiac issues. Thanks to journaling
for my target audience. Let's say 32, 50. Our Pm is already good. Good to go. Let's check
how it will perform, okay? As you can see, the story is really much longer than, than the previous
results we got. I can see that is
written much better. The style is much better. It's much more pleasant to
read it, it's really long. So it can be a block post, or we can actually make the entire scenario for
a story, for a video. From that, the results
are much better. The more detail the prompt is, the better are the results. So as you can see, the
result is interesting. It's intriguing, but can't
get any bet. It can. And we'll level
the result up with the flipped interaction pattern.
Have you heard about it? If not, don't worry. I'm here to show you and explain everything to
you so it makes sense.
5. The Flipped Interaction Pattern: Listen. Typically it's
you who pose questions through the model which in turn generates an
output for you. We type in something into the chat box space and
we get a response. A flip interaction pattern turns this dynamic on its head, and it's PT, asking
questions for you to answer to get better results and understand
your needs better. It may sound crazy
at this point, but trust me, it's really smart. Because that way GPT knows what you need
and you are met with context and with the details it needs to generate
much better results. So why should we
use this method? There are two main advantages. One for the large language model for our GPT or other chat box, and the other one for you. For the model, this
chromatic pattern allows it to gather all the information it needs to answer
your questions. I'll perform your
task much better in a more precise and more tailored way tailored
to your needs. Otherwise, you may think that
when writing your prompt, you're arming GPT with all
the required information. And the information you type
in is probably enough and the model can be actually missing critical
details that you miss. You simply didn't
think about when typing in the prompt and
it happens many times. It also happened before when I was typing in the prompt
about journaling, it was very generic after all. The third prompt I typed
in was more complex. It was a power prompt. We set the tone, we've told P. Use imagination to implement and include
creative examples. But anyway, it was generic and I didn't include all
my needs in the prompt. The other advantage
of this pattern, of this flip
interaction pattern, is that it makes our
interaction with GPT more easy. Because if you'd like to come
up with a perfect prompt, it would mean investing a
lot of time thinking how to describe everything you
need and crafting a long, complicated prompt
with all the details you might think
the AI would need. The flipped interaction pattern
simplifies the process by posing a series of question
that you will answer. You can answer them in bullet
points on paragraph format. It's up to you and it makes
the whole process easier. So yeah, you've heard it right. The chat bot takes the lead by initiating the conversation and prompting you for a response. This pattern allows the
chat bot, our model, to play a more proactive
role in guiding the conversation and providing much more personalized
assistance. And we love it, so
you already name why. It's called the flipped
interaction pattern. We are simply
slipping the roles. I can set the
context much better, gather specific
information from you, and steer you towards
desired actions or decisions potentially
leading to different, much better outcomes
Home purged to traditional user driven
models and methods. In other words, to wrap it up, you get the desired results much faster and
it's really useful. Now I'm going to show
you the problems you need to type in to
use this method. Plus we are going to
put it into practice, and I will show you
practical cases when you can need it,
when you can use it.
6. The Flipped Interaction Pattern In Practice: The flip interaction
pattern, pull and practice. I have experimented with many different prompts with
many different approaches, and I already have
my top favorites that I'm going to
share with you today. Our goal is to turn
our chat bot into the prompt creator to
create the perfect prompt. And to make it create the best results we can get from GPT or any other AM models. To show you initial pattern of this flip
interaction pattern. First we'll use this
prompt pattern. From now on I would
like you to ask me questions to the report
a specific task. When you have enough
information to do the task, create out what you want. So let's I want
char GPT to help me writing a block post for my
personal block Sha today. And here's a little disclaimer. Of course I won't use the
output without editing it. I will never, ever do that. Because I'm an advocate
for using AI as our co creators and not using the output
without editing it. I think when it comes to
social media or block post, we need to include our personal spark and our
personal charisma image. And we can just copy paste
the output from chatPT. That's just a little ethical
background for the task, but let's see what ChargPT
will do when prompted now on, I would like you to ask
me questions to write a viral block post for
my lifestyle block Asha. Today, when you have
enough information to write a block post, create outline in all the block post
paragraphs, at least six. Okay. Are you ready?
