All language subtitles for 10-BONUS - Mastering Prompt Engineering 2

That's where I'm leaving.

I get that the chat GPT 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 ICP.

Sorry, Mark asked, um, I can't
remember, read the questions.

Um, so, uh, you can use the
ICP throughout the rest of the

chat in the variable example.

Yes, you can.

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

Oh, that the,

oh, that was the extension
of your earlier question.

Um, I hope I answered that for you.

All right.

I think, is that the last question, Mark?

Uh, you want to start over?

Cecilia, you're lost.

Would I lose you?

I'm sorry.

No, I was saying start back
in March, the beginning.

Oh, you want to start over, over, over.

Great.

Um, I do see, I see we have a
question, um, from Nan in the Q&

A about writing a marketing quiz.

Need a question for a
quiz with four responses.

Yeah.

So Nan asked that earlier.

I need

to know the specific case and we
could probably sit and, and do that.

And this is part of what, um, Jeff was
saying last week too, like the, one of

the things that's going to happen is
there's going to be a new career for

people who can write effective prompts.

Right.

So, um, then yes, we, so what we have.

Uh, because it will need to be
extended from where you are.

So, and do you want the entire quiz?

Or just question need a question for
I really want the format in essence.

So I've got a quiz.

I want it to make up a quiz question.

I wanted to make up the 4 responses.

There's 4 categories or buckets.

And so, therefore, the response.

I want tied to an individual
bucket, but I want all four of the

buckets represented in the question.

What it tends to do is break
it up into here's a question.

All responses go to this bucket.

Here's another question.

All the responses go to that bucket.

And I want every question.

Have each of the four responses
tied to each of the four buckets.

Yeah, I gotcha.

Embracing that, it just, I have to write
it and rewrite it, and I'm like, geez.

So, it's going to get
confused there, okay?

And the reason it's going to get confused
is, and I can guarantee, I haven't

seen your prompt, but I'm going to
promise you one thing about your prompt.

Your prompt is actually contradicting
itself in some places, and that's

what's going to confuse it.

Okay.

Uh, and that's not anything personal.

Um, Linda and I worked through a
prompt a couple months ago where

she had the exact same thing.

And to her, it was, Oh no, this
seemed like one, one thought.

But then when you actually looked
at how the computer would look at

it, it was started conflicting.

Okay.

Um, the thing that I will tell you
is an example is going to help you.

So you're going to have to, so you're
going to ask it, you not what dropped

your, did you drop your prompt on
my, um, form on the, in the, uh,

in the Google doc, no, I don't.

No, because all I saw was your handout.

I didn't.

Okay, so I finally got it.

But now I'm like, how did I get it?

Because I do this over and
over again, you know, I've got,

because I do, I write quizzes.

And so I know there's a way to
exactly write it and I just thought

you might be able to help with
the language of it of the prompt.

So, write down what you say.

Okay, but but so hold on, but
you figured out the answer.

Yes, several times I have, where
it has done it, but now I'm

like, how did I, what did I say?

How did I ask that?

Okay.

Do you have a version of when it
did it right that you can share?

I probably can.

Yes.

Yeah.

If you're willing to share that with me, I
can help you understand why I did it right

and how to make it repeatable because
that's ultimately what we do because

we represent so many different clients.

We have to figure out
how to take a prompt.

And make it in a way that it's,
it's going to repeat great answers,

no matter who the client is.

I can probably find that, just a second.

So, if you may want to go on to
something else, let me look for a minute.

Um, up, up, up, up, up, up, up.

Questions, Mark?

Uh, let's see.

Oh, let me make sure I've unmuted.

I actually asked a few and
I was muted the whole time.

So anyways, uh, Stephanie has a question.

I may have missed this one on, of the
trainings I missed while traveling,

but have plugin recommendation
recommendations been made.

I keep hearing about one that supposedly
makes it so chat GPT gives current

info, but I don't know what it's called.

So I'm guessing current info
would be, uh, full live info.

Yeah.

Yeah, so the, um, the ability to browse.

Um, we'll get you to current info.

Um, what's hysterical is, in case you
didn't know, Jasper now does that too.

And they released it before,
ChatGPT, and they're Jasper Chat.

Um, and so we'll still see how that
actually works and, and how you

actually can use the data that it
gives you or how, how great it is.

Um, the plugin I'm super excited
about is, uh, I'm gonna sound

like a real old school nerd.

Um, this APR plugin,
that's got me excited.

Sorry, I missed the answer.

Hatiba, did you say what it's called?

Whatever plugin we need so it will look
for information from 2023 instead of 2021?

