All language subtitles for 10-BONUS - Mastering Prompt Engineering

Thank you Mark, and thank you,
um, everybody for being here.

Um, but first I gotta say this, it's great
to see so many of my friends on here.

Hey everybody.

It's good to see you guys.

Nice to see y'all.

Nice to see y'all.

Thanks for being, being here.

Um, as we dive into this and also,
man, I've gone through the content that

y'all have gone through and it has been
amazing, like, Just amazing all this

stuff that you've learned so far and,
and so I'm, I'm super excited about

that, and I'm super excited to be a
part of it and take it the next level.

Okay.

Uh, let me figure out
sharing my screen here.

Now, I, I want to be completely
honest with you guys.

Can y'all see my screen?

No, probably not.

Hold on.

So one of the beauties here is when you
are an old school geek or even a marketer,

you might be good at those things, but
you're not always good at things like

hitting the share button when it's time
to share screen, which is what I just did.

All right, can y'all see that?

Can I get a thumbs up if you all can see?

See the screen?

Fantastic.

Okay.

Um.

So I'd like to start, start here by
saying, um, my team knows, I don't usually

talk a ton about myself, but I do wanna
give a little bit of a background today

as I start to kind of frame some of this.

Um, I started coding when
I was in seventh grade.

In ninth grade I took a class called,
uh, analytics and Data Structures.

And in that class, Mr.

Yee, over the course of a year took
us through a process by the end

of the year where we had a test
and had to produce a program that

he would give us a set of data.

Right.

It was just this big set of data and, and
we call it big cuz we were 14 years old.

It was like 50 terms.

It was this big set of data
and he then would say, okay.

With this data, I am going to give
you a term and I want you to write

a computer program that would
predict which term I'd say next.

Now, we did that by then also having
other data that said, Hey, oftentimes

dog followed cat, but if you said
bird and then dog or bird and

then cat, then fish came, not cat.

What we didn't know back then was
he was teaching us literally the

super, super, super, super, super
junior version of what AI does and

teaching us to think in that way.

So I, I start today by
saying thank you to Mr.

Y.

For being able to do this
presentation based on something

that he taught me God, years ago.

Okay.

Um, won't give up how long that was.

Um, and we're.

Prompt engineering today.

But first, lemme tell you what my goal is.

My goal is to arm you with some
prompt fag fragments and modifiers

that will help you enhance your G P
T output and answer your questions.

So I wanna answer some of the questions
that you ha have had, um, about.

Prompting about chat g p t about
the responses that you get about

even why that wonderful thing of it
forgets why I do say certain things.

Why is it forget?

I wanna answer some of those questions
to help you understand what's actually

going on so that you can write
better prompts to get better answers.

Is that fair, everybody?

If that's fair And if that's cool,
do me a favor, put, um, in the chat.

Put it's fair.

Let talk to me here if
that's cool, everybody.

Fantastic.

Thank you.

Thank you.

Thank you.

Appreciate that.

Super fair.

I like that.

Super fair.

Thank you.

Okay, now I got a bunch of slides and
I'll be honest with you and tell you

that, that, um, Jeff Hunter last week
touched on quite a bit of this, and so I'm

gonna run through some of what he touched
on, but if you have a question about

something that he said or something that
I say, please stop me, ask the question.

I wanna get to some more advanced
stuff, but I have to make sure that

everybody's on the same page with
some of the baseline stuff that,

that Jeff covered last week as well.

Okay?

All right, so, oh, next thing, before
I forget, I want you to drop some

prompts, so I do want to help you
if we have time and if Mark will

allow us at the end of this all.

Okay, at the end of this all, and
Mark, I think I sent you the link

to the, um, the Google sheet.

Um, if you could drop that in the chat
for me cuz I forgot to copy the link.

Um, I want you to drop the prompts
that have given you challenges.

If you got a prompt you've been working
on, it's not doing what you wanted to do.

Drop that prompt in there for me so
that I can, um, thank you, Michelle.

See, that's why Michelle wears
the Superman hat like me.

Okay.

Um, thank you, Michelle.

All right.

Drop the prompts.

Like, it's like, it's hot for me.

All right, let's start off
with understanding the model.

Um, and I want you to kinda understand.

And take you back so you understand
what is actually happening.

When you type something in to chat g
p t, you type something in and it's

called the prompt and the computer quote
unquote, the computer takes that prompt

and it tries to encode that prompt.

That's a fancy word for saying get
rid of as many words that I don't

need to pay attention to as possible.

That's what it means.

So you put in, how do I
start learning to code?

And the computer encodes that too.

How Learn code,

it cuts all those other characters
out, all those other words out.

It then takes.

What it encoded and it decodes it and
tries to figure out, okay, now what order

do we need to put this in so that we can
actually understand what they were asking?

And then it performs the operation of,
in this case, answering your question.

It gets that, and then it takes that
answer and then it encodes it back

and generates an output for you and
sends that back for you to read.

Does that make sense to everybody?

Anybody have a question about this?

Cause I want you to wanna make sure
you understand just this, we're not

gonna do too much theory today, but
understand just this piece of theory.

If you understand, gimme a
thumbs up in the chat, please.

If you don't understand or you have
a question, gimme a thumbs down.

Ask me the question.