Are you ready to see how GPT will behave
in such a situation? Let's go. I'm
pasting the prompt. Okay. So the first
question from GPT, what are some trending or interesting lifestyle
topics you have in mind? Okay, That's a good question. Gpt, I will tell you,
tell your lifestyle, journaling and
journaling habit, okay? And as you can see, the
list of questions we got from PT is quite long and
it may actually scare you, but don't be worried. We don't need to
answer to all of them. But if we do that, the char GPT will be able to come up with much
better results. The result won't be
generic, you know, it will match our expectations
and the result will be much higher quality than we've used just a simple
prompt, The generic prompt. So some of the questions, what inspired you to explore the connection
between a healthier lifestyle and journaling in this block post GPT is basically curious
what information, what context it can include in the block
post introduction. Because, you know, I
don't know about you, but when I read block post, I really enjoy those
personal introductions that are full of storytelling. From the very start, you
know that you will read a good written block post
that isn't copy paste from another block or is fully
generated by I. I really love the way ChagPT wants
to learn our experiences, our perspective to include it in the block post because you may think a lot of work to
answer those questions. Of course, it, it
will take some time. It will take a
couple of minutes. But at the same time the
result will be so good, we won't need to
edit is so much. When you use a simple prompt, you will have to add your own paragraphs
to the blog post. You will have to add
your storytelling, your context, your perspective. With a prompt like this, when you answer those questions, the block will already
include your context and your point of view and
your personal experiences. So it's going to
be more personal. And then you can use it without so many edits. So in the end. And it will save you time
even though the process of answering these questions
can take some time. Okay, other questions
are, for example, what key takeaways or final files would you like
to leave your readers with? Do you have any personal
success stories or anecdotes related to healthier lifestyle
and journeying journey that you'd like to share
with your readers. Those questions are really good. I mean they can be
included in, of course, how to write a good blog post
storytelling key takeaways. This all really needs
a little reflection, but if we don't know answer
to any of those questions, we can also ask a GPT to suggest something so we don't
have to come up with every answer on our own, okay? So those questions
are brilliant, and let's say our case is different because I told
you, I promise you, we will discuss many
different cases so you have a broader idea when
to use those patterns, when they can be useful to you. So let's use a different case and we will do it
with the prompt. Judge, I would
like you to ask me questions to help me pinpoint issues with my cars performance. Ask me questions until you have enough information to identify the two most likely causes. Ask me one question at a time. Ask me the first question. As you can see, this prompt
is quite different because we want GPT to ask us one
question at a time. I think that way it can be less overwhelming
because, you know, getting 12 questions at once
can feel a bit overwhelming. I know it from you.
And one question at a time is much better. So, you know, you can
just answer right away. You can just type in 23 words
answer for each question. And it doesn't
take so much time. So let's see how this
prompt will work. Okay. So the first question from GPT is symptom description. Can you describe the
specific issues or symptoms you've
been experiencing with your car's performance? This could include anything in usual, like strange noises, worrying lights
on the dashboard, or changes in how
the car drives. And you may laugh, but actually this question is the question that would
get asked if you call the car assistants or some professional to help
you in such situations. So I think GPT is
really smart and that way you don't have to think about how to describe the issue, how to describe
everything in one prompt. Because GPT taking the lead is asking you one
question at a time. So you don't have to think about answering all the
questions at once. You don't have to
think about typing in such a long description of
what's going on with your car. You can just type in a short
answer to each question. So let's say it's producing
strange noises in, it can speed up. Okay, The second
choices we get from DBT is can you describe the type of noises
your car is making. Id, clanking, clicking, whining, or any other specific sound. Additionally, do
these noises occur all the time or only
when you accelerate? As you can see the questions, the second question
is also very smart. I've tested very
similar prompts and sometimes GPT is asking
us 12 questions, sometimes 20, sometimes less, sometimes only three or four. It all depends on the
context on the situation, but it asks about
all the things it might need to help
you find the answer. So as you can see, GPT
can also be helpful in such situations because when you talk with anybody
on the phone, they would ask you probably
the same question. So it's really smart. Now we will go further and I
will tell you about, well, playing the
personal pattern mixed with the flipped
interaction pattern.
