If you use ChatGPT before, but you weren't that
impressed with the results, you're gonna love this video because this really takes it to the next
level. Automatic GPTs that work by themselves and get your results without you having to
interact with it. That's the topic of this video, and what they provide is the following:
usually, if you interact with ChatGPT, it's you and ChatGPT. You give it commands, it
gives you results. And I find that most people that are unimpressed with the results didn't give
the best commands to begin with, or they might have needed internet access to complete the tasks
they were looking for. The apps that we're about to look at solve both of these, kind of. And I
say that because they're anything but perfect, but they're fascinating, and for certain use
cases, they're actually really useful already. So with that being said, it's time to sit back and
relax because I'm about to lead you for the world of automatic GPTs and what is possible today.
Okay, so first things first, there are many of these, and I would put them into two
categories. There are the ones that run locally, where you kind of need the basics of coding to
run them. Auto GPT and Baby AGI fall into this category. And then there's all the web interfaces
that have been popping up over the past few weeks. That's the ones that we'll be looking at this
week because realistically, for most people, these are the only ones that you're gonna end
up using. Now, there are different ones. They come with various limitations and capabilities, so
we're going to talk about that in this video. But what all of them have in common is that there's
the settings icon, and then you have to link up your API key for this to work. You can find your
very own API keys here. And before we begin, I would strongly recommend you set up usage limits
under billing usage limits, and then set these up to whatever you're comfortable with because it can
happen that you forget about these. They just keep running and they'll drain your entire account. And
if you want a step-by-step tutorial, I did a video on Agent GPT a few weeks ago where I showed you
exactly how to hook it up and how to start using it. But essentially, once you have your API key,
you're good to go. And now we can start exploring. Okay, so we looked at Agent GPT as soon as it came
out, but now they upgraded it. So you can see they added different GPT personas that are particularly
good for one task. They also added web search, but more on that soon. And in the settings, you
can see that we have way more settings now, which I absolutely love because now you can actually
pick your language. And in the advanced settings, the most important thing in here is the looping.
So you can pick how often it can prompt itself, the temperature, something like the creativity of
the model. How much out-of-the-box thinking do you want? Where one represents super creative and zero
represents super conservative. And then obviously, you get to pick the model. And I'll just
say, all of these work way better with GPT-4, and I signed up on the first day and got
lucky, so I have access to it. So if you don't have access to that API yet, follow along
with this video, and I'll be doing a livestream where we explore this together with my GPT-4 API.
Alright, so that's some nice additions to what we know already. But beyond that, there's
godmode.space That you might have heard of already, and this is a different version
of it where it asks for more of your input, which I kind of like. And at this point, we
should start talking about how these work and how to use them. So overall, you give them a
goal that you want to achieve. This is not about coming up with instructions and context like
with usual prompts. This is all about giving it the final destination to where you want to get.
So instead of saying, "Write me a sales email for my company," you would say, "Send out sales emails
for my company." Again, many limitations to this, and it's not going to work perfectly, but it's
fascinating, and this is where the space is going. So it's important that we start learning
about this as soon as possible, which is now. Okay, and then there's a third one that popped up
a few days ago, and this is a web interface for Baby AGI. And I find this one to be the best,
except it doesn't have access to the internet yet. So let's talk about that. It's obviously
a key feature in some of these, especially when you're trying to do research. It's really
good to have access to the internet. Right now, the deal here is if you were to set up Auto GPT
on your very own machine manually, which honestly might be too much to ask of most people that never
used code before, you could access the internet, and you would get all the custom options that
you get from these different apps. But honestly, even for myself, having that and having this,
I end up using this interface most of the time just because it's so convenient. And often, tasks
I want to explore come up spontaneously, and over the past few days, just pulling this up, switching
this to GPT-4, and inputting your objective down here is what I find myself doing the most.
But if you do need internet access, well, Agent GPT added it, but right now, this button is
disabled. On the roadmap, it says they'll have it back up and running permanently within the week.
So by the time you see the video, it should work again. God mode has internet access, but it really
depends on your input if the results are useful or not. So this one, I would just experiment with.
And Baby AGI, as of today, doesn't have it yet, but apparently, we'll be adding it soon.
Alright, so we're getting deeper and deeper on this topic. But before we go into some use cases
and show you what is actually possible, I want to show you two more apps because you should totally
take note of these. So one of them is iBabyAGI by Nate Chen on Twitter, and he took the Baby AGI
code and he put it into an iOS app. So if you go on here, you can totally download the brand
new application on the iOS store and have this thing with you on the phone at all times, which
is pretty amazing. And a lot of things we look at is built on top of the shoulders of Yohe here,
which created the original Baby AGI. And by him open-sourcing it, allowed other people to build on
top. And this is the last app I want to show you before we get into the use cases, and that is Baby
AGI for us. So you might have heard of this GPT-4 model that is completely free but quite limited.