So, that, that would be the browser plugin
that, that um, ChatGPT is releasing.

Um.

It's not available yet?

It is not.

At least.

It's not widely available.

Few people have it.

I'm not one of those few people.

James might be though, putting
you on the spot, buddy.

It's funny.

James always gets access to
everything like the day before me.

And he's like, Oh, I got this.

I'm like, yeah, I don't have it.

Okay, so I did put that in
chat in question, the Q and A.

So I think that is the one that
is going to get me the response.

But then I gave you an
example of a question

on what it produced with
the responses it produced.

Is there a better way to ask that?

Okay,

so

your prompt was, write a question
on challenges with business strategy

and planning or branding and customer
journey or marketing and sales.

I would rephrase that
whole sentence to start.

Yeah, just tell me and I'll write it
down should be three responses for the

question with tied to one of the child.

But okay, so but you didn't
define you say tied to one of the

challenge areas, but you didn't
define what the challenge areas are.

Yeah, that's the business strategy
and planning, branding and customer

journey, marketing and sales.

Those are the challenge areas.

Yeah, it looks like what you need to do is
separate all this into separate questions.

I was afraid to put a comma in there.

So you think I need to just
say there are three challenges?

And yeah, and the three challenges
are this, this, this, and then

I'll reference that to be able to
answer the question in those areas.

But the whole thing is you should,
this should be like three different

prompts is what I would do or four.

Really?

Oh, wait a minute.

But I only want three
responses to the question.

Yeah, but the way you're producing
it is just, um, clunky is the word.

Yeah, that's why I'm asking
you to rewrite it for me.

Okay.

Um, so you, at the end of the day, you
want to, uh, a question for a challenge on

business strategy, planning, or branding?

No, that's what it.

I want, I want a question with the
response going to one of those three.

I got that.

So before we get to, to that,
you're not giving it the question.

You're asking it to figure
out the question for you.

Correct?

Yes.

And you're then asking it to
figure out the response and

which category it goes to.

Yes.

Okay, so from a logical perspective,
it can do that and it can do

that in one prompt for you.

Okay, the logical perspective is the
first thing is, hey, this is my topic.

I'm looking for a question, just
one question, except I'm going to

say 10 questions, but that's okay.

Yeah.

Start with one.

Okay.

Run it 10 times.

Okay.

This is my topic.

I'm looking for a question
for this for a question.

The question will have
four possible responses.

The response categories should be,
and then give the name of the category

and a description of the category.

And I should say X, X and X, right?

Define X.

Well, X is the three challenges.

Yes.

Writing them out.

The response category should be
branding and blah, blah, blah, blah.

Should be X, so X, X, and X.

Defining what X is in each case.

Yes.

Um, okay.

And that will help.

I'm trying to find
something else for you here.

Um,

for an actual prompt.

And I want to be sure
there's no duplication.

Because once in a while, it'll give me two
of the same one, and then the other two.

And leave one out.

And that's where now the example comes in.

So, after you write
everything that I just said.

Say an example would be and then tell
it question and give the question

that answer one is and give an answer
and said this is this category answer

to is this category answer three is
that category to be very specific in

the response that that you give it.

Does that make sense?

Uh, yes, it does.

Uh, but

I'm afraid what it will do is not
really tie the response to the category,

that response to that category.

I can try it and see if it'll do it.

And as long as I have an example, I think
I can just say, use this example, can't I?

I'm going to say something in,
in, in absolute, absolute love.

Okay.

Um, but it may sound a

Right.

You said, I'm afraid it's
going to do something.

Yeah, but we don't know what
it's going to just try it.

If it does, we'll figure it out.

Okay.

I'm going to drop a prompt that

may help you.

Okay.

You're putting it today.

No, I'm going to put it in the chat.

Yeah.

Okay.

If it doesn't help you, just ignore it.

Okay.

All right.

Um, there anybody else you can have
that to, um, this prompt, what it does

is it actually creates carousel posts.

Is this the right one?

Make sure I put the right one in here.

I think I did.

I think this is the one that creates
carousel posts that, that then

creates, turns a carousel into a pole.

Carousel into a pole.

All right.

It's,

um, I don't think you
actually want that Indy.

I think that would hurt your head
even more than it hurts mine.

Um, actually, no, I think
I did drop the wrong one.

And that one, I think
does just a carousel post.

This one here.

Here, take that one.

Uh, let's see.

Okay.

Um.

Actually, while you're trying
that out, you want to check

on, uh, Stephanie's question?

Sorry.

Yes.

Thank you.

Uh, do we want to help ChatGPT by limiting
the number of words we use in our prompts?

