Everyday AI Podcast – An AI and ChatGPT Podcast - EP 313: How To Use Custom GPTs Inside ChatGPT

Episode Date: July 12, 2024

Win a free year of ChatGPT or other prizes! Find out out.Wondering what on earth a GPT is and if you should use one? Yes! GPTs kick ChatGPT up a notch, letting you make custom workflows to fit your needs. Whether you're a newbie or you've dabbled with a few GPTs before, this episode is for you.Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan questions on GPTsRelated Episodes:Ep 183: Turning GPTs Into Gold – Monetization Strategies and Practical ApplicationsEp 217: 7 Steps on How To ACTUALLY Use ChatGPT in 2024Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTimestamps:03:00 Basics of GPTs08:52 Knowledge workers can benefit from GPTs.11:49 Creating custom GPTs allows fine-tuned control.15:40 Many publicly available GPTs for custom use.19:04 Customized ChatGPT utilizes industry-specific knowledge.22:01 Multiple GPTs used in chat for autonomy.25:45 Sell custom GPTs and make some money.27:47 Quick rapid-fire Q&A for live audience.32:08 Identify repetitive tasks to create GPT.35:01 Customize ChatGPT for specific tasks and needs.39:44 Provide input-output pairs to improve GPT performance.43:41 Use web reader GPT48:59 Using GPT helps complete tasks efficiently.Topics Covered in This Episode:1. Custom GPTs: An Overview2. Utilization of GPTs in Everyday Tasks3. Creating Custom GPTs4. Monetization of GPTs in GPT Store5. GPTs in Action: Automating Work OperationsKeywords:artificial intelligence, podcast, Bing, priming the AI, GPT, chat GPT, custom GPT, GPT store, workflow automation, repetitive tasks, knowledge tasks, ChatGPT Plus, GPT customization, business development, career growth, Slack's AI tools, Boston Dynamics' robot, Microsoft's AI model, GPT course, Copilot Pro, monetization, browser plugin, document recall, targeted research, configuration instructions, conversation starters, custom knowledge base, paying users.Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the Everyday Podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All In One Creative AI Studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. What the heck is a GPT?

Yeah, we've heard of chat GPT, but did you know that you can create your own? And that it only takes sometimes a minute or two to create a custom version of chat GPT that can honestly completely change the way that you work. So we're going to be showing you how and going over some of the basic questions. in today's episode, a beginner's guide to custom GPTs inside of chat GPT. What's going on, y'all? My name's Jordan Wilson, and I'm the host of Everyday AI.

And Everyday AI, it's your guide. It's your guide to learn what's going on in the world of artificial intelligence and how you can leverage all of that knowledge, all of these updates, to grow your company and to grow your career. So if that sounds like you, welcome. You are in the correct place. and we do this live stream and podcast every single morning, but also know that there is a free daily newsletter.

So if you haven't signed up for it, you have to go to your everyday AI.com because there's a lot of great information there. So more on that here in a minute. Let's get into a beginner's guide to custom GPs in chat GPT. So whether you haven't created a GPT yet, you're brand new to chat GPT, or if you've created five or 10 of them,

I still think that you're going to get a lot of value out of today's show. So I'm just putting that out there. This is a beginner's guide, right? So also, I want to hear from you all. I want to hear questions. And, you know, I referenced the newsletter earlier, right? And why you should be reading it.

Well, number one, there's exclusive insights that literally are only found in our newsletter. There's dozens of AI newsletters. They all cover the same thing. We bring you exclusive content every day, including I let you all pick the show. And by one vote, this show won out over perplexity. So was that fun for those of you that voted that you get to control, you know, my morning and our team's day?

Hopefully, hopefully that's fun for you that you guys are in the driver's seat. But I'm also curious, you know, to our live stream audience or, you know, if you're listening on the podcast, hit me up. How many GPTs have you built? And what's been your experience so far? And what are your biggest questions? So to our live stream audience that's, that's joined. joining us like Anili and Jay and Tara, Brian and Harvey joining us from the YouTube.

Thank you. Thank you all for joining us live as always. But let me know what are your biggest questions. And also, how many GPDs have you built? I'm always curious. I feel people have either built like one or two or like 50. And I think a lot of that lies in how useful people are or are not finding these GBTs. So I hope that what this, that's what this episode.

can help you with. So let's just dive straight in and talk a little bit about some of the most common questions about GPTs. All right. And at the end, as long as I don't accidentally go on some side rants, at the end, we will jump in live. But let's just go over the basics, y'all. This is for beginners. So some questions, some questions I will post to myself. But get your questions in as well. So some questions I will post to myself an answer. Why would I use a GPT in chat GPT? Well, to put it very shortly, a GPT is a customized version of chat GPT. So if you've ever found yourself using chat GPT and you're like, hey, this is okay. However, you know, it's struggling. It's, you know, I can't use my information about my company or my brand voice easily or, you know,

there's a certain part of chat gpti that I don't like. It's acting in a way that doesn't work for me. That's what gpties are. Think of them as a very easy way to go under the hood of chat gpti and change things in a very easy way. All right. So that's why you would want to use a custom gpt inside of chat gpt is to make it your own, is to maybe add or remove features from chat gpti to adding your own knowledge.