All right?

Fantastic.

Getting thumbs up.

Appreciate that.

Appreciate that.

All right,

so let's do on a couple
frustrations that people have.

I hear it.

People say it all the time.

Um, one of them is, it forgets,
the darn thing forgets.

Um, I was watching Jeff, and Jeff
was talking about how it forgot

his proprietary framework on it.

How do you forget?

Then it asked him again, or he asked
it again, Hey, use my framework,

and it was like, Still forgot.

How did that happen?

Why did that happen?

Well, it happened because in the
model, especially when you're

talking G P T three and 3.54 is
going to be a little bit different.

But in the model, what's going on is,
number one, you put in a prompt, right?

And it chopped out a whole bunch of words
like I just explained it, figured out what

you wanted, and it gave you a response.

Cool, we got that.

But then it, it also looks at it and
says, Hey, I've got a maximum of 4,000

words that I can remember or respond with.

Combination, how much it remembers
and how much it responds.

And so every time you send in another
prompt in what we call chain of thought

prompting, which is what we do in
chat G P T, it has to make a decision.

What am I gonna remember?

And what am I going to forget?

Every time you prompt it
has to make that choice.

And so when you start to say, Hey, it,
it forgot, it's because it's reached

the point where it's got too much
data and it's gotta cut something.

And unfortunately it cut the thing
that you thought was super important.

And so by the end of today, I'm
gonna show you how to get over that.

The very last prompt that I'm gonna
give you today is going to show you

exactly how to, it happened to you today,
Vicky, I'm sorry, but by the end of

today, you'll know how to overcome that.

Does anybody want to know
how to overcome that?

If you do wanna know how to overcome
that, put a heck yes in there for me.

Put a heck yes in the chat if you
wanna know how to overcome it.

Forgetting, yeah, everybody wants that.

Right now.

What about it ignores?

Well, guess what?

The fact that it cuts the majority
of the words that you say,

that's how you know it ignores.

It's designed to ignore, unfortunately,
but it has to in order to, to actually

be able to perform its function.

And we'll talk a little bit about
that and how you can help that,

because there's, there are a few
ways around it, but not that many.

We will talk about that.

Um, what about inconsistent
output structure?

Anybody run into inconsistent
output structure?

Like, you asked something and it gave
you something that was beautiful in

terms of the, the response, right.

Um, sometimes you'll ask, Hey,
give me a, um, an I idea for

I, um, we do video marketing.

That's what my agency does.

And so we may ask, what's it,
give me a video script and it

will give you scene selection
and voiceover and image thoughts.

And then other times it just gives
you a paragraph of pros, right?

We'll talk about that as well.

And then we'll also deal a bit with
the robotic and generic context, right?

Um, as well.

All right, so I got a
prognostication for you.

And all of you fit in
the first line of this.

Smart people are going
to get smarter and dumb.

People are going to get dumber with ai.

And smart people are gonna get
smarter because you're going to

learn to ask better question.

That's all prompts are.

The better we ask questions, is the
smart you'll get and the more you'll

be able to have the system do for you.

No, no.

Michelle, you're one of the
smartest here, so you're fine.

Okay.

Um.

So what the heck is this thing
called prompt engineering?

Well, prompt engineering is the process
of designing and crafting the text.

Yeah, yeah, yeah.

Okay.

That's a whole lot of speak.

It's asking questions.

Y'all Prompt engineering is
the art of asking questions.

Now, it isn't just the, and especially
if, if you're married, it isn't just the

learning how or just asking a question
to your spouse, although technically

it kind of is because Mark, you know,
if you, there are certain ways that

you ask a question to your wife,
you're gonna get a certain response.

True or true.

Yeah.

Very, very, very specific way to ask.

Right?

And so true.

And that's prompt engineering is
understanding and which is why

I started with that first slide.

Once you start to understand that the
computer wants to, it wants to cut out

most of what you say, then you have
to start to, you start to understand,

okay, what's really important to it?

And as you start to understand
that, then you start to get better.

Responses or in, in our case, for
those of us men who are married.

Cause women never have this problem
when, uh, with their husbands,

men who are married, we learn
how to ask questions better.

Women ask perfect questions
every single time.

Right Mark?

All right.

Um.

We're master destruction,
uh, the first thing.

And, and so again, here in this
first set of slides that we're

gonna go through, um, we're going,
yes, we asked Mark, right Mark.

Um, we're gonna go through
some of the basics here.

Now, as I go through these basics, I'm
going to kind of go through some quickly.

If you've got a question, please,
please, please raise your hand,

get my attention, um, explain it.

Um, Actually, we're gonna talk about
explaining it a different way, by the way.

Um, raise your hand.

Get, get my attention.

Mark.

Is it okay if they come
off mute and interrupt me?

I don't mind.

Do, do you mind?

No, no, I can do that.

Fine.

Great.

So y'all have permission.

You can come off mute and ask
me a question if you need to.

Okay.

Um, to get my attention.

So, Number one.

First thing is what is GT's
role in this conversation?

So we are gonna tell it act
as a, or I am a what are you?

Okay.

Act as a YouTube copywriter
that knows video seo.

So here we even took it
a little bit further.

Sometimes people just say,
Hey, act as a copywriter.

No, no, no.

