Improve Your GPT Building Skills - Challenge #22

It. It's sa. Sam. It. Oh, hello. Okay, okay, okay. This is it. Hello everybody. Welcome to another edition of the other live stream. There's been a waiting screen, so we don't need to waste any time. We can just dive right in and I spilled some water. That's okay. That's okay. Okay, never mind. How is everybody doing today on this blessed Tuesday, June 4th? How's everybody? It's good to have you guys live here. I feel like there's always such a disconnect on YouTube, you know, between like viewers and creators and, you know, that's just the beauty of a live stream is the audio coming in. Well. Can you hear me? Well, I took all the feedback from the last few live streams. I suppose the audio should be better. We're go. Sorry for that sound. We're going to be building GPTs today. So that's the plan for today. I'm going to go, you know, take the next step into showing you how to build custom GPTs. We're going to do about 10, 15 minutes of like lecture and hands on presentation. Then we can maybe build one that you guys suggest and then I'm going to challenge you yet again. There's going to be a new challenge and this is going to be all about GPT building. So, yeah. With that being said, how's everybody doing in chat? We're not going to be doing the thing as last time. I know last time I went a bit, you know, into all kind of the fun and entertaining things with like the video wall. This is going to be a bit more instructional, but nevertheless it can still be fun. If you want, you can share what country you're from. I know last time I like shamelessly abused that and just took all the countries and generated a map through the code interpreter. We're not going to be doing that this time, but nevertheless, you can still share where you're from. I always love seeing that in chats. I'm in Lisbon, Portugal. Beautiful weather here. A little too beautiful to be honest. It's very been. It's been so hot this week. Anyway, I see a lot of familiar faces in here. So let me, let me, let me settle in and then we can get into the content here. And for anybody, you know, for anybody watching the recording here and being like, oh, Igor gets, get to the content. I came here for the content. You know what, man, you can gladly skip ahead and keep those comments to yourself because we're here to have a good time while learning. It's not just about content. Content, content. That's what the edited YouTube videos are for. These live streams, they're more human, they're more natural. You know, I started the stream. I wish I could show you. I spilled some water just to start the stream off. I mean, you know, happens. That's life. That's okay. We're gonna do live GPT building demos here. Things might go wrong. They probably will. That's okay. We'll adjust, we'll take the feedback from the chat. It is all right, because we're live, baby. So, yeah, a lot of countries coming in here. I always love seeing this. We're so global. Canada, Latvia, Uzbekistan, Brazil, Saudi Arabia, Texas, Mallorca, or Mallorca, should I say Matthias sitting up in Majorca. Sweet home Alabama. Yeah. Good to see a lot of familiar faces in here too. What's up, Felipe? What's up, Carol? Larry. Yeah. Good to have you guys here. And Euromastro, the, you know, the runner up and last in the last competition. Good to have you in the community. Good to have you around, Jim. Jim Wells coming in strong with the what's up jerks. What up jerk? All right, but don't worry, we are, we're talking all things GPT today. You know, we'll have a little fun along the way, but like, truly, this event is going to be about dissecting my GPT builder a little more. Then we'll use it and then I'll show you, you know, what else you can do with it and I'll challenge you to use it yourself. And then actually next week we scheduled a follow up session. Well, I scheduled a follow up session just now because I figured this is, I think this is going to get really interesting. People are going to come up with all sorts of creative ideas just like they did last time. And we're going to review that. We're going to be, you know, today I'll be showing you some basics and then next time we'll be reviewing what people do with it and we can build on those basics. I mean, I suppose last time we did basics and now we're building on that. So this should get interesting. All right. All right. Good to have everybody here. I believe I have everything in place that I need right here. So let's begin the live stream, let's begin the content. First things first, what are we going to be doing today? Two things, really. First of all, we'll be using the GPT builder this time. I will go into the prompt behind it a little bit. We'll be going a little deeper last time. I was just like, hey, I'm not here to kind of explain, explain the prompt engineering behind this, what makes it work, et cetera. This time we'll do a little bit of that. Not a full deep dive, that would take like an hour, but we'll do a little bit of that. Then we'll use it to build a GPT. I'll explain some of the improvements that happened over the last two weeks. We made several tweaks under the hood of the. Of the prompt of the GPT. By the way, last time it was just a prompt. Now it's a GPT. So we'll talk about all that then. We'll be creating GPTs. We'll be customizing them. I'll be showing you some little tweaks that you can make here and there, and then I'll open it up and I'll essentially give you a list of various use cases that you could be building. Maybe you have your own. And I'll just, I'll go out there and I'll. I'll give. You shall be warned already. I will challenge you to build a GPT of your own. So. Oh, and there's one more surprise. I have one surprise ready. It's kind of like, let's see if it works. But, like, once, you know