So what are custom gpts for? Well, custom GPTs are really for anything, right? Any, anything that you can do inside of chat GBT, you can create a custom GPT around that, right? And I think that also goes to a common misconception that people have about chat GPT. I think a lot of people think, oh, it's, you know, it's for writing blog posts or it's to help me brainstorm ideas. Yes, it is, right? But that's like saying, okay, this Ferrari is an umbrella. It keeps me dry when it's raining.

Yeah, sure. But that's like not even scratching the surface, right? So what are custom GPDs for? Think of this. So I talk a lot about knowledge work. And so many of us, myself included, but so many of us and probably so many of you out there listening are knowledge workers. So what that means is a company pays you for a specific skill.

or a type of knowledge or expertise, and you generally sit in front of a computer and perform that knowledge or that expertise somewhat repetitively throughout a course of multiple knowledge-based tasks, right? So that's what a knowledge worker is. And that is where GPTs can really help. And we're going to go through some examples of that. But that's what GPTs can really help for. Think of all those manual tasks that you do repetitively, that require your knowledge.

And then think, what if you inputted that knowledge into a customized version of chat GPT that you can use at any time, then you can kind of train it and go back and forth and, you know, kind of unload the knowledge that's in your brain that your company's paying you for or maybe, you know, you're working on a side project or you're an entrepreneur, right? Take that domain expertise that you have as a knowledge worker, give it to chat GPT, right, which I know that sounds weird. give away all your expertise, right? But that's where the world's heading. And you can create multiple specialized versions of your own knowledge, right? Or of your company's public data, right? We're going to get into that, but don't upload confidential proprietary, sensitive information, PII, PHA, don't upload that into chat GPD.

But all your public data, all your knowledge, your expertise, your own internal decision-making, you can very easily create a custom GPT that can then perform those manual tasks. that we all do day to day for you. Anywhere inside chat GPD. We're going over the basics. We're going over the basics, y'all. Cecilia said, I need this course to help build my confidence to build my first GPT. Well, hey, get your questions in, y'all.

Get your questions in. I'm going through the basics here. And then we're going to build one live. All right. So let's go over some more of the basic questions about GPTs. So can I create my own chat GBT? Well, kind of, right? So you are creating a version of chat GPT, right?

So you are essentially starting with the base model when you are creating a GBT. And you are giving it kind of custom instructions, right? So a lot of people were confused when I said don't use custom instructions, right? So if you're new to chat GBT, that was a feature that chat GPD released, I don't know, about seven or eight months ago called custom instructions. And you could essentially give chat GBT custom instructions. However, I told people don't use that. It's going to be bad, right?

Because if you're using chat chbtee for a variety of purposes, which I would encourage you to do, right? So for writing, researching, analysis, to help you plan your day to connect to third-party services, right? So we've always had these things called custom instructions. However, those applied unilaterally to every single chat, right, which was bad, right? It was a marketing gimmick, if nothing else, right? So that's where creating your own kind of quote unquote version of chat GPT through GPTs really comes in handy. Because essentially you get this concept that we had many months ago of custom instructions that I think was, you know, marketing gimmick and not very useful for very many people. But now it's it's fully fledged, right?

So now you have more kind of fine-tuned control over, you know, you can build an army, right? If you have 10 specific skill sets that you do on a daily basis, you can build all of those skill sets into little GPs, and then you can call them or mention them at any point, which we're going to show you live. So yes, you can create kind of your own version of chat GPT, but it's essentially you're either adding some features, adding some knowledge,

or maybe taking some features away from chat GPD, which we'll get into. Are custom GPTs free? Well, the answer to that, that is yes and no. You do need a paid version of chat GPD plus, which starts at $20 a month. You could also opt for chat GPT teams, which I would encourage everyone to do. It is $30 a month and requires a minimum of two users and a maximum of 149 users. However, when you do that, it is you have a little more control over sharing your GPDs,

which we'll get to it in a second. But however, and you do get to it. double the usage, right? Double the message cap, which is something that we, you know, talk about. We go over in our free prime prompt polish PPP course. So they are not technically free because you do need a paid chat chbt

account. However, once you have that paid chat GBT plus account, they are at that point technically free to make, right? So it's not like you have to pay for each one that you create or you do not have to pay for each time you use it. Once you have a paid account,

You can, I don't even know if there's a limit. You know, you can create hundreds of GPTs and use them for all different purposes. So, yes, they are technically free once you have a paid chat GPD account. Can they only be used in chat GPT? Yes. Technically, there are kind of ways around it if we want to get a little more technical by instead working in with the assistance API within Open AIs playground. But by default, if we're talking beginners, if we're talking about the kind of quote-unquote

simple GPTs that you can create inside of chat GPT, they can only be used inside of chat GPT. They cannot be used elsewhere. They cannot be embedded on your website. And yes, you can share them. But anyone that is using them has to also have a chat GPT account and a paid version at that. You can't just have a paid account, create a GPT, and then just share it with people who have a free account. So to use a GPT, whether you create it or not, you do need a paid chat GPT account.