I was specific.

I am a YouTube copywriter and guess what?

This is what I know.

Why do we do that?

We do that because as these models
get bigger, they know and know

and know and know more stuff.

Some of us here are old enough to
remember going to the library, and

when you went to the library and you
walked into the library, you asked

the library and you, you said, Hey,
I'm writing a book report on x.

And the librarian said
that's upstairs to the left.

In this, this section would be
the reference books on that topic.

So what we're telling G P T
here is, Hey, this is who I am.

Go reference that data
that's up there on the left.

Ignore everything else for now.

Go reference that.

That's why act as A is so important.

Okay.

Had a librarian, but they had computer.

Yes.

Yes.

Uh, you, you had a librarian.

My librarians didn't have computers.

They, they just had the um,
um, Dewey decimal system,

which I still don't understand.

Um, number two, who are you speaking
to, especially in marketing?

So we're creating content.

A lot of what we do here is.

Content creation of some sort.

We do video content.

We're always ideating.

Yes, microfiche, we're always
ideating through, uh, video

content for our clients.

Now when we're doing that, we are
sitting here thinking about our

client and their ideal customer.

And so it's important that you understand
the perspective of the content that

you're creating, as we all know,
and, and we had some early sessions,

um, in this James at a, an amazing
session on, on having chat G p t even

help you figure out your customer.

Um, but.

You gotta tell it who your
ideal customer is, right?

My ideal viewer is business owners
who want to hire virtual assistants.

Okay, great.

And then what do you want it to produce?

A blog, a social post, an email.

A video, an ad, a, a video script.

These is the basics which I
think most people, given all that

we've covered so far, most of us
understand most of these already.

Fair point, true or true?

Drop it in the track.

True or true?

Then obviously, what's the topic?

What's the topic?

What's your question?

Okay, what's the question?

So these are all the basics
that you understand so far.

All right, now let's start
talking about desired outputs.

Um, this one here I love because
we, we also do use this for emails.

Um, I don't know about y'all,
but email is one of those things,

and Mark, cover your ears.

Email is one of those things that we
struggle with, like we can ideate through

video content like nothing but man to
come up with what to write in an email.

We're like toast.

Absolute toast.

And so we've used Chat, G P T,
and OpenAI to help us with that.

But now you may write a beautiful
email, but then you still

need a subject line for it.

Right, and so here we would
give it examples, so we have

a prompt that produces subject
lines and we give it examples.

Where did these come from?

Yeah, these are digital marketer
subject lines because, well, why digital

marketers creates great subject lines.

I know because I'm always opening their
emails for products I already owned,

and it's like I already have that.

I don't need to buy it again.

But the subject line was so great, I just
wanted to read it and so you can give it.

Examples of what your
desired output would be.

Now what about structure?

You've gotta dictate your structure.

If you want a list, tell it.

You want a list?

If you want a nested list, tell it.

You want a nest, a list.

If you want to create a table, tell it.

You want to A, okay, one of the
ones, we use a ton for social

posts because nobody wants to sit
and read a two page social post.

That's all in one paragraph.

No, you don't want that.

You want a social post,
especially on LinkedIn usually

are one sentence per paragraph.

Well, if it produced great content
and I still had to go in and hit enter

a bunch of times, okay, I'm lazy.

Don't want to do that either.

So you can tell it.

Each sentence is a paragraph
and it will format that for you.

Going back, um, two slides ago,
if you wanted to give it structure

examples, you can do that as well.

You can say, structure it in
the example of, and give it the

example that of how you want the
structure and it will return that.

Um, anybody have any specific structure
questions before I move on this?

Cause I saw somebody ask
about it a little bit earlier.

Actually, we did have a couple
questions if you wanna do that now.

Okay.

Uh, let's see.

We have nd saying, uh, what
sections changed between your

first talk and this talk?

Maybe just wanna, you know, kind of
go over where kind of you started.

So, So that's a great question, Andy.

Um, and, and here's what changed.

The first eight slides are all the same,
and in terms of when I'm numbering here,

um, then I think when we get to about
number 10, it starts to changing and

then the end of this is very different.

Nice.

Uh, and then if you wanna look at,
uh, can you see the Q and As Tiva?

Yeah.

There's a question by, uh, Saprina Boy.

I use prompt now each
time I start in the chat.

Act as a topic.

You have all the information as a topic.

Okay.

Um, so good.

So you're using this to, to kind
of train on who you are, is that

what you're saying, Sabrina?

Yes.

Yes.

So that's perfectly fine, and I'm
gonna give you a better version

of this in the end as well.

Nice.

Okay.

Um, go back to step four.

Uh, mark, did you have Mark se Selzer?

Did you have a question on step four?

Oh, just didn't catch it.

Okay, sure.

Oh, okay.

You probably keep going.

We'll, we'll touch base
on the questions again.

There it is.

Um, what, what is your topic?

Okay.

Um, Jeff spent a ton of time
on this one, so I'm not, I'm

going to zoom through this.

What is the writing model
that you want to follow?

Um, yeah, that was his whole
presentation, so that was awesome.

Um, that, that Jeff did last week.

And then this is one of my favorites,
topical, um, or additional constraints.

So, and we are gonna go a little
bit further than we did in the M

three presentation on this one.