Well, that love here on Twitter went ahead and hooked it up to Baby AGI, and this one only has
the basic functionality so far and requires basic coding knowledge for you to run this. But the
result would be that you have a simplified version of GPT-4 running on your local machine that can
prompt itself, where you deploy various agents, and it just works by itself. So really, that's
kind of the alpha version of the ultimate thing that you would want to have, right? An automatic
version of GPT-4 that can work by itself and complete tasks on your computer by itself. I
personally wouldn't recommend this yet; it's too early. I find this free version of GPT-4 called
GPT for All to be a little too basic for most use cases. That's why I didn't create a separate video
about it. And again, it's also early days for all these auto GPTs, agent GPTs, however you want to
call them. So they're not extremely useful yet, although there are some use cases.
So now that you have an overview, let's finally get into the practical part of
this video. Okay, so I would say, let's just dive right in with a prompt that we used on the
channel here before: "Develop a marketing plan for my AI YouTube channel that spreads awareness
of my AI tutorials beyond YouTube." So up here, I'll pick GPT-4 as it is the best. You can totally
experiment with 3.5, but let's not kid ourselves, GPT-4 is so much better inside of these
automated apps. And then, to begin with, we'll say three iterations, and that's what
I like to start with. And for the first task, developing a task list is usually the best.
So this is it. I hooked up my API. There's not much more to this. I'll just say "run," and it's
gonna start doing it. And the amazing thing about Baby AGI in the web interface, it actually works
and it's reliable. It doesn't get lost in loops, and the results are very useful compared
to some of these other web apps that are more experimental, and you get useful results
sometimes. But from Baby AGI, I was able to retrieve useful results regularly. So, as you'll
see, first, it's creating a task list. Here, we'll just have to wait a little bit until it's done.
Okay, so there you go. The first result is we have a task list. Here are the tasks: "To find
the target audience, research the competition, create a content strategy." And here, you
can really see the difference between just prompting ChatGPT and doing this. Because really,
getting the best results from ChatGPT and GPT-4 is all about being sophisticated about your
prompts and knowing what exactly to ask. Here, you just lay out a goal, and it finds the
different prompts that it's supposed to run to get to that goal. Is it gonna find the optimal path
and the optimal prompts? No, it's not perfect, and it doesn't have all the context it needs. But
is it going to get you better results than just running this inside of GPT-4? Absolutely, yes.
And also, let me say this: one of my favorite ways to use this is I just start with something
generic or a short prompt like this, and then it's going to lay out all the different parts.
And I'm just gonna take these one by one and craft my own prompts from these, right? Like,
I could take "defining the target audience" and turn this into a 10-line prompt that is
super specific to my organization, my needs, and I would also provide the information that I
already know about my customers.And then I could run this prompt sequence by myself inside of GPT-4
and get similar results. But this is really going to get you started, and it's just fascinating to
look at. And as you can see below, it's already at work. It's developing a channel branding strategy:
create a consistent visual identity and tone for your AI YouTube channel to increase recognition
and credibility. And it's just doing that all one by one. And now I gave it three tasks to complete,
right? But if I say 50, it's gonna go through and make this entire to-do list happen for you.
Okay, so that's the basic approach. You just take the prompts you have for ChatGPT and you move
away from providing instructions and some context to giving a final goal and some context. And just
that alone is gonna get you very far. But you can go further and actually use the internet search
capabilities of these things. So, for example, if I go to godmode.space and I paste in the
prompt that we shared in the Weekly Newsletter that went out yesterday, and it looks something
like this: "Use Google Trends (then the URL to Google Trends) to analyze the search volume and
interest for keywords related to (and who are going to replace) the product or service.
I'm going to say generative AI education, and we're going to say in the AI industry. And
then we round it up by saying identify emerging trends and potential niche opportunities."
Now, we did pick GPT-4, and we're just gonna launch this. And as you'll be able to see, God
mode has a little bit of a different approach. It's going to give us the tasks to pick from. So,
this is more interactive and less automated. Okay, so here are three suggestions. So, I then fire
11 keywords related to generative AI education in the AI industry. Sure, use Google Trends.
Yes. And then emerging trends and potential niche opportunities. Sure, let's launch this.
And now it's going to start working. It has these three tasks that we defined. And by
clicking this button, you can just lean back, have a drink, because this is gonna auto-approve
all the requests for the next 10 minutes. And a good thing about this interface is that you're
going to be able to give it some feedback while it performs. You can add context while it goes.
But as you can see, it proposes the first action, right? Let's use Google and input keywords
related to generative AI education in the AI industry. And it's going to Google that for
you. And now that we have auto-approve enabled, it's just gonna do all these things one by one.