All right. And hey, as a reminder, we're doing this live. We're doing this live for you all. So I do see a couple questions in here. But make sure if you do have questions, even as I go along, if you have clarifying questions, or if you are listening on the podcast, you know, we always check out the show notes, always put my email, my LinkedIn. So feel free.

ask any more questions. I'm going to do my best to answer them. All right. So next common question. How do you use your custom GPTs? It's a great question. So there is a default GPT mode, which I'm going to show you here in a minute

when we kind of dive into this live. But at any point, you can mention a recent GPT or a GPT that you can use, right? It's also important to denote there right now are millions. I think the last count I heard is, something like four million. There are millions of publicly available GPTs that you can use that are already kind of built, right, that other people have gone through this process and they've built their own and they make them publicly available. So the way that you use your own custom GPTs,

and I'll show you this, your most recent ones are going to be on the sidebar on the left hand side of your chat GPT Plus account or at any point in any chat that's using GPT4, Adobe just introduced. an entirely new way to create, bringing the power and precision of its creative suite into one conversational experience. Meet Firefly AI Assistant, now live in the Adobe Firefly app, the All In One Creative AI Studio. Powered by Adobe's Creative Agent, Firefly AI Assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the Assistant. The Assistant orchestrates multi-step workflows, drawing on 60-plus Prograde tools across Adobe Creative Cloud apps, including Photoshop, Illustrator Premiere, Lightroom Express,

and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director. Adobe Firefly AI assistant now in public beta. See it today at firefly.adobie.com. You can at mention and use a GPT that way.

So you can use the dedicated GPT kind of interface, which I would not recommend. And if you've taken our PPP course, our updated version two, PPP course that we just updated this month. You know why. But you can use it in the dedicated GPT mode, which I would not recommend, or you can use it in the default mode or anywhere just by at mentioning

that GPT. So you're going to give it a name, and then you can call on it at any point. All right. Next common question for our beginner's guide to custom GPs in chat GPT. Do I need to know how to code to create a GPT? Absolutely not. We're going to go over that if you can talk to chat GPT, you can build a GPD. It's actually kind of meta, right, because you can actually use the,

gpt builder to build a version of chat gpti inside of chat gpti that then you'll later use inside of chat gbt and you can improve it with chat gpd it's it's this weird loop of of you know gpties but short answer is you do not need to learn how to code uh to to learn to use a chat gpt to build a custom gpt uh can you add extra features and in functionalities if you are a coder or a developer absolutely uh but you can create a basic yet extremely useful GPT in minutes without even learning to code. You can do it conversationally, which we will show you here in a second. All right. Next common question.

How are custom GPTs different than chat GBT, which is a great question. Some people are confused. Technically, you can make it not different at all, right? You can go in and essentially give it no instructions and just say, all right, here's my version of chat GPD, right? You can make it only slightly different. You can make it extremely different, right? The options are endless in terms of how they can, how custom GPTs can or can't be different than chat GPD

because we're going to go into the back end and just go over some of the simple options. But yes, you can make it extremely different. You can limit chat GPT's capabilities. You can limit kind of the features and functionalities that it has depending on the results that you want to drive from it. right? So also in the configuration instructions or by tapping into third parties, you can make it extremely different, right? In theory, you know, maybe the leveraging the GPD technology is only 10% of the power of the GPT that you're going to create in 90% of the power maybe is in the configuration constructions and third party services that you can connect

via actions, which we're going to get into in a minute. Or maybe it's 95%. Just a chat GPT, the default clone and you just have a couple things tweaked to give you better outputs. So we're, you know, maybe all you have is you're just uploading, you know, a couple documents to, you know, kind of do what I like to call, which is not technically accurate, but I like to think of it this way, you know, doing some like mini rag, right? So some mini retrieval augmented generation, right? So just like an easier way to bring in kind of this domain specific or industry specific knowledge that you want chat, GPT to refer to, right?

Because there are downsides, right? There's downsides to using extremely large language models that are trillions of parameters because sometimes they know too much, right? So that's another great benefit in a way that custom GPs are a little bit different than the default version of chat GPD is because you can upload your own documents, your own knowledge base, and you can kind of tell or encourage chat TVT. Sometimes it'll work, sometimes it won't, to first refer to your documents versus first referring to its knowledge base, right?