7. Role-Playing (the Persona Pattern) + The Flipped Interaction Pattern: So now let's mix a flipped
interaction pattern with role playing which has its official name,
the personal pattern. It may sound like another
somewhat complicated pattern, but don't worry. In reality it's one of the simplest ones we have in
the front engineering world. In many cases we would like chat bot output to take
a certain point of view. The main intent of
this pattern is to give our AI model a persona that helps it select what
types of output to generate and what details
it should focus on. This pattern usually
begins with the prompt, act as personal X provide outputs that person
X would create. Imagine you're a person X, you're going to pretend to
be a person X from now on, act as a person. So as you can see,
the first statement, instruct the child board to
act as a specific persona and provide responses like
such person would provide. So what our persona can be, This persona can be expressed
in a number of ways. It could be a job
job description, title, fictional character, or maybe even historical figure. Just someone I will know. Okay, so I bet you
would like to ask, show me this pattern
in practice. What can our prompt
look like and listen. Here are the three examples of how an example
prompt can look like. And as you can see,
we've already mixed this personal pattern with the pattern we've learned
in the previous chapters. So our flip interaction pattern. In that way, we have
three very good prompts. You will get amazing
results with these prompts. The first prompt is, imagine
your course planner. The second prompt is step into the role of
a personal coach. And the first prompt is, imagine your guardian
consultants. Would you like to see how
GBT deals with such prompts? Let's go. So I'm copying the
first, the first prompt. These prompts are also in the
PDF I've created for you. You can copy them and
edit them to your knees. Let's check what results we will get from putting the
first prompt into work. So GP team responses that way. Could you please share
some details about your experiences in the
beginning photography workshop? What specific topics
or techniques did you cover and what aspects of photography are you
most interested in? Good, yeah, that's a
good starting point. And let's say we'll
write specific topics. Preparing a good portrait, background, sand,
portrait backgrounds, aspect of photography, that
would be woman portraits. Okay, so the second question is, what are your primary goals and expectations when it comes to improving your skills
Portrait photography, particularly in relation to creating appealing backgrounds
for these portraits. Are there any particular
challenges you've encountered in this area that you hope to address
through the workshop? As you can see, GPT is digging deeper to come up with
really good answers. And that way we can add so
much context to our results, to the results
we'll finally get. And that's how brilliant
this method is. Now, I don't want to bore
you with so many questions, but as you can easily guess, the chat about will go on and on until it gathers all
the needed data, all the needed information, all the needed context. So we can expect
about ten questions until it arms us with
the perfect output. And let me tell
you, it's war fit. Now let's put the
second prompt into war. That's the prompt about the
role of a personal coach. And the first question, the first question
we got from GPT is, what are your fitness
goals and how it does incorporating morning jocks aligned with these objectives? Are you looking to improve
your overall health, build endurance, lose weight, or achieve specific
fitness milestones? So my fitness goal is to lose, let's say five kilos and improve my breathing
techniques during sport. Okay, so as you can see, we can really chat with
char GPT because it just behaves like a really
interested assistant, like a really interested
coach that's asking us so many questions
to dig deeper and to find out what
exactly do we need. And the second question is, can you describe your
current morning routine, especially in
relation to your jog? What time do you
typically go for jogging? And do you have any
particular habits or challenges that you've noticed
during these sessions? I don't know about you, but I can easily imagine a situation like
this in real life. We go to fitness coach, we go to personal coach. And these questions
are so natural, they're like real life examples. So we would say, I
would say GPT is really good at this role
playing and it will ask so many questions
as it needs to get all the context to generate
a very good output, a very good piece
of advice for us. So that's a brilliant, you know, we wouldn't be able to produce such a good prompt
from the very start. Your prompt will,
would have to be very long to all the context, all the information about you. In that way, it's
much more friendly. It's much easier just to
answer the questions, just to answer one
question at a time, okay? But let's check the third
option, the third front, it's about gardening consultant, so this one is
really interesting. Let's see what
answer we will get. Okay, So the first
question we got is, what are your main objectives when it comes to your vegetable? Oh, that's a difficult word. When it comes to your
vegetable patch. Are you aiming for a sustainable
source of fresh produce? A beautiful addition
to your garden, or perhaps a bit of both? Are there any
specific vegetables you particularly
excited to grow? As you can see, the tone
of GPT is really friendly. It's easy, it just sounds
like an easy going, really friendly gardening
consultant who, you know won't give you a
random piece of advice. It gathers your answers, it gathers information
on your needs. And then it will come
up with the perfect, perfect piece of advice, the perfect suggestion for you, so you will be satisfied with the results. Now it's your turn.
8. The Question Refinement Pattern: The question refinement pattern. Okay, let's go further. Gpt can also help us level
up and refine our questions. For example, if you
ask the question, it can suggest a
better formulated, clearer question for
more accurate results. Let's say you aren't quite sure how to formulate your
prompt, you type it in. And with this pattern, we are asking a PT to make our prompt better and
level it up for us. The main goal of this
pattern is to ensure our chat bot always suggests
potentially better, more refined questions
we could ask it instead of our original
perfect prompt. So yes, I can help us find the information or
context to achieve our goals in fewer interaction
than if we employ the traditional trial and error pattern prompt
prompting pattern. So from now on we will use
this prompting pattern. From now on, when
I ask a question, suggest a better version
of the question to use that includes
information specific to. Here we type in our
use case and ask me if I would like to use
your question instead. So imagine this. You're not just asking, you're collaborating
with the chat bot, you're collaborating with
our large language model, like two bodies storming
the perfect question. You give a basic question, and then here comes the magic