So topical exclusions have you.

Written a prompt to return content.

And you know that there are certain
contents you don't wanna cover in

this prompt or in this response.

There's certain things you don't want
it to talk about, and sometimes those

are the, the most common answers.

Well, guess what?

You can tell it not to.

Okay?

You can literally type in either
one of these bullet points here

of do not include these topics.

Put 'em in parentheses, right?

Because when you put 'em in parentheses,
it lets the, let's the system know that

it's not two separate words, but it
is one word, and comma, separate them.

Okay?

This will help you, especially
when you're dealing with the

issue of, well, my topic's pretty.

Broad or pretty popular, and people say
the same things about it, which leads to

very generic content, topical exclusions
and saying, Hey, I know this is broad,

but don't talk about this, this, or that.

Forces the system to ideate further and
give you new topics, which will take

you out of being just, um, generic.

Okay.

Um, next is, uh, phrase exclusions.

This drives me nuts and I may be the
only one, but please write a blog

post in this blog in this video.

Drives me nuts.

So just tell it not to, if it's doing
something like that to you where

there are certain phrases that you
don't want it to use, you just like

the topical exclusions, you can say.

Don't use this phrase.

Now, what I'll tell you about
phrase, phrase exclusions

is it doesn't always work.

It works, my opinion,
70 to 80% of the time.

Okay?

Um, especially the, in this video, if it,
if you're writing a video script and it's

a long video script, it may not start the
video within this video, but paragraph

number five here comes that in this video.

Okay.

Um, so phrase exclusions,
tonal suggestions.

Jeff, um, talked about about
this, uh, before, and I know

some other people have as well.

Um, that you can get into great tonal
suggestions, uh, with the, the chatbots.

Now, this is one of my favorites as well.

Let's think about.

And you can take each one of
these exactly as they are.

You don't need to change 'em.

This is how we use them.

Okay, so let's say you have a passage.

Let's think about this
passage using an example.

So sometimes we want it to rewrite.

Well, maybe a rewrite
isn't just what you want.

Maybe you want something that's
completely different and or, or

a rephrase isn't what you want.

You want something that's
completely different at a different

perspective about the same thing.

Okay, great.

Let's think about this
passage using an example.

And it will come back and create an
example slash story for your content.

Let's think about this topic
from a reverse perspective.

Again, the thing that we do at our agency
is a ton of ideation, and when you have

a bunch of clients in your ideating
form, it's so easy to get pigeonholed.

And even in your own brand, it's so easy
to get pigeonholed and only see things

from one perspective because that's the
way you've looked at it forever, right?

You did your I C P, you did all
your doctors and, and you only

think about the world in this way.

These give you the opportunity to see
things from a completely different

perspective and, and sometimes
realize things in ideation that you

would've never considered before.

Okay.

Um, another one is, is think
about this from a bigger context.

So

this helps us with clients.

Have you ever talked to a
client and, and Oh my gosh.

I'm thinking of one right now.

Have you ever talked to a client and
you're trying to understand what they do

because you're, you're trying to create
copy for them or a video, what, what have

you, you you're trying to create content
for them and all they can do is give you

the details of how they make their widget.

Like they're so in depth in what they
do that they can't actually see the

theoretical side of what they do.

Right, and they can't picture bigger
because they're so in the weeds.

This helps you get out of the weeds.

You can take their weeds, stick
it in, and it will take you up.

Didn't go up high enough.

Great.

Do it again.

Run it again on whatever it
produced, and just keep going

until you get that bigger concept.

Okay, why?

Where is that helpful?

Well, guess what those bigger concepts
are usually what you use in the awareness

stage of the CV J four content in the
awareness stage of the CV J for content.

You're not deep diving
in how this thing works.

It's big theory of why they need it.

This prompt helps us a ton with that,
or I should say this prompt fragment.

Because you're taking this,
you're taking your regular

prompt and you're adding this on.

Okay?

Um, let's think about this
prompt using analogies.

Let's think about this
prompt from multiple angles.

Got a few more of these.

Let's think about how does, let's
think about, so these, let's

think about working for y'all.

Y'all, y'all feeling with these?

I if, if these are, are doing
something for you, say, writing the,

the chat, doing something for me.

Thank you, Kathy.

Um, let's think about, about this topic
by looking at its impact on, so this

is another great one right now, right?

Um, I have a client who, he works with
kids and helping kids through transitions,

but he sells to who, he doesn't sell
to kids, he sells to the parents.

And all he can do is tell me about the
transformation that he gives to kids.

And he's kind of doesn't
understand the parent side.

And so we're able to use this to
say, Hey, think about how these

sorts of transformation for kids
will have, what kind of impact

that will have on the parents.

Helps us ideate.

Let's think about this topic
from a historical standpoint.

This works great if you're
trying to tell a story.

Sometimes.

Sometimes this topic has existed in
a different way in history and chat.

G P T pulls that out and all of a sudden
I've got this story I can relate to here.

Okay?

Um, in terms of economics
or the negative perspective.

Now someone's gonna ask
negative versus reverse.

Those are not the same.

Those are not the same.

So on the first slide here, it
said, think about 'em from a

reverse perspective, reverse and
negative produce different results.

Okay?

All right.

Um, any questions?