Which is essentially what RAG is, right? That's the, you know, kind of all the rage over the last year with large Hague which models has been around fine tuning, has been around RAG, you know, bearing in your company data. But this is a very overly simplistic way of, hey, you can bring in your company data pretty quickly with custom GPDs. And that's a way that it's definitely different. So how are custom GPDs different?

then plugins. Well, plugins are gone RIP. All plugins have been depreciated as of April 10th, so as of about a week ago, super sad. And I'm, I'm super bummed about it. I'm going to go on a short, can I go on a super short rant here. I'm looking at the time. Live stream, you're so with me, right? I can go on a short little rant here. I truly believe Open AI made a big mistake when they got rid of plugins. There's reasons I don't understand, obviously, you know, they wanted to kind move everyone into GPs versus plugins, they cited a variety of reasons. One reason that GPs are extremely different is you can only use one at a time, right? So technically over the course of one conversation while keeping the results of everything

in chat GPT's context window, you can work with an unlimited amount of GPTs, right? So that was kind of the counter argument when OpenAI faced a little bit of backlash by getting you're in a plugins. The difference is, right? You have to have more manual human input when you switch between GPTs, right? So I'm going to give you examples, but you could, as an example, ask for something from GPTA, right? It spits it out. Then you then have to take that information. And then you can manually then, hey, based on this information from GPTA, use GPDB, and do this task, right? So yes, you can work with multiple GPDs in the same chat, but you used to be able to do it autonomously, right, which is something that I only think maybe our listeners and a couple

other people on the internet were actually using it for. So I think with chat GPT plugins, we had a very small, very, you know, I'll say archaic or a very duct tape version of agents, right? You could choose any three plugins in one prompt. You could have all those plugins work together and with each other autonomously. I just, we did it live here on the show like a year. ago, right? And I don't think people understood the power of plugins because those three plugins could work back and forth autonomously, right? And they could, you know, hey, plugin A could

go retrieve, send it to plugin B, send it to plugin B, send it to plugin B could complete a task, could send it back to plugin A, plugin A could go look something up, send it to plugin C, right, and do that all autonomously with one prompt without any additional human intervention. Now, however, with GPTs, you can only mention one of them at a time. So I personally, miss plugins, right? But that's one way that they're different. Yes, it is technically easier. You know, the average person couldn't create a plugin, right?

You might have had to pay a developer. You might have already had to kind of know your way around development to create a plugin. So it does make it easier to create a GPT. However, it does greatly, I think, reduce or limit the output of what you could get with chat GPT by using GPTs instead of plugins. Next common question. And hey, thanks. You know, Hector, I see a couple questions.

TC joining us from YouTube. I'll get to your questions here in a minute. So if you do have questions, I'm going to get through mine and take all yours at the end. So make sure to get them in. Can I publish my GPT in the GPT store? Yes, of course you can. However, no, that if you are doing it, if you are, like I talked about, there's, I think, more than 4 million GPs in the GPT store.

You can search them. you know, anything that's in the GPD store is open for use. And it's free for anyone to use. However, if you don't know what you're doing, it's actually pretty easy for someone to copy or clone your GPD. If you do publicize it in the GPD store, like I said, if you don't know what you're doing. And there are obviously some security things to keep in mind.

That's another episode for another day. But anyone can publish their GPT in the GPD store. But I caution you, you should really know what you're doing if you make your GPD public. but yes, it is fairly simple. Also, can you sell custom GPs? Oh, yeah. So you can't sell them in the GPT store.

That's the thing, because once you publish them, they are public and they are searchable, right? They're indexed. So anyone can then use them for free. You can monetize off the GPD store right now. That is for the US only. It is based on usage. And according to early reports, it's not a lot of money.

It seems like the most popular GBT's in the world among millions have gotten a couple thousand dollars so far. So, you know, if you think of this as kind of like the App Store, right? And there's literally, you know, companies that went from zero to 10 billion plus dollars in revenue just by creating apps in the iOS App Store. And they did that over the course of a year or two, right? It's not currently a feasible, at least right now with the monetization structure. within the chat GPT, chat GPT's GBT store.

I sure do hope that OpenAI increases the monetization amount and looks at it more like a display ad, right? And you get paid, you know, I don't know, a fraction of a penny every time someone queries, you know, your GPT or something like that. I don't know what a fair model is. But I would say if the most popular GPs for a company like OpenAI that is, I think, valued at 90 plus billion, they're doing a hundred. hundred plus million dollars in ARR annual recurring revenue.

I would hope that the most popular apps or sorry, the most popular GPTs get more than a couple thousand dollars, right? These are GPDs that are being used by presumably hundreds of thousands of people. So can you make, can you sell custom GPs? Well, you can, if you're lucky and if you, if your GPT gets super viral and tens of thousands or hundreds of thousands of people use it, you might get a couple bucks, but not a lot. You can sell them externally, right? So there's options and I'll show you that.

And eventually, yes, we are going to have a dedicated course. So those of you, yes, y'all, I talked about this months ago. I didn't forget. I promise, you know, all of a sudden we had, you know, big companies and Vida saying, hey, Jordan, bring every day A out to California and, you know, a couple of Fortune 500 companies saying, hey, Jordan, can you come teach us this and that, right? So I promise you we're going to have a dedicated, in-depth, multiple step, you know, GPT course inside our community.