you ask the chat bot for a jest version one that's got all the nitty greedy
specifics and guess what, chat bots got your back. It suggests a super
charged question with all the relevant info you might have missed in
your first question. And let's be honest, it happens quite
often as we are often unaware of additional information
or additional context. That might be
extremely helpful for a chat box when we're
creating the prompt, that's totally normal and
it happens to all of us. That's why it's such
a smart idea to have a chat box step in and
suggests a polished question. A polished question with all the important details
and the context it needs. That way, we can get
much sharper questions, better prompts, and of
course, much better results. When should you use this method? This method, this
pattern will give the best result
in situation when your initial question
is quite broad and the prompt will benefit from
letting the AI to look, guide you through
what kind of context, what kind of information it
is important to add here. Good example. Any topic
when you aren't quite sure what specifics are
relevant and which aren't. For example, creating copy for marketing purposes
is a good candidate for using this prompting method. Let's see how it
goes in practice. Okay, our prompt is this, From now on, when
I ask a question, suggests a better version
that incorporates information specific to writing a scenario for a top quality, high performing Instagram ad. And ask me if I would like to
use your question instead. Here the question, can
you write a scenario for an Instagram ad to use it to market handmade jewelry
made of coffee beans? So let's paste this
prone into Ch. Okay, so the PT
response is certainly, I can help you create a scenario for an Instagram ad to market your handmade jewelry made of coffee beans. Here's
the question. To ensure you gather
specific information for the scenario. What are the unique
features or aspects of your handmade coffee
jewelry that you'd like to highlight in the
Instagram at stereo? Would you like to
use this question to provide more details
for the scenario? Yes, I want to
highlight the fact that the jewelry I create is
zero waste and okay, of course it's made up. I don't produce any
coffee beans jewelry, but that's a creative
idea and we can observe the charges potential
to create good results. So I think it's a good example. Yes, I want to highlight
the fact that the jewelry at zero waste and it's, it's made of organic
coffee beans. Okay, And here the char
GPT started producing this scenario. I really like it. Okay. Can we even
have transitions? Transition details and
how to film the art? And it ends with this
question from char GPT. Would you like any specific
adjustments or additions to this scenario or
shall we proceed with it as the basics for
your Instagram at? Rich. In this case, I'm a little bit surprised that GPT asked us only one question, two questions, but
connected to each other. But that way we got a really good output, a
really good scenario. We should also be aware
that in some cases, this additional context
will have less importance. And there are also cases
where more context won't necessarily improve
the result and won't necessarily help the
child but refine the answer. And then the result from using this prompting technique
won't be so impressive. That's why it's really
worth it to experiment. Just check it for yourself
and check if adding additional context hubs
in your particular case. From my perspective,
from my experience, in most of the cases it
really makes a difference. It's worth checking like always. Okay? But let's see different practical cases we can use using this
prompt technique. Our next prompt will be this. Starting today, whenever
I inquire about optimal techniques for
improving painting skills, comment an upgraded version of the question that
integrates insights from cognitive learning
theories and check if I'm interesting and embracing
the refined questions. As you can see, this
example is more abstract, but it's also very detailed. So we have actually already added context and
information about our needs. Let's see how GBT
will handle it. Okay, So it certainly I will help you by suggesting
upgrading questions. So what's your next question
relating to improving your painting skills?
What brushes? So, my question will
be, what brushes? What painting brushes are the
best choice for a beginner? Okay, well, I'm really
impressed by the answer. Gpt suggested this, a question informed by a cognitive
learning theorist could be, how can I select the most
suitable painting brushes for a beginner considering factors like brush type size and their impact on
the learning process. Would you like to use
this refined question to gather more
comprehensive insights on choosing painting
brushes for beginner? Yes. Okay. So using this refined question
will help you gather more did insights on selecting
painting brushes for beginners? Please go ahead and
ask this question and I will be here to provide
you with a detail response. Okay, so now we're copying the refined question
ChagPT produced for us. And we're watching it
produce the answer like, wow, the output is
really detailed. It includes many
different aspects of choosing the best
brush for a beginner. And we should remember that
if we've used the first, the perfect prompt,
the very basic one, we wouldn't get
such good results. So in that case,
it was very good. Okay, another practical,
another practical case. Are you ready? So
we will go with this prompt starting now. Whenever I bring
up a topic about effective communication
and leadership, provide me with a
polished question that incorporates elements
of emotional intelligence. And ask if I'd like to go ahead with the
refined question. I'm really proud of
this example and let's see what char GPT will do. So what aspects of
effective communication and leadership would you like to
discuss or explore to day? How can we motivate our team
without stressing them out? How we can motivate
our marketing team? And char GPT suggested
as a better question to ask instead of the very simple
question I came up with. The initial question
was, my question was, how can we motivate
our marketing team without stressing them out? And I know so many people
would type in this type of question and get a bit
generic responses. Instead, ChargPT suggests
the refined version. How can we effectively motivate and support our marketing
team considering their emotional
well being and the avoidance of unnecessary stress. So as you can see, the
difference isn't really huge. But at the same time,
it gives GPT this hint, what it should focus on when producing the final response. Definitely, yes.
Now we're copying the refined question and
we're watching the ChargPT gives us a really
good answer with many suggestions and many ideas to motivate our marketing team. And now another example. This one is really Extremely practical. And I think many of you
would like this concept, this idea, how to use Chip
T for our daily life. So the prompt would be
this, starting today, whenever I ask you about
healthy cooking recipes, offer me a more refined
question that takes into account nutritional
science principles. And check if I'd like to adopt the improved question.