For example, would it help if instead of
typing the word and, we used an ampersand?

Or does that not matter?

So you're not helping chatGBT,
you're helping yourself with that.

Okay?

Because remember that the point
of it is, especially in 3.

5 and 3, you're limited in
terms of the number of words.

Right.

Um, and really it's about characters
more than it is actual words.

But, you know, as humans, we don't
think in characters we think in words.

Um, and so it's an approximate
of about 4000 words.

Right.

Um, so that's not actually going
to when you do that to you.

You limit the number of words so that
it can return more response to you.

That's the bottom line.

And and versus ampersand is not going
to make enough of a difference to care.

Have you counted the words
in your, like, killer first

prompt that's three pages long?

I haven't.

It just works every time.

And believe me, I have some
prompts that don't work every time.

I have, we uh, we did a prompt.

So, real talk for a moment.

One of the things that we're working
on right now is LinkedIn prospecting.

And, anybody use LinkedIn Sales Navigator?

The whole point of, of, A GPT model
is that an AI is that it takes a bunch

of data and finds commonalities and
data that the human eye can't see.

Just the whole point.

And then it predicts what's, what
the next set of responses will be.

Okay.

So we got the bright idea in LinkedIn
Navigator, um, because we always just

get stuck with the, um, the criteria for
Navigator to find the people that we want.

And so we got the bright idea.

Could we write a prompt that took
the profiles of Five LinkedIn

navigator people that are our
ideal customer and then create the

LinkedIn navigator search for those.

Anybody interested in something like that?

Yeah.

We were super interested too.

And we got it to the point
where it will only take two,

it will only take two profiles.

We go to three profiles and it just craps.

Every time, but two profiles is not
enough to actually give you enough

data to find anything worthwhile
that the human eye couldn't see.

So we're still working on that.

And if we get it done, I'll let you know,
let me rephrase when we get it done.

I'll let you know,

but that's kind of my point about why
prompt engineering is, is important

because it allows us and why having
these conversations are important.

Six weeks ago, we would have never
thought of that, but because we're using

prompts to help us ideate our generic
content or not generic, general content

for our clients now, and that's sped
up and given us so much time back.

Now it's allowing us to solve
problems that we didn't even

realize we could solve before.

That's the beauty.

Of what AI can do for us.

Yes.

Actually, I just got another
question from Michelle.

I must have probably in
the chat or something.

Uh, do you start every chat with
the master base prompts or do you

start a new chat and tell it to
refer to the master base prompt

chat that you've already saved?

So that's a good question.

Um, no, I don't start every
prompt with the master chat.

The master chat is for when,
like how Jeff was doing when

he had his different clients.

And you have a chat that's
generating content for this client.

That's when you do it.

If you're just asking a
one off question, no need,

no need to go through all
that because it's a lot.

It takes, it's going to take a good
45 minutes to set it up correctly.

No need.

But if you're going to return to
this for the next eight weeks,

then yeah, it's worth the time.

All right.

Let's see.

We got Vicky too.

Another question, Bard or ChatGPT.

This space is moving so much.

Uh, how the space is moving so much,
how much have you played with Bard

and what and who is it better for?

So that's a really great question.

I'm going to answer it in this way.

Um, so in case you hadn't noticed,
Jasper uses four different

models to produce content.

And if you actually listen
to them, there's actually

five models that they use.

If you're really, really listening, okay?

Why is that?

Because every single model does
something different and does

something better than another one.

So if, if you want to write in my
opinion, just straight writing, Bard wins.

If you want ideation, chat, GPT wins.

We're building an app right now
that is designed to help you with

ideation of video content, basically
everything that we do at our agency,

it's encapsulated inside of one app.

And so we're going to be doing that.

We're going to be using two
or three different models to

help produce content because.

Bard's going to do a better job
writing a video script, whereas chat

GPT will do a better job writing
the, um, the, the, the title, right?

And who knows?

Anthropic may do something else
better than, than the others.

And so we may use two or three
different things to pull together

to create one piece of content.

These are tools.

They're not AI,

like mid journey in your process?

So prompt engineering and mid journey.

I'm not, you know, and, and, um, honestly,
here's why I spend so much time in, um,

written prompt response and engineering.

And I know that world.

And the reality of the fact
that I am visually a dunce.

You just haven't practiced it.

That's all it is.

It's when I need some, I'm just going to
call Mark cause he's super good at it.

I will say that the prompt
engineering definitely has an

application within mid journey.

The big difference between something like
chat, GPT and mid journey or any of the

graphic kind of creators is there is some.