So I didn't forget about that. Don't worry. But you can also sell custom GPTs in a different way, right? So it's not really a beginners, but you can essentially make it private and then use a third-party service to charge for your GPT. All right. Then how can I share my GPD with my team? Well, I'm going to show you that, but essentially there's different options when sharing your

GPT. you can make it public, you can make it just for you, or you can make it unlisted. And essentially, if it's unlisted, that is the way that you can share it with anyone on your team. There are some better controls if you do have a team account, just so you know, it's a little easier to share or revoke access to those GPTs with team members. All right, let's go rapid fire on your questions. And then don't worry if you're still, if you're still, hey, is anyone still there? I always wonder, right?

sometimes if the episodes go a little long, are you still there? Should we dive in live here? I think we can make it quick, but I think it'll be helpful. And if you are listening on the podcast, if, hey, if our live stream audience wants to keep going and explore things live, I'm going to do my best to describe it. But this might be one of those that's most helpful to obviously check out the show notes and watch the video. All right. So let's go over some of your questions, rapid fire here.

All right. So TC, I'm wondering if custom GPs could streamline the price. prime step in the PPP. So without getting into details, yes, kind of, it can. There are some, I'm going to say there's some ambiguity with how the document retrieval process impacts tokenization. In our testing, it's different every single time. So it's hard to keep memory in mind when using GPTs. So without getting into the weeds, kind of. Hector, where can I find a guide to build my own custom GPD? There's tons of. of them. We did a video on that Hector. I'll make sure to include it in the newsletter.

Another question, seeking custom GPD for job seekers, where can I find them? So yeah, you can just go into the GPD store, search anything, search job seekers, search career builder, right? Like whatever, and there's probably dozens or hundreds of different GPDs. You can quickly try them all for free if you do have a chat GPD plus account. How can I eliminate another question? How can I eliminate hallucinations when building a custom GPD. Well, you need to increase the quality of your data. You need to know how your instructions impact how chat GPT acts. And we'll go over that here in a second as well. Another question, Hector, love all the questions. Where can I find a searchable database of custom GPDs available for

specific tasks or use case? Well, the GPT store is not the best. We do also have our own custom directory on our website. So just reach out to me if you want access to that. And then, Then Hector, another question. How can I publish a custom GBT as a Chrome extension? So you can't. There's not the way that you would build it inside chat GBT with the GBT builder. You could build it in other ways, but that's a different question for, or a different show for another day.

All right. Frank says, how can we get the latest free and previously paid PPP? Well, hey, our prime prompt polish course is always free. We do paid engagements like this for companies. knees, right? But our free one, we mention it in the newsletter every once in a while. So, Frank, if you've ever registered for our course before, it's the same link. You can go up there. All right. Looks like one other question from Jay, how do you know which custom GBT in the store would be best to start without trying them all? Luckily, a couple of weeks ago, and if you follow

our show, we mentioned this in the news. ChatGBT just released some more statistics on GBT's. before there wasn't a lot of that information. But now you can see the average users, the average score. So normally, just like with searching an app, right, you would want the one with the most reviews, the highest score. Same thing, right? Not saying that, you know, new GPTs with no reviews and no usage can't be good. But if you want to know a good place to start, find something with, you know, 4.5 stars and,

you know, 10,000, you know, uses, I think is a good place. Another great question from Doug, Doug, saying, are custom GPDs created in Copilot Pro available to non-paying users? So this, what today's episode is about. And Doug, thank you for the question. We are not talking of co-pilot pro. There is kind of a way to build a version of custom GPDs in co-pilot pro. That is not what today's episode about.

We're just focusing on them in chat, GPT. All right. Woo, that was a lot. Let's do this quick, y'all. Let's build this live, shall we? we're not going to build one live from scratch, but I want to show everyone a quick example, use case, and maybe something that I would use a GPT4. So go back and I said, what is it that you do repeatedly as a knowledge worker, right?

You have some level of expertise. You have some level of knowledge, some level of domain-specific experience in a certain sector that someone pays you for, right? So think to categorize that knowledge and think, okay, what are these repetitive tasks that I do that involve either, you know, writing, research, analysis. What are these, you know, repetitive tasks that I do over and over that I can make a GPD? So just as an example, something I do a lot right here as the host of everyday AI is I plan a lot of podcasts, right? We do this every single weekday, Monday through Friday, right? So as an example, am I an expert at planning a podcast?

Nah, I mean, I'm okay, right? I've done it 250 times. But I use chat GPT to help me, right? So let's look a little behind the scenes here, if you all don't mind. And let's just give an example here. All right. So let's dive into this here. So I have an example here of a podcast planner.

Okay. So I am in the GPT mode here. And you can always access. your GPTs, like I said, on the left-hand side of your toolbar. So there's some of my GPTs, and this is the one that we're going to be doing an example of here. This is called the podcast outline planner. All right, we're not going to build this from scratch, but I'm going to show you

kind of how it's built and kind of show you around the interface, right? So if you are brand new to GPTs, it's very simple. So there's essentially a create tab and a configure tab. You can do work in either, right? So essentially, the Create tab is more of a, it's like using ChatGPT to build Chat GPD. So you can just have a natural language conversation with Chat GPT to build a custom GPT. Right. So you can say, hey, I want to build a GPT to help me plan podcasts, right?