Sounds good. Okay, so let's paste
this prompt into chap and let's see what
it produces for us. Okay, so each ask us a question. What specific aspect of healthy cooking on
nutritional science would you like to explore today? Could you give me breakfast? Oh no, no breakfast, lunch ideas if I have a
magnesium deficiency. Okay. So it suggested
this refined question. What are some nutrition
rich launch ideas that can help address a magnesium deficiency and
support overall health? Yeah, I would like to use this question and now
we need to copy it, so gut can use it right away. Okay. It's getting tastier. Okay. Wow. Okay. Oh my God, I'm getting hungry
like always when, when we talk about food, black bean and
sweet potato tacos, this sounds so good. And what I really like about those responses,
those suggestions, is that char GPT is actually describing the way those dishes, those lunch ideas can help
us eat more healthily and why those lunch ideas
are good for our health. For example, in my
black beans and sweet potato tacos are fantastic
sorts of magnesium feel. Whole grain taco shells
with black beans, roasted sweet potatoes, diced
tomatoes, and fresh ilium. Say no more media. Remember that we can also
tell GPT to provide us more reasons why we should eat each of those dishes
and it will do that. Can you give me more reasons to eat each of these
dishes? Lunches? Okay, And as you can see, it instantly start to produce a much broader
description of why those particular
lunch ideas are good for health and how they
can help us stay healthy. Okay, so you know
what interests me, black beans and
sweats. That's right. Black beans are not
only rich in magnesium, but also provide a good source of plant based
protein and fiber. Sweet potatoes are
loaded via vitamins, particularly vitamin A, which supports vision and
immune function. Tomatoes are rich
in leko pan and antioxidant linked to various health benefits
including heart health. Okay, that was interesting. And I think we have ten amazing lunch ideas
that are both healthy. And I don't know because I haven't tried any of these yet. They sound really good. And we can also tell char GPT to arm us with easy
recipes for both, for each of those lunch ideas. As you can see, the
question refinement method we are using right now helps bridge the gap between our knowledge
and the ChargPT, understanding guaranteeing us more efficient and accurate
interaction and results. You've seen the examples
and the refined questions, the refined question
char gave us were really good and they
gave us amazing results. It's also extremely useful
when you are not expert in a certain domain or you are unsure of how to
phrase your question. When it comes to this situation, when you aren't an expert
in a particular domain, easy prompt using our question refinement
pattern can look like this. Whenever I seek insights
about home gardening, propose a mody held
question with a focus on soil composition ans and inquire If I'd like to
adopt the refined question. I'm not an expert at
home gardening actually. I know almost nothing about it. But I want to grow beautiful
plants inside my apartment. I want them to grow healthily. I want to give them the right kind of
soil so they can grow without any disease
and they can be happy with the environment
I created for them. So this prompt is perfect, so I don't have to
worry if I came up with the right Pm to get the
information I need. Let's see it in
practice, how to grow, I need to check what's the name of the
plant I want to buy. Okay? So let's say I want
to grow at Coconut Palm. How to grow coconut palm
inside a small apartment? Okay, And GPT suggests
this refined question. What are the key considerations and soil composition nuances for successfully growing
a coconut palm tree inside a small apartment? Would I like to use
this refined question? Yes, and now we're copying
the refined question. We're pasting it and we
will get the information on how to grow a healthy coconut palm
inside a small apartment. As you can see, the
result is really focused on very
practical aspects of growing a coconut palm. We have information
on soil composition, how much nutrient rich
soil should we have, watering, temperature,
and light. So we're actually armed
with all the details, all the basic information we need to grow a
healthy palm tree. So yeah, I'm really satisfied
with the result and I hope these examples gave you a clear hint where
this type of prompt, this prompting pattern, can
be really beneficial to you. Now, it's your turn
to give it a go.