And again, don't expect any miracles. These things are developing on a daily basis, and they came
out a few weeks ago. But this is the direction everything is going in. Okay, you're gonna
be setting goals, and ChatGPT is going to be prompting itself towards that goal. So, the skill
of writing clear prompts is not going to go away. But it's not going to be necessary for every step
of the way anymore because it's going to be able to come up with these tasks and execute on top
of them. But as you can see here, it's always going to be able to take additional input from
your side. Usually, giving additional context in some form is the way to go here. And again,
that brings us back to the basic prompt framework of instructions plus context. Because if the
instructions that it's doing are fine, you don't need to touch that. All you need to do is add
additional context if it's necessary. But if it's going in the wrong direction, Yes, go ahead and
change the instructions. So that would be my pro tip on how to get optimal results. Okay, and the
next step we're going to look at is Agent GPT. And what I want to do here is actually take one of my
newer prompts: "Create a budget." And as you can see, this is going to help us create a budget for
my $3,500 monthly income by allocating funds to rent, groceries, transportation, health insurance.
Then we assign categories to the various expenses we have. And then in the end, we say, while taking
into account saving for a wedding in three years. Okay, we're just gonna paste this into the goal.
And as you can see, we have the instruction here, right? "Help me create a budget." And then all
of this is context, and this is a classic prompt. And as I said before, you could also reformat this
to be focused on the goal. But in this particular case, where it's "create me a budget," and then
you have all the constraints, that is kind of the goal. You want the budget, that is your goal.
So just always post the question, "Does my prompt result in some goal?" If the answer is yes,
it's probably going to work super well in here. Okay, I'm gonna switch to GPT-4. We're gonna
have it on automatic mode, and I'm gonna set the loop to three iterations to start with. And
in this case, I'll take the temperature down to about 0.4. And you can experiment with this, but
generally speaking, with something like a budget, you don't want that much creativity, right?
It's just like if you're hiring an accountant, you want a super conservative and
orderly accountant. That's why I took the temperature down here. I'll just call this Budget
Assistant, and we're ready to deploy our agent. And it's going to start working on this. As you
can see, it got right to it. It added the task of calculating the average cost of a wedding right
away because we need to keep that in mind. And if you're paying attention, you would realize, "Oh
wait, maybe I could have provided the cost of the wedding that I wanted to calculate for, right?
Then it doesn't have to do the step, and it's going to be more precise." So if you wanted to do
that, you would stop the agent, add that to your prompt, and rerun it. But let's just go with this.
Next up, it researches and suggests a percentage-based budget allocation for essential,
discretionary, and saving investment categories. So it respects our categories. That's fantastic.
And only then, it creates a monthly budget plan. So now it's working on the first thing,
and it came up that in the United States, a wedding costs around $28,000. And
then it divides it into the three years, coming up with the number of $777 that you
need to save every single month. Fantastic! Alright, the next step, respecting that situation,
that context that it just came up with itself, it starts splitting up your budget into
different categories and gives you all the reasoning behind it. So look, this is fascinating
and actually really useful if you're creating a budget and don't have much experience with that.
A prompt like this is really gonna get you going. And to round everything out, I just want to show
you some of the coolest use cases that some of the people across the internet came up with. So for
example, this guy hooked up all the GPT-2 11 Labs, which allows you to produce realistic voices, And
he used it to order a pizza successfully. Why is it always pizza? I don't know. If you're familiar,
but the first real-world Bitcoin transaction was a guy buying two pizzas for 10,000 Bitcoin back
in 2010. So I guess we're at that moment with AI right now. Anyway, here's another one that
I'm experimenting with, and once I perfect it, I'll show it off on the channel. But you can
essentially use these agents to research your podcasts and your videos. So here, you had five
different searches in 15 different web browsers, and GPT just prepared all the topics for
the All-In podcast, which, by the way, I couldn't recommend enough. It's
my favorite podcast. And then here, Nicholas shows off how to use Agent GPT, that
we just looked at, for competitor analysis for specific products. So all these things,
you can explore. I'll include all the links below. There's so much fun to be had with this.
So now you have all the links, you have all the basic resources, you know how it works. And
even if you don't have GPT-4, don't let that hold you back. Take the 3.5, start playing with
it, and see what you get and how it differs from ChatGPT. Because this is the direction the space
is evolving in. And by you becoming proficient at prompt design and these primitive versions of
these Auto GPTs, once the better versions come around, you're going to be way ahead of everybody
else because you're going to understand how this stuff works at a fundamental level.
But if you don't want to do it yourself, here's the live stream where I'll be
exploring some of these with the GPT-4 API. And if that's not out yet, turn on
notifications because it will be soon.