And then it's going to ask you some questions. It's going to go back and forth, right? And then you can answer those questions. And essentially what happens is chat. GPT or the GPT builder is going to take away what it believes are the most important instructions, essentially from that natural conversation that you're having with it. So like I said, you don't have to know how to code. You have to know how to have a conversation with a bot, which is extremely easy because

the GPT builder kind of guides you through it, right? And maybe you played around with GPs when they first came out in November and, you know, early December and you haven't been back since because it was a frustrating experience, right? The experience has improved a lot, right? I remember it used to sometimes take 15, 16, 17, especially in the first week or two, generations, and the GPT builder would just time out. It's not like that anymore.

It's pretty reliable, right? So in natural language, you can essentially create custom instructions and tweak and make a version of chat GPT for a specific purpose. So let's just say, as an example, you're a CEO, and you wear a lot of hats, right? and you're managing or you're helping manage, you know, your marketing department, your finance, your sales, your HR, right?

In theory, in theory, right, you could create a custom GPT and upload custom knowledge in each one of these things. So like I said, after having a conversation with the create kind of tab, it creates configuration instructions. Okay. So if you're on the live stream now, I have the configuration instructions for this GPT that I use that helps me plan podcasts, right? So I'm not going to go through these like step by step, but it essentially is a series of steps. And there's some formatting that helps chat GBT kind of better understand this.

But essentially, I'll kind of read the steps, one through nine, right? I'm not going to go into it. But it says prioritize guidelines is step one, and there's a description. Step two is received topic. Step three is document recall. And I'm going to show you the document that I uploaded. Step four is initial research. Step five is in-depth analysis.

Step six is engaged in conversation. Step seven is further research. Step eight is prepare outline draft. And step nine is present final outline, right? So I've tweaked this GPT a little bit by using both the builder and the custom configuration. You can also, especially if you've used chat chitp t a lot, you don't have to go through the create process because it can be time consuming and especially if you know what you're talking

about. What I like to do is I write my own configuration instructions from scratch because I've done this a lot and I'm pretty okay at it. So then I'll do that and then I'll use it. So essentially on the left side, you have your two different tabs. So you have your create and your configure. And then on the right side, you have a preview. Right.

So I can actually, in real time, as I update something, I can test it out to see if it's working as I wanted to. And then if it's not what people don't even, I think one of the most underutilized features of the GPT builder is using it in this way, is essentially, especially if you've built one or two. So I know the beginner, you know, this beginner's guide, maybe a lot of our audience, you've built a lot. So this is a tip that I think a lot of you can benefit from, especially if you built a lot because you know how these work and you'll get better. So here's a little tip, right? I like to write the configure instructions from scratch. So I go straight to configuration.

I don't use the GPT builder, right? So you have your instructions, which you can type. You have your name. You need to give it a name, a description, and then a photo. You can upload your own photo or there's a Dolly option that you can just click on. and it'll generate a photo using Dahl Lee. You have conversation starters, and those are, you know, I'm just going to say, let's plan a podcast, even though that's not going to do anything.

But now when I update this, that's going to be a conversation starter the next time that I use this. I believe my view is a little too small there. Okay. Also, then you have your knowledge base, right? So this is where you upload documents, right? So when I say this is mini-rag, it's not really, but it's the concept of this, you know, retrieval augmented generation where you can tap into the power of chat GPT. And this is where I think GPD has become extremely powerful.

Again, don't load private, sensitive, confidential, PII, PHI, don't upload that type of documents unless, you know, someone at your organization has told you to, otherwise don't do it. However, this is where you can upload custom documents. your knowledge, your brand voice, you know, examples, right? So in this case, this PDF that I have here is examples of a podcast outline. So that's essentially what this custom GPD is. So I give it all these instructions, right? And I'm going to bring up these instructions because I want to point a couple of things out.

And maybe this will understand, oh, here's why it could be useful. And now I'm going to show you how it's used. So this is really why it's useful. So I'm having it in step three, doing it. a document recall. So it's looking at the document and in the document, I give explicit examples of exactly how a podcast outline should be formatted, what it should include. So I give it, you know, we talk about shots, quote unquote, shots and prompting and how zero shot prompting is this terrible and you're going to get bad results. So in theory, I'm giving it examples.

I'm giving it input output pairs, uh, which are technically called shots, which should improve outputs. But I'm doing that, um, in a document, right? Because sometimes when these configuration instructions get super long, the performance can go down. So in this, in my series of steps, I'm telling it at step three to always, anytime that there's a new podcast topic that I type in, to always go in and do a document recall to look at that document, to understand that this GPT needs to be outputting a podcast outline exactly how I need it. Also, a couple other things here.