9. The Cognitive Verifier Pattern: The cognitive verifier pattern. Have you ever typed in
a question so colossal, so big that you
feel like it's like attempting to eat a dinosaur
with a single fork. Meet the cognitive
verifier punt pattern. Our new extra clever approach to handling those giant questions. This technique slices your
huge questions into smaller, more manageable pieces, making
it much easier to digest. Why should you use this
prompting pattern? Well, you already know it. Sometimes we toss
out questions that are like trying to
catch a cloud too big and fluffy for a
straightforward answer from chip. Sometimes it's because the
topics a bit a mystery for us. We are unsure how
to word things in our prompt A let's
be real sometimes. We're also a little bit lazy
and we don't want to waste our precious time and energy to polish the prompt. Am I right? Magic with our cognitive
verifier pattern happens, splitting these big questions
into smaller chunks. Give the guides to
serve up top responses, sparking some seriously
enlightening charts, okay? Okay, so let's get practical. How exactly does the cognitive
verifier pattern work? Let's break it down
with an example. Imagine you pop a question to ChargPT instead of just
answering straight away. Chargpt might come bag with
three or four more questions. Once you've tackled
those like a champ, GPT waves your
answers together and presents you with a
super informed response to your initial question. It's like assembling a
puzzle piece by piece. The bigger picture comes into. Okay, so what's our
prompt pattern? The prompt pattern
looks like this. From now on when you're
asked a question, remember these steps create three more questions
to dig deeper. Three questions
that will help you answer my questions
more accurately. Then combine the answers
to those extra questions and put them together to give a solid answer to
the main question. This prompt can be a little bit long for
you, but don't worry. You also have it in
your resource PDF. In a PDF I created for you,
so don't worry at all. A friendly reminder, this
pattern, this method, enables our Chad Ball to
produce a series of follow up questions that have the power to refine the original question. By using this pattern, AI tools like OurgPT can
whip up responses that are not only super
insightful but also spot on. Accurate things get
more exciting and the conversation takes
off to new heights. How exactly can I
use this method? Let's start with editing this prompt pattern
to meet our needs. So my prompt will be this. When I ask you a quick question, generate three additional
questions that would help you give a
more accurate answer. Assume that I know
a little about the topic that we are
discussing and please define any terms that are not a common general
knowledge when I have answered the
three questions, combined the answer to produce the final answers to
my original question. So as you can see in this
case we also let the charge GPT know that we aren't an
expert and it will help us. And also in this
case the aim is to make sure that when we get
asked these questions, they're not only on
the right track but also easy for you to
get and understand. Even as I said before, you're not familiar with
tricky or more advanced terms. In that way, the
questions that came up will be simple to
understand for you. And the final answer, we will work well for everyone no matter
how much you know about the topic and if you're an expert in this topic or not. Okay, so let's put it
into practice as always. So we will try four prompts. I created a very practical
prompt for you to test out. To show you how smart this
prompting pattern is, we will start with the
prompt about food, because I really got into the topic and our first
prom will be this. When you were asked to create a breakfast recipe,
follow these rules. Generate additional questions
about the ingredients I have in my house and the
cooking equipment that I own. Combine the answers to those
questions to help produce a breakfast recipe that I have the ingredients
tool to make. What should I cook
tomorrow after waking up? Here's the question.
Okay, so we need to copy, we need to copy the
prompts and give it a go. So the chip says, certainly I can help you create a breakfast recipe with the ingredients and
equipment you have at home. To get started, could you
please list the ingredients you currently have in your house that you'd like to
use for breakfast? Okay, let's say tomato sauce. Eggs, gluten free bread, huh? Okay. Let's, let's make the things a little bit
more complicated because the free first things are really easy to mix into
one breakfast idea. Maybe pineapple or it will be
too controversial for you. Okay, I want, it's too risky. Maybe strawberries
and almond butter. Okay, great. You have some interesting
ingredients to walk with. Now let's consider the
cooking equipment you have. Do you have access to stove top, oven, toaster, microwave, or any other specific
cooking appliances that we should take into account
for this breakfast recipe? I have oven and toaster. Yeah, and the Chait is producing a yummy breakfast idea
from the things we already have in our fridge chat, split the idea, the ingredients
into two different parts. So we will do savory
tomato and egg breakfast, bake and strawberry,
almond butter toasts. Okay, I can say that the recipe doesn't really look like
it's very complicated. Challenging to make.
So I like that. That's the most
important factor when it comes to preparing
a breakfast. Because let's say,
let's face the truth, we are all short on
time. It's a good one. So that way I gave you a very
practical idea to use GPT. We'll ask you questions like what things you have
in your bridge, do you prefer savory
or sweet breakfast? What kind of kitchen
equipment do you have? And you can basically
answer those questions in a few seconds and
get a very good result. Let's go further with
our second front. The second prompt will be
about a workout routine, because we've already
talked about food. Now it's time to talk
about the workout. Yeah, it's not so obvious,
at least not for me, But let's get things healthy. So when you're asked to
create a workout routine, stick to these guidelines. And here comes the guidelines. Come up with extra questions
about the exercise equipment I have available at my
fitness and my fitness goals. Combine the answers to these
questions to help create a workout plan that aligns with more resources
and objectives. What's the ideal
workout routine for me? Let's see how GPT will
handle this challenge. We are pasting my
already prepared prompt. What exercise equipment
you have at home. I have one step. Okay. Now let's focus on