Let, let's jump in here.

Do a lot of long form content,
specifically help with clients.

Okay.

Yeah, if you wanna read the, the whole
question or I can if you want Lati.

Yeah.

Oh, sorry.

Sorry.

I do a lot of long form content,
specifically helping clients write

12,000 word non-fiction books.

GT four is a godan.

Yes.

It would be GT three and 3.5
weren't very helpful in that sense.

Um, any thoughts on how much even better
it's going to be once they release pl?

Oh my gosh.

Um, so memory's not gonna change
once they re release plugins.

Plugins aren't gonna help you with memory.

Um, that, that was, sorry, that
was the end of of his question.

How, um, things are gonna be
different once they release plugins,

especially in terms of memory.

Nothing's gonna change
with memory and plugins.

However, the thing that plugins are
going to allow us to do is to, again,

getting back to that slide where I say
smart people are going to get smarter.

Your ability to be able to consider now
the content that was created and how I

can then marry it with what this plugin
can do is going to allow you to create

even better and bigger and faster.

Paul, I hope that answered,
answered your question there.

Uh, my friend, um, writing a.

You know, I was trying to remember
Bucket do io earlier today, and I

could not, I just came off of mute.

Is it OK if I ask if I
ask a follow up there?

Yeah, yeah, go ahead buddy.

This is Paul.

Thank you for asking me
for answering the question.

Um, do you, do you see, or are you
aware of any, um, improvements, whether

it's within GT four or, um, and maybe
no one knows this yet, but in terms

of that memory and being able to go
back further, more than 4,000 words?

Oh, oh, oh, I understand your question.

Now.

I misunderstood your question.

Um, however, the answer's the same.

Um, so it's gonna to, it's going
to be interesting here with G PT

four because G p T four, everyone's
like, oh my gosh, 32,000 words.

Now, you know, all, all this wonderful
stuff and, and that seems wonderful.

But if you actually read the fine print,
the prompt side of this is still at 4,000.

Now that's greater than it was
in G P T three or 3.5, where the

prompt and response was 4,000.

So it's gonna be able to remember
more prompts in that sense, but

it's still going to have a limit.

That make sense?

It, it does.

And, and, and for example,
I've, I've, I've really enjoyed

this training so far, and I.

Took from, I forget the person's name,
but he showed us how to create the,

um, ideal client profile, including
demographics and psychographics.

And so now when I'm writing a
book, I'll paste that whole thing

in and, you know, who knows, maybe
it's 1,500 to a thousand words.

Um, do, is, is that the best approach
or do you think that that should

be simplified into something more?

Concrete.

In other words, can't, is it?

Is it really if I say, Hey, here's my
deal, client profile, write a paragraph.

Considering this person in mind, is
that too much information for it?

So I'm going to answer that
question with the very last slide.

Okay, perfect.

Thank you.

Cool.

Hold me to that too.

Okay.

Um, man, writing a marketing quiz
for bucket io need a question

for a quiz with four responses.

Each response should be tied to
one of four specific categories,

no duplicate response categories,
but all four categories should

be included in the question.

So you want a prompt for that?

Is that that's what you're saying, man.

Do we have Nan, you do, I, I
thought I just put yes in the,

uh, in the response there.

So I thought you saw that.

Sorry.

Uh, no I didn't.

I'm sorry.

I'm so sorry.

Um, okay.

We can talk about that one at the end.

Hold, let's hold that cuz that's gonna,
um, we have a prompt that does something

similar to this, but not exactly.

Um, to hold that, let, let
me, let's talk about that.

Cause that, that's gonna
take a little work.

Okay.

Which is fine.

Which is fine.

Okay.

Besides commas, how do other
elements like parentheses, brackets,

and impact, input and output?

It depends on how you, what you tell us.

So, and d ask, besides commas, how
do other elements, like parentheses,

brackets, et cetera, impact or, um,
the input slash output, um, Depends

on what you tell it with chat g p t.

Um, y'all gonna get tired of this answer,
but the last slide is also going to answer

that for you there as well, um, because
you'll see how I'm using them in the last

slide and how I tell it in the last slide
that, uh, a curly bracket means this

versus, um, a square bracket means that.

So you can, you can absolutely tell it.

Um, so that's my answer for the moment.

Uh, also this is Paul.

Now also I see that G PT
four just released templates.

Uh, that seems like it's going to
completely undercut companies like Jasper.

Any thoughts on the direction of open ai?

Open AI seems amazing to me, but
would love your thoughts on it.

Um, no.

It's not going to undercut Jasper,
um, would be my short answer to you.

My longer answer would be I haven't seen
the templates, so that's my longer answer.

Go ahead, mark.

Oh, I was gonna say y Yeah.

Uh, Jasper really focuses
on small business.

So in terms of how it generates
answers, it's trying to generate

answers for business use, which is a
big differentiator from, um, chat G P T.

But the whole system itself,
again, is a research platform.

So my theory is that open AI's entire
business model revolves around licensing

agreements with outside businesses.

So they're not planning
on making money with.

Consumers using the product.

They really just wanna
get those APIs out there.

Cuz pretty soon, once everybody has an
app that's driven by OpenAI, they're gonna

start charging those companies for it.

So that's my theory.