Right. And if you've taken our free PPP, you can see where some of this is coming in handy, right? I'm telling it to do initial research, and I'm telling it to use the browse with Bing tool to source recent information from the year 2024. So I'm doing what we call a targeted browse with Bing call. I don't just say use the internet. I say use Browse with Bing and find recent information from the year 2024, okay? And then I'm saying for it to visit at least 10 pages and to gather comprehensive details

about the podcast topic that I suggest. And then you'll notice, hey, and for anyone that's, taken our PPP course, I'm not asking it right away to give me a podcast outline. I'm saying first go do research, right? So think, rag, mini rag, right? And this is priming, right? This is essentially what we call priming. This is a very small version, right?

I still personally prefer to do this via a chat. But then I'm telling it, first understand the topic. Go read the documents, go research, then ask me specific questions. based on your research that will help you best fill out this document. Do you see how much time we're already saving? That's already, I'm telling it to go to 10 different sources, read the outline. I'm already saving probably an hour or two, just, and we're not even done yet, right? Then it's supposed to ask me questions based on what it finds.

I, the user, you're right? If you're reading this, this is what I'm telling it to do in each step. Then I answer the questions. Then based on my questions, it goes out and it uses Browse with Bing again. here in step seven, based on my answers, right? So again, this is the concept of priming, right? And you don't just want a cold prompt because you're going to get bad results. So first, it researches everything, comes back, ask me questions based on my answers. It's going to go do another round of research, at least 10 pages. Does it actually use 10 pages? I don't know.

I've researched. I say 3, 5, 10. Do you get better results? Sometimes, right? It's arbitrary. there's no way to know if it's actually using 10. But I do find at least when I give it a higher number, it usually comes back with better results than if I just don't give it any guidelines. And then after all that, it's going to prepare a draft. All right. So let's do this. I'm going to do this real quick, y'all.

So I'm going to go into the default mode. All right. So I'm going to start a new chat. And hopefully this makes sense. So I'm going to start. So we started. We started by talking about. how you should be using this, not in the dedicated GPT mode.

All right, so I'm in, I'm now going to call a web reader plugin. My chat, GBT, is being a little finicky here. Let's try it again here. All right. So now I am mentioning the web reader plugin. Okay. And what I'm telling it to do is I'm telling it to visit a specific URL. Okay.

Oh, I didn't do my full. Let's redo this one. I just realized I didn't copy my whole prompt. It's nothing crazy here. Let's try this again. All right. So what I'm doing is I am using a web reader plug-in. The reason that I'm doing this is because Browse with Bing cannot visit specific URLs.

Other GPs, again, this is one, a very popular one that's in the GBT store, can visit a specific URL. So I'm giving it a URL that takes it to recent AI news from the past 24 hours only. So I'm saying, I'm using the web reader GPT. I'm saying go to this page and give me 20 podcast, potential podcast episode titles based on recent news. Okay. So now it's visiting all these, this recent news from this Google news results page, which I think gives you 50 results. And it's suggesting different podcast episodes, right?

So I'll do this a lot. You know, yes, I can do this manually. And it still gives me, right, people are like, oh, Jordan, is everybody? everything AI know. This is just a way for me to more quickly research. Right. So as an example now, instead of me reading through 50, you know, news articles, I have, and I have links that I can always go back and read these. So I have links to recent news articles and ideas, right? So let's just say number four here, generative AI in the future of content creation. Okay, so I copied that. Now I'm in the same chat.

Okay. So I'm in the default mode of chat. So now I'm going to mention our new podcast outline planner. So this is called an at mention just like any social media. I type the at key and my recent GPs are going to be there. And there's the one that I just built, right? Or the one that we just kind of went over, you know, by default. So the way I built it, right, is I just give it a topic.

Right. So now what this. GPT should do is it's going to right now, it's going back and it is going through all those steps. So this might take a while. We'll see if chat GPT is being a little finicky today. I did just do this right before the show started, I swear, but this is sometimes how it goes. So now it should be going through all of those steps. So the first thing, obviously, it didn't do it, even though I did it earlier.

So now instead it just spit out the entire description, which is funny, right? Because I literally just did this. But hey, that's another thing to know about generative AI. It's generative. It's deterministic, right? You're not always going to get ideal results, right? So now as an example, I'm just going to go into, I'm going to try the same thing in the dedicated GPT mode.

I actually have a couple podcast planners. Hopefully I didn't just call on the wrong one. So here we go. So now when, for whatever reason, maybe I should have done a hard refresh there. But now when I'm using the dedicated GPT, it's giving me, it's going through correctly. It's asking me questions, right? So I'm going to try to quickly answer this quickly. So one, it's asking me a question, who is your intended audience?

So I'm just going to say, U.S. professionals, 35 to 55. We're going to go quick here. I don't like typing live. Are there, so it's, this GPT is doing his job, right? So this is what I would normally be doing in my head, but it's doing it for me. And it's making sure the result is going to be customized. So that number two, are there specific aspects of generative AI? So I'm going to say, I'm going to say creative aspects in general.

I'm not giving it good feedback because it's going to take a while to type live and I'm typing like a T-Rex with short arms. Number three, do you have any guest speakers in mind? I'm going to say, no, I don't have guest speakers, but you can suggest guests. for this podcast based on your research, right? So then question four, what type of format do you prefer for this podcast? I'm going to say fun, fun and conversational for the live podcast, right? And then number five, it says, is there a particular tone that you're aiming for? So I'm going to say casual.

