How to Leverage ChatGPT & Generative AI

we're told that we should be putting out.
at least one piece of content every.
single day this is how you do it faster.
with AI you find a topic that works so.
here we have hiking boots let's say.
you're Eddie Bauer a client I've worked.
with how do you generate content so you.
find different topics there's a bunch of.
platforms that do this hrfs sem Rush.
Niels Uber suggests that will give you.
great keywords that will rank you go and.
find a topic.
and then you get you go I personally.
when I'm working with clients from.
Vancouver I'll go to Miami and search.
for the same keywords and see what.
contents ranking I then take that.
content and I put it through it an AI.
paraphraser so a lot of the AI tools.
that you guys know and love craft like.
quill bot does it really well they have.
a one-click paraphraser so no need to.
spend hours writing that blog every.
single day you could literally go get a.
high ranking one paraphrase it add your.
keywords add your little formulas for.
success and have it good to go then you.
have ai doing email Outreach so here are.
the results for one of the campaigns of.
a client that I have all of these.
keywords a lot some of them with super.
high traffic are coming from the blog.
and now imagine these results replicated.
over 30 days in a month we do this every.
day and it takes us five to ten minutes.
per day.
now are a program prompt programming how.
to get Ultra specific spot on results.
from AI rather than the generic.
repetitive stuff that you see.
I'm going to go through five prompts and.
then go through my personal favorite and.
what we've built our whole company on.
first prompt is I want you to act as.
this one's one of the most common so not.
my personal favorite but you start out.
by saying I want you to act as like a.
dermatologist or an NLP expert and that.
really sets the stage for telling chat.
gbt you want an expert to write about.
that topic then you do I will which lays.
out the relationship between you and AI.
so I will give you a product and a.
Target customer and then the last part.
of this is you will and this is what you.
are going to provide the AI so you or.
what you want the AI to generate so you.
will choose the product slogan or.
channel for promotion.
then the last part is laying out a.
specific response so you telling it okay.
this is my Target demographic this is my.
audience.
now the second one I'm going to show you.
is when Chad gbt provides a prompt so.
this one's a fun one you get Chachi EBT.
to interview you and give you a prompt.
based on that so ask me questions to.
solidify a woman's clothing brand this.
is an example.
and the fun thing about tattoo BT is you.
could then ask it follow-on questions so.
you could lay out you could use the.
first one where I want you to act as an.
expert and then you could keep going and.
asking it questions based on that.
now this is we're starting to get into.
the better prompts these are more of my.
favorite because you're getting so much.
more accuracy but feeding it data so.
let's say you are a real estate agent.
and you sold 150 homes over asking you.
could give it this data and tell it to.
provide you different outputs based on a.
like in the format of a blog or an ad.
and I'll go do that.
now we're getting into few shot learning.
this is when you give it a couple.
examples.
here you're getting better it's getting.
more accurate because you're showing it.
how you actually want it to perform so.
what I always tell people is do you just.
want an email or do you want an email.
that mimics one of the highest.
converting emails of all time so we're.
constantly trying to beat out the.
general public of people that are using.
this tool and the one way is to give it.
great examples.
here is a graph of the results of a test.
that they did on these students they got.
AI to answer questions about math and as.
you can see the accuracy when you're.
giving it barely any examples it's like.
being like a mirror 20 that's just like.
designing nice prompts but then once you.
give it examples it shoots up there.
fine tuning when you want to give it.
larger subsets of data which had gbt you.
could give it a couple examples but.
there's that limitation of how much you.
could give it so if you really want to.
get technical you could go to the source.
the GPT 3 playground and you could.
actually train it with hundreds of.
examples from your industry and this is.
when you're getting that super.
customizable results where it knows your.
brand your style your voice everything.
about you and it's actually giving you.
outputs based on all those formulas that.
we know and love.
here is a graph showing you with the top.
line that's the latest from open AI.
That's the model that we're currently.
using as you can see xeroshaw is just.
giving it a fancy prompt one shot is.
giving it one example fuchsia is where.
we get crazy amounts of accuracy and.
then fine tuning is even Beyond this.
where you're giving it a hundred.
examples that's where you're going to.
get that next level of content.
so takeaways from here ask for tons of.
variations the more the better and.
actually when you're finding an AI.
copywriter that you want to use for more.
technical results try to find one with.
reinforced learning so with basically.
when you're liking favoriting when.
you're then taking it into their text.
editor it's actually using that data and.
feeding it back into it I know some aui.
writers do that and that means that it's.
fine tuning on your data that's.
different from chat gbt but based on the.
same technology but it's different.
because chat gbt you go to it and each.
different time it's not able to remember.
some AI writers have gotten around this.
by fine tuning based on whatever you're.
using.
10 examples pull from a variety of.
different sources so I know we were.
building a LinkedIn post generator for.
our client and we fed it all the best.
business entrepreneur LinkedIn posts.
that we could possibly find the results.
were terrible it sounded like a.
marketing bro like it did not make sense.
at all and then we tried feeding it.
different samples of data from like.
political LinkedIn posts or different.
LinkedIn posts from different Industries.
and even though we weren't using that.
for that use case it was so much better.
because it was pulling from a variety of.
different data and styles and giving you.
so many more options.
now one of my favorites is go straight.
to the source so gpt3 playground if you.
can get technical that's when you're.
going to be able to really customize.
your attach EBT.
continue to refine based on the output.
so make it funnier uh when you're.
talking to chat gbt you could say.
summarize this or expand or provide me.
more outputs.
another big takeaway is beware of data.
contamination so one time I had this.
marketer training a tool that was super.
enthusiastic all the time and she put.
exclamation marks on everything and I.
kept seeing the outputs of the tools and.
it was non-stop exclamation marks and I.
realized we were fine-tuning it but.
because she was fine tuning it the data.
was contaminated by too many exclamation.
marks.
and then lastly chain training to hack.
long form this is one of my personal.
favorites so when you're dealing with.
long-form copy you ask chat gbt or the.
technology to generate a 5000 word essay.
the quality is going to taper off over.
time and so this one is hacking it where.
if you ask it in steps or if you use an.
AI copywriter that basically with a Blog.
generates the headline first and then.
from the headline generates the intro.
then generates the outline it takes a.
couple more minutes but you're getting.
that like short form results with long.
form.
[Music].
[Applause].
[Music].