Well, guess what?

They're already charging us.

Yes.

I mean, I pay the bill monthly.

They're already charging us.

Um, yeah.

And, and it's going to be to,
to Mark's point, you know, a

company like Jasper uses open ai.

Right.

And so you, you, you're not gonna really
piss off your customer base in that

sense and compete directly with them.

Right?

Like digital marketer is not
gonna start doing marketing for

agent or for small businesses.

They teach us how to do it.

That's what their job is.

That's what they do.

Right.

Um, all right.

Stephanie, at what point in the prompt do
you use, let's think about terminology.

That's, that's a great question.

Um, Because there, let me, let
me think about how I answer this.

Well, there is no direct
order to the prompts.

If we go back to that third slide
where the first thing it's gonna do

is figure out what words to to, to
cut and then reorder what you say it,

what it's reordering, what you say.

Now, logically, It may make sense
to put it in a certain way so that

it's readable to you as a human,
but the computer doesn't care.

Okay.

Stephanie, uh, Kathleen, would you
please clarify your meaning with regard

to the angle of reverse perspective?

Um, that's a good question, Kathy.

So.

Um, what I'm trying to do is
think of a good example for you.

Negative means negative, right?

So, um,

here

is the Samsung watch.

Good?

No.

Okay.

Let me go a little bit further.

Studies have proven that the Samsung watch
is good for children under the age of six.

Is that true?

Let's say that was your prompt.

Think about this from
a negative perspective.

What type, this is the response.

What type of idiot would give a
Samsung watch to a six-year-old

child That's negative.

Reverse would be where studies have
shown that children under the age of six,

um, Would benefit from a Samsung watch.

The concern exists of what the long term
effects of that would be on the child.

Does that make sense?

Kathy,

as an example, I.

Thank you.

All right.

Awesome.

Okay, Michelle, uh, can you provide
suggestions on how to create calculators?

Ooh, that was a great example
of a calculator, uh, by dm, but

want to learn how to create them.

I'll be honest, I've never
created a calculator.

What type of cal?

Um, do me a favor, put it in the
chat, um, in the regular Zoom chat.

Michelle, what type?

I love this.

Like gimme something I've not done
and make me think, um, yeah, just so

everybody knows, Dean Davis, he actually,
he covers that pretty extensively and

his, and that'll be Thursday don't, or
Wednesday, but let me, let me confirm.

So lemme actually get into more app
development is, is how I looked at it.

Yeah, yeah.

No, absolutely.

So let me also say this one of, and
this is one of the things that I, I, I

appreciate with the, um, doing, um, this,
uh, AI bootcamp and more so creating the,

the, um, the, the Slack channel as well.

AI is collaborative, you
know, for human history.

We've learned stuff and we've kind of
siloed and, and I learned something,

but Mark didn't know it, and I'm not
sure I'm gonna share, share with Mark.

But the point of ai, the, the, the, the
genesis of this all is a collaborative

experience of all of us working together,
learning from each other of I learned

something, I shared it with you.

You learned something you shared.

And we go back and forth in on
this process that there isn't now

just this one guru that knows.

Every freaking thing, it
isn't possible anymore.

Right.

That's one of the things I love about
how, of what you've done here, mark

pulling all of us together because yeah,
I've learned a ton from everybody else

and I can't wait for that one cuz I've
never thought about calculators, so yeah.

I wanna learn that too.

Okay.

We all learned together.

All right, moving on, let's.

Variables, and these aren't
true, true variables.

These are more placeholders
than variables.

And these are designed to
one, help you with some of the

input concerns that we have.

Okay.

Especially in, in three, 3.5.

And if heck, if you're writing a book
in four and you're having some input

concerns, could help you there too.

The purpose of the variables
slash placeholders is to

cut down on the amount of.

Text that you're sending
in with the prompt.

Okay, so if you're going to use the same
term multiple times, then a placeholder

slash variable is the way to do it.

Likewise, being an old school programmer,
it just makes life more readable.

It makes things more readable for me.

Okay.

So that's also secondary
reason why I use them.

Now, the way this works and, and
OpenAI is actually pretty smart

with this cuz I screw it up all the
time and it still gets it right.

Okay.

So you can see here I've got ICP
right on a second line, right?

Right.

Uh, YouTube video description to attract
I c P and then down below I have I CCP

equals and I define who the I C P is.

Now, um, I've tried this both ways of
putting it in parentheses cuz you'll say,

well, shouldn't that be in parentheses?

Because we, we wanted to consider that
whole thing of business owners who

want to hire virtual assistants as one.

In our experience so far, using
the variable slash placeholders,

it treats it all as one and it
doesn't return different results.

At least not different enough
results where it was like, oh, you

completely ignored what I said.

Okay.

Um, so.

Variable slash placeholders help you
with readability, but then also saving

characters, which is the main reason,
saving characters on your input prompt.

Okay.

That make sense to everybody?

Any questions on that before I move on?

Carrie, can you just
repeat that one more time?

Sure.

So variables and, and placeholders.

Okay.

Same thing.

They help you to cut down on the
number of characters that your

prompt is sending into the system.

So it saves characters that
are, they're sending in.

Cuz again, that's wh when we start talking
about how much can it taken and return.