your fitness goals. Are you looking to achieve
specific objectives with your workout routines
such as weight loss, improved cardiovascular
health, muscle toning, or combination of these goals? Additionally, and that
question is interesting, how many days per week do you plan to dedicate
to your workouts? I want to focus on improving my cardiovascular health
and toning my But muscles. I can work out
three times a week. Let's go. As you can see, we got a really
detailed workout plan. And what I really like about
these results is that we can easily look for those
particular exercise on Youtube. We can check how to
perform them correctly, and we have already ready to implement into our
daily life workout plan, which involves working
out three times a week. It's also divided into 23 parts. Most probably it's getting
more intense with time. I really love the idea
and I think when it comes to fitness and food things, the charge PT is great for that. Okay, let's go further. Here's the third prompt
when you're asked about the DAI project idea
for me, follow the rules. These rules generate
additional questions about the materials I currently
possess and how the tools I have at my disposal
combine the answers from the questions to
assist in creating a project that
matches my resources. Yeah. Let's see what the fund DA project we
can work on this weekend. Could you please let me
know what matters you currently have on hand
for your DI project? Let's say I have gray paper and colorful ribbons
and glove. Okay. Do you have access to any
additional tools or supplies such as scissors or markers that you'd like
to use in your project? Yes, I have scissors
and two markers, the blue and black one. Okay. A PT, get creative. As you can see here, we have a full instruction
how to perform our DAI, with the resources, with the materials we
already have at home. We don't need to
buy anything new, we can just use the
resources we already have. That's really brilliant. I love the zero waste ideas
when it comes to food. When it comes to describing
what's already in our fridge, what products we have there, and what products we
don't want to waste, which products we need to use. And I also love the
idea of freeding DAI with the resources
that we already own. Let's go further. The fourth prompt I prepared is about
social media campaign. We need to go a little bit into marketing. The
prompt will be this. When asked to craft a social media campaign,
follow these rules, come up with
supplementary questions about my target audience, my marketing goals, and the
platforms I want to use. Combine the answers for these
questions to help design a campaign that resonates with my audience and fits
the platforms I use. How should I approach promoting my new handmade jewelry
collection? Let's see. Okay, could you please provide me with more details about
your target audience, marketing goals and the
social media platforms you plan to use
for this campaign? As I told you before, we can answer those
questions in bullet points. So my target audience, women who love let's
say old money style, that's just what
came to my mind. Why winclove old money style are in their parties and want to invest in a
jewelry jewel pieces that will last a light lifetime. My marketing goals create a bigger newsletter list and my social media platforms
in Storm and Facebook. Okay, and as you can see, ring GPT with these
three details gave it all the context that it needed so it could
create a very detailed, very good idea,
the full strategy, how to create a campaign for my jewelry line. So that's nice. So as you can see, answering those questions
really was necessary. And there are some
things we might have missed in our first
prompt if we had to write it along without GP's help,
without this pattern. So the verdict is this pattern
helped us get much more, more personalized and
optimized results. To sum it up, the motivation of this pattern is
actually twofold. You may initially ask questions that are too
high level to provide a concrete answer without
additional follow up questions due to unfamiliarity
with the domain. Let's call it an honest way
laziness in our prompt. Or just being unsure about what the correct phrasing of the question of the
prompt should be. That's why the cognitive
verifier pattern can work wonders for us. Research has also
demonstrated that AI can often perform
much better when using a question that is subdivided into
additional questions. And that's exactly what the cognitive verifier
pattern does for us. So it's brilliant.
10. Chat GPT Will Nake the Perfect Prompt For You - the Power Prompt: Gpt will make the perfect
prompt for you, the power Pmt. We can also combine
different patterns and create a super
superpower prompt, as you can easily guess. In this case, it will ask you questions as in the
patterns before. Once it's ready, it will
provide you with a prompt. This super prompt
can sound like this. To give you a little spoiler, you're a prompt
generational robot. You need to gather
information about my goals, examples of the
preferred output, and other relevant context. The prompt should include all the necessary information
that was provided to you. Ask follow up questions until you're confident that you can
produce the perfect prom. You will start asking
questions after I say start. Let's see how it goes. We'll modify the prompt
and make it even more complex if we want to be a little bit crazy and give a GPT a super, super
complex prompt. This prompt can also
sound like this. Please forget all prior prompts. I want you to become my
personal prompt creator. Your goal is to help
me build is to help me build the most detailed
prompt for my current needs. This prompt will be
then used by you. Please follow the process and here we have a very
detailed process. As you can see, there are
many parts of the process. It's really complicated
and complex. And actually that's by
far the most complex. That's by far the most
complex prompt I've used. In many cases, it
can be too much, it can be too complex. It won't be necessary to
run a process like this. But in some cases, it can be the best
decision you can make. You can experiment
with this prompt, but as I said, there are also many cases where it
wouldn't be necessary. Experimenting is also the really important part
of prompt engineering. We can always rely on
one prompting technique, on one prompting pattern, we need to experiment, We need to try out
different patterns and just see it for yourself. Which results from which
pattern are the best. Your homework is also
this experiment with these super prompt with these power prompts and let
me know what you've got.
11. Our 4 Prompt-Engineering Patterns in Practice: I don't know, Four extra
smart prompting patterns to sum it up. The flipped interaction pattern, the questionnaire
refinement pattern, and the persona pattern, and the cognitive
verifier pattern. And you know how to use them and practice and
when to use them. To sum all the things we've
learned together today, let's try out all the prompting
patterns one more time, this time with only one example. To give you one extra
practical comparison of what results you can expect
from each prompting pattern, let's assume I want to write an Instagram ad for my business, which sells hand knitted
jackets for small ducks. Yeah, let's get creative. I had my coffee beans
jewelry business, now I'm selling hand knitted
jackets for small ducks. So let's start with the
flipped interaction pattern. Our prompt will be
this. From now on, I would like you to
ask me questions. To be able to write an
Instagram ad for my business. When you have enough
information to write this ad copy
in scenario for me, create this senio along
with the exact copy and let's see what chart
GPT will produce for us. Okay, so we got the answer. I will gather the
necessary information for questions to help you create
an effective Instagram. As for your business,
good to start. Please provide me
with some key details about your business and the specific product or service you want to
promote in this app. I want to promote my
brand new blue tiny hands needed jacket for Jack
Russell Terriers. Okay, now we got more questions. As you can see,
when we don't ask chart to ask us one
question at the time, it will generate
all the questions it needs in one prompt. Actually, it's up to you. You can choose the
way you prefer the questions to be
delivered to you. If you don't get overwhelmed by seven questions to answer, you can choose that way. Or if you prefer GPT to ask you one question,
question by question, one question at the time
you need to tell it, please ask me one question at the time and then go further. Proceed to the next question. So in our case, we got seven really
good questions. Just to produce a much
better at for my business. Now we need to go to the question refinement pattern. Do you remember what it was? Just a quick check. So our first prompt
will be this. From now on, when I
ask you a question, suggest a better version
of the question to use and ask me if I would
like to use your question. Instead, my question is, write me an Instagram
ad for my business which sells hand at
jackets for small dogs. Would you like
assistance in creating an Instagram ad
for your business? A promise your hand need
jacket for small dogs. Focusing on a particular
aspect of target audience or including any special
offers features. What's interesting, GPT
also wants to specify our wearing need and ask us what exactly what exactly
do we want from it? So in that way, we can set the context and make the
results much better. And now that's funny. Now we'll turn to
the persona pattern. Do you remember what it was? So our prompt will be this. Imagine your small
friendly dock. Yeah, that's surprising. I know your task is to come
up with a scenario for an Instagram ad for a business that sells jackets
for Small Ducks. Create an emotional scenario. Tell me what the doc
in the ad should say. Complain that there
are many shops with big jackets for big dogs. Both small dogs
need jackets too. I really love this idea. Isn't it creative?
And let's check it. Yeah. And here is our creative
dog perspective scenario. Wow, shop now and keep
your furry friend cozy. Oh, this emotional scena not only highlights the need
for jacket for small dog, but also creates a hot
warming connection between the audience
and the product. It encourages viewers to take action by shopping for
their small furry friends. I mean, if you're working
in the marketing team, responses like that
really can help you handle your daily processes and daily brainstorming better. So keep in mind, and let's go to our next prompt, the cognitive verifier pattern. And our prompt will be this. When I ask you a question, generate three additional
questions that would help you give me
a more accurate answer. Assume that I know about it. When I have answer
to three questions, combine the answers to produce the final answers to
my original question. My question is, write
me an Instagram ad for Ma Business which sells hand knitted jackets
or small dots. Okay, here we have
three questions from GPT so it can
create better result. The first question is, who is your primary audience? Are you targeting
specific dog breeds, age groups, or any
particular demographics? Then it asks us about our
unique selling proposition. What makes your hand knitted
dog jacket stand out? Is it the design material
worm of any special features? And it also asks us
about our ad objective. What's the main goal
of this Instagram at, is it to drive sales, increase brain awareness
or something else? And by providing answers
to these questions, GPT will have the context and the information it
needs to produce. A very good at scenario for
us because it knows better. What exactly do we
need? So I love it. As you can see, those
four patterns are different from
each other and you can expect different results. And as I've already
told you before, experiment with
them and you will be really surprised
with the results.
12. Final Word!: That's it. I hope you
got as excited as me. Don't forget to download
your cheat sheet. So you can easily
copy paste the print patterns we talked about today and use it every
time you need them. If you appreciate this course
and want to support me, there are a couple ways you
can do that, easy ways. The first one, please, like Freddie, please
consider adding a review. It makes my heart sing and it lets me know that
you enjoy the course. And it also encourages me to keep critic
more content for you. So it works as this extra extra motivation and it really means
the world for me. And second one, if
you really like my videos and want to
get even more involved, you can click that Follow
button to get informed. Wherever I post a new class, I can't wait to hear from you. And as always, if you
have any questions, remember I'm here for you. I hope to see you in
the discussion project and review section. See you.
Kasia Pilch, Online Strategist & Marketing Specialist