All right. So now I answered all these questions and now based on the configuration instructions that I gave this GPT. What it should be doing now is it's taking all this information. It should be going back to Browse with Bing, find sources from 2024. And now the next generation that I should get is my podcast outline. And it should follow the exact template that I put in our knowledge base. So here it is.

And it's doing it right here. So it is giving me the exact podcast outline that I wanted based on the document that I uploaded in the knowledge base. And you'll see here it's just kind of, it's going to keep spitting it out. It did all my research for me, right? So people always ask, Jordan, how is it possible to do a podcast every day?

It seems like you do a lot of research. Yes, I do. Do I use AI tools like custom GPs, like perplexity, like Google SGE, absolutely all the time. Do I use, you know, do I visit web page, pages manually and read them like a human. Yes. Do I visit web pages manually and have saved prompts via Chrome extensions? Yes. But hey, here's one way that here's a task that I do repetitively, literally every single day. And if I didn't use AI, this would probably take me

three to four hours longer than if I do use AI. So this is a very simple way that you can use a GPT based around a knowledge task, right? And And the more that I use this GPT, I'll go in and add more configuration instructions. And kind of like one of the quote unquote hacks I was talking about is going in and talking in the create tab and saying, hey, this GPT is working pretty well. However, this step is producing bad results. Here's what's happening. Here's what should that should be happening. Can you please update all of the configuration instructions accordingly?

Don't take anything away, but help me fix this one problem. right so then on the left hand side you can have a conversation with chat gpte about your version of chat gpte to make it better right uh so who that was that was kind of a long one y'all that was a long one so uh i hope this was a good beginners and maybe we got a little bit into intermediates but a good guide on custom gpts in chat gpti again to quickly wrap it up gpc are a version of chat GPT that you can configure on your own. It doesn't require any coding experience. All it requires is you need to have, at minimum, a paid account, $20 a month to chat GPT plus.

At that point, you can go in the GPT store and use millions of GPTs that have already been built. Or you can create your own. Like I said, to create your own, you can create a decent version in a minute or two. If you spend more time, obviously it can be much better. There's more advanced customization options that we didn't really get to, such as using third-party APIs, different actions. You can use JSON, YAML, you know, yet another markup language. So there's more advanced things that you can do within these custom GPs, tapping into

third-party services via their API, via JSON, et cetera. But for basics, anyone out there, you don't need any technical experience. Think of the knowledge work, what you get paid for, right? Think, what does not require private data, right? What in my head? What is that expertise, domain-specific knowledge, those repetitive tasks I do over and over? Can I create a GPT that can help me cut that down, cut that time down by 30%, 50%, 80%, 90%, 95%. And the answer to so many of those questions is, yes, if you do it correctly.

So a GPT is a customized version. You can limit kind of chat GPT's functionality by default. You can add more functions to it. You can add your own data, your own knowledge base, your own documents, and your own configuration instructions. GPTs, I still believe, to this day, are one of the most useful, one of the most useful generative AI tools out there and one of the most underused tools. I think people have this thought that they're difficult or that they are extremely time-consuming.

And don't get me wrong. To create a very polished one that is consistently giving you accurate results, it is difficult, right? Especially when you're talking about tapping into third-party services, right? Maybe your company has an API. Maybe you want to use Zapier, right? But I think there's like 20,000 different actions within Zapier, right? So, yes, they can be very very. very complex and great ones can take a lot of time. But you can create simple time-saving GPTs to grow your

business, to grow your career in a matter of minutes or hours. And those are going to pay off exponentially because the more you use them and the more you go in there, you know, tweak the settings, add more knowledge, tweak your configuration instructions, the more time you were going to win back. All right. I hope that was helpful, y'all. If it was, please, Hey, we spend a lot of time on this thing. I sometimes don't sleep a lot. If you could repost this, we greatly appreciate that. If you could, leave us a review.

If you're listening on the podcast, on Spotify, on Apple, we'd appreciate a review. And like I said, if you are listening on the podcast, this is one of those that might help to come back and watch. I try to do my best to narrate. But some of these things are very visual. So if this was helpful, please consider telling others. Tag someone if you're on LinkedIn or Twitter, acts, whatever, it's called, repost this to your friends, family, co-workers. And more importantly, go to your everyday AI.com.

We did cover a lot in today's beginner beginner's guide to custom GPTs in chat GPD. We went over a lot, went on a couple side tangents. So if you want the too long, didn't read version, make sure to go to your everyday AI.com. Sign up for that free daily newsletter. And hey, maybe one of this thing where we ask you. all, you know, what we should be covering. Maybe we will. So, hey, this thing is for you. This everyday AI thing is for you. So make sure to get in on that daily newsletter. Make sure to join us tomorrow and every day for more everyday AI. Thanks y'all. Meet Firefly AI assistant. Now live in

Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us.

If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your Everyday AI. and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.