Every time I repeat I c p, all of
those words, it counts multiple times.

Here it doesn't.

If I have to repeat i c p
multiple times in my prompt.

That make sense?

Yes.

Thank you.

You're welcome.

Okay.

What about hallucinating?

Anybody have a hallucinate on you?

Anybody wanted to stop
hallucinating on you?

This prompt will make
it stop hallucinating.

Now it's gonna be in the slide, so
you don't have to go copy or try to

write all of this down right now.

Um, but here it is, where I've
actually only taken the, the

very last three lines of it here.

Cause you the last three lines are
really, what are the workhorse of

making this thing stop hallucinating.

But here it is.

Okay?

And so if you can't
answer because you don't.

Have the information, output, a
clarifying question, and allow me to

respond by providing the information.

So we're telling it.

You don't know the
answer, don't make it up.

Figure out something to ask me to help
you answer the question, then ask it

to me and let me give you information
that you can answer the question.

Right.

Then it tells me it understood.

What's my question?

When is Atiba de SU's birthday?

And now I was, I was really hoping
that this wouldn't work, but it

has no idea where my birthday is.

Of course it doesn't.

Right?

And it asks the appropriate question,
who is this famed Atiba de Suza?

And why don't I know him?

Okay.

I didn't say all that.

This will help it stop hallucinating and
then also tell you how it needs help.

How it needs help.

Okay.

If this was mind blowing to you,
put mind blowing in the chat for me.

It was to me when I first learned it.

Okay.

Y'all ready for this?

Now?

This is, I think this is the second
to last slide, so I know I've, I've

promised a lot in, in the, the last slide.

Um,

We talk a lot about training
chat, G P T and one I.

I somewhat hope that you understand
from what I've said, that the concept of

how we've been thinking about training
chat, g P t is a little bit flawed.

Okay.

Because if we're asking it 5, 6, 7
prompts and making it respond to a bunch

of stuff and hoping that it remembered
everything from prompt number five,

because that was the one where it
finally gave us the right answer, and

then we started doing stuff from there,
doesn't really work because it's going to

chop, chop, chop, interpret, interpret,
interpret, and quote unquote, forget.

Forget.

So how do we help it?

How do we help it remember what
we needed to remember and then

produce what we wanted to produce?

Well, the answer to that question is
actually quite simple in the model, and

I, I don't know why more people don't
talk about this, but in the model, the

first prompt is the most important.

What's unfortunate about that
is for most of us, the first

prompt is the most generic.

So if you saw my presentation in in in
Austin, one of the things that I was

talking about is getting to the place
where instead of thinking about this

as chain of thought prompting, how
do you write the perfect prompt that

produces the result that you want?

The first prompt is the most important.

When it has to decide what to
prioritize, it's always going to default

to remembering what you said first.

If your first prompt was the most generic,

that's why I forget sometimes.

So there's a, there's a, a format to
this, to your first prompt format, macros

and style, format, macros and style.

Okay?

So that, this is the last slide.

Doesn't it answer everything?

It's a joke.

Y'all.

Somebody laugh.

Put, put laughter in, in the,
in the chat bot in the, uh,

I love it.

Oh, let me show you what this looks like.

And, um, in the handouts for today, I
have, um, a Google doc that has this

prompt in it because listen, this is
the prompt here, and let me see if

I can make this a little bit bigger.

This is the prompt.

All of this, all of that.

It keeps going.

It's still going.

Still going, y'all.

All of this, this is
your ideal first prompt.

Now let me break it down for
you and you can modify this.

This is a very, very, very, very
generic version that you can modify.

Okay?

Um, you are my general answer device.

Okay, you told the who, what it is.

This is prompt number one.

Your response will be output number one.

Your next response will
be output number two.

Okay.

Number.

Your response is every time.

It does not always like this very well.

Just being honest.

Okay.

Um, here is the format of my inputs.

PP means previous input prompt from me.

You're doing a prompt.

You wanna reference a previous input,
just told you how you can tell it.

That PP means the previous input prompt.

That's what I'm talking about right now.

PO means previous output from you.

So in your prompt you can type
PO and it knows you are talking

about its previous output

expert.

This is just a term you can use
any term that you want here,

expert means, and then you define.

So think, um, ideal customer.

Think, what am I, who, how, what's
my perspective on the world?

Anything like that you can put in here.

We use expert.

Again, this is a generic
prompt that you can customize.

We use expert because most people
understand what an expert is, right?

Relevant, um, 20 years in the
field, multiple PhDs, and then

we even tell how it, they think
they're, um, they're un orthodox,

less known advice in their answer.

Right, um, style.

When I use the word style, it means use
the following style guide in the writing.

I'm not gonna read all this.

Y'all have it in in in the doc.

Y'all can go through and read,
but it it, it kind of goes through

and tells you tonality stuff, all
metaphors, et cetera, et cetera,

of how we want our style response.

So now think of what Jeff was
doing last week in his session.

Right when he had all that.

Jeff can put all of that right here.

First prompt.

It won't forget.

Then you can define different
terms of what you wanted to do.

Critique.

I want to, for whatever reason,
I wanna scan this material and

tell me a list of issues with it.

Tell me things that are
concerning about this.

You know, you, you write a blog
post and, and, um, you wanna know

that I include everything Again,
it will critique what you wrote.