Abstract aspects to it, like the
stylized function on mid journey.

What the heck does that mean?

You know, it's, uh, it's kind of
a term that's thrown around, but

you can't really decide what it is.

So, whereas you could be hyper
specific or at least a lot more

specific with the chat GPT, it's
a little harder with mid journey.

And a lot of those, because it's.

It's an art platform, essentially.

So you have to incorporate
that lack of science into it.

But with that said, there is
a way to phrase information.

There's a way to manipulate,
uh, you know, specific prompts.

Uh, and it does have a few more of
those, uh, kind of tailored aspects

like the stylized function, uh,
that you don't have on chat GPT.

Can I share one last
thing, Mark, before we go?

All right, so showed you
some a little bit abstract.

Okay.

Um, so here's my master prompt again,
uh, big thing that I changed this time

is look at the, um, the expert, right?

So my expert is an expert
in content marketing.

Looks like Clark Kent.

But has a friend, Atiba, who is Superman.

They have run a content marketing agency
and currently are the president of digital

marketer, educate an education company.

Anybody know who that is, by the
way, just a random friend of mine.

Um, the rest of this stayed the same.

Then I asked it to produce a video
script about how to make money.

raising chickens in Texas.

Um, and it wrote a pretty decent,
goofy, silly little script.

And he said, okay, that's
just another silly example.

Let's get, get real though, T,
but we're not writing scripts

about chickens in Texas.

Okay, great.

So what about a blog outline about the
customer journey, customer value journey?

Well, he gave me a blog outline
and by the way, It knew that

this should have eight stages.

We're on the right track, because
it's not one of those four stage

customer value journeys, right?

So it gives me this blog outline.

Okay, great.

What about, um, something,
let's get a little SEO now.

Let's get some, some keywords,
some, and semantic keywords.

How does that work?

Okay, so I'm looking for keywords and
semantic keywords, and the expert for

a blog, and same blog title here, and

it's going to give me some stuff as well.

So you can absolutely use this
for very, very practical things.

As well.

That was my last example.

Awesome.

Well, thank you so much, Atiba.

I think, uh, you know,
the whole format is huge.

Um, people will definitely need to
rewatch this episode if you're not

familiar with, uh, you know, prompt
engineering, but once you understand

it, it makes a huge difference in
terms of how well you're able to

consistently produce the results that
you're looking for, uh, with chat GPT.

So Atiba, thank you.

Uh, where can people find out more
about you and maybe get some more

of that, uh, prompt magic going.

Yeah.

So listen, all seriousness,
I want to help you.

Okay.

I want to answer your questions.

I know you're going to have a lot and
they're probably stuff that I've run into.

Um, the best way to do
that is go to meet Atiba.

That's A T I B I A.

That's my name.

Meet Atiba.

com.

That's going to take you to my LinkedIn.

You're going to go
directly to my LinkedIn.

Once you get there, connect
with me, send me a message.

Let's talk.

Drop me a prompt.

Let me help you.

Love it.

Well, thank you so much, Dipa.

I definitely met recommend, you know,
since this is fresh in your head, that you

go ahead and, uh, try some of these out.

Uh, that way you have them saved
in the chat and you can see if

they work or, or they don't.

But, uh, or you know, if you need
to rephrase, uh, but you know, at

Dipa, like I said, excellent content.

Uh, love that you continue to
develop it since the last time

I heard it, and I am definitely
gonna start employing this as well.

Maybe I'll make a version for.

Uh, you know, mid journey or
those, uh, kind of visual aspects,

uh, but it's all applicable, so
very much appreciate your time.

Oh, my pleasure.

Thank you.

All right.

So make sure if you
can drop your examples.

We want to see your work, the show your
work discussion in the slack channel.

Really appreciate if you
made some of those and maybe

just summarize your results.

See if this methodology really
improve what you were doing.

We'd love to hear those case studies and
also your review of the session itself.

Uh, as a reminder, this call
has been recorded and will be

available within the next 24 hours.

It's should be up, you know, instantly on
the Slack page, but, uh, within an hour,

uh, probably no less, and then you could
rewatch it because you might need to a

couple of times and pause, use the pause.

All right.

On Wednesday, we have
our podcasting session.

So it's going to be Jeremy and he
has a bunch of successful podcasts.

He helped podcasters and he actually
uses a variety of completely different

tools that we haven't even discussed yet.

So definitely don't miss Wednesday.

You're going to learn a ton about
podcasting and about how to really

simplify your podcast creation
process using Jeremy's techniques.

So tune back in, in a couple of days.

And have fun with chat GPT in the
meantime, really appreciate y'all.