Anything in parentheses is
the perspective from which I

want you to write your answer.

So I just told it what
parentheses mean to me.

You can put anything that you want.

You can change this to make parentheses
mean, whatever you want it to make.

Same thing with curly brackets.

Anything with, uh, between the colon
is the input to use for writing.

And then it goes through three words
or less and some more information here.

Now we give it examples.

Now we're gonna feed an example.

So here's a prompt y'all in,
um, in square brackets, right?

Is what I want you to accomplish or write.

I want add copy.

What's the perspective, right?

Is expert inside of, um, parentheses
and then it's going to be a Facebook ad.

And here is after the colon.

Here's what I want you to write about.

That's the prompt.

And then if we give
how it would translate.

The question mark symbol
means this example.

So this now says previous output.

I got a question about it.

Will trans will translate into, explain
your reasoning for your last output.

All you typed in was that now
you wanna talk about saving.

Um, characters.

We're saving characters and we're
getting the computer closer to

what it already understands.

We have the minus symbol.

I'm not gonna go go through all of these.

There's, there's a ton, right?

Um, and then we have examples
of how this all works out.

Okay.

Uh, one of their favorites here is
you could take the previous output.

So let's say it wrote something and
you want it to be funnier, or you

want it to be sillier, or it was too
silly and you want it to be serious.

Well, you can give it a
perspective and tell it that.

Okay?

You can even go to the point
of giving it a percentage.

So I wanted 20% more serious.

Here's my goal here, okay?

Because I know this is a lot and
this prompt is I'm not even done.

The prompt keeps going.

My goal, as I told you guys a little while
ago, there is stuff that you can do with

AI that I know that's different than what.

Um, other people know, and we're all
working collaboratively together.

And the purpose of all of us working
collaboratively, collaboratively

together is so that we can
create greater, there is greater.

And this is a, this is a basic advanced
prompt, if that makes any sense to you.

Okay.

When we get into advanced
prompting, this is the, the,

this just scratches the surface.

There's so much more that you can do.

There's so much more that you can learn.

So I'm also going to challenge
you that as you're learning, there

are a lot of people who are giving
you the same generic answers.

Go deeper.

Go deeper.

There's a lot more here to learn.

Okay.

What does that all look like?

All right, so we're gonna write a
video script from an expert opinion.

It's gonna be YouTube short about why
digital market asserts are the best.

Okay.

I like it to include a hook,
scene, selection and voiceover.

So I took the, the basic prompt
and I then I said I want specific,

um, output format as well.

Um, the please write in English
is optional, whether you want

that or not, and it did that.

Okay.

Then I wanted it to be funnier
because that's pretty dry.

Are you tired of sifting through endless
online courses, wondering which ones

are actually worthy or worth your time?

Eh, that's pretty dry.

Mark wouldn't write that, right?

Mark?

Or maybe he would.

No, he wouldn't.

Okay, good.

All right, so stick of scrolling through
endless online marketing courses.

Get certified with digital marketer.com
and watch your marketing game.

Go from zero to hero.

Right.

It goes on.

And then the, the, um, funnier actually
came in, in the voiceover, right?

Don't be like Dave.

They thought he could handle marketing
with just a few flyers and a prayer,

but now Dave's business is struggling.

Don't be like Dave.

Get certified with digital
marketer.com and take your

marketing skills to the next level.

It's time to become a marketing superhero.

Absolutely love that.

It uses superhero motif.

I didn't ask it to, but it it, it did.

So it was cool.

Okay.

Um, all right.

I know I just gave you a lot, so I'm gonna
pause right now and take some questions.

Okay.

Vicky with the limit on the data that
chat g b t has, let's say, ingested.

Um, our topic is something, is something
now that has been growing since 22 to 20.

Okay?

Uh, so there isn't much content in chat.

G b T go to Bing, not, not
bing, go to, um, go to Bard.

That's my answer for you, Vicky.

Go to Bard.

Hopefully when chat g p t comes out with
Ablo browser plugin, that would help.

But, um, right now I'd say go to Bard.

Have you gotten back

you saying something, mark?

No, I was gonna just go
to the list of questions.

Oh.

Uh, What is the benefit of
saving the number of characters?

Um, Harry.

Um, thanks for your question, Harry.

Uh, the benefit of saving the number of
characters depends on what your output is.

If you prioritize your output and need
a lot of output and you're out, you're

starting to reach the output limit.

Then the num, you wanna save as
many characters as you can so that

you can get better and more output.

Okay.

Um, I guess one of the things to also
understand with me, I, I get that you

can write more pleas and all of this.

I'm very much into writing one
prompt, getting my answer and leaving.

I get that, that chat g p t
was created so that it could

be chain of thought prompting.

I get that.

However, chain of thought prompting
is based on single thought prompting

and they just extended the model,
so they're trying to get you to that

anyway, so, I just tried to go there.

Yes.

So you can use ic.

Sorry.

Uh, mark asked, um, I can't
remember to read the questions.

Um, so, uh, you can use the I C P
throughout the rest of the chat.

In the variable example.

Yes, you can.

Harry, you just said by using
variables, so I'm not sure.

Oh, was that the, was that, oh, that was
the extension of your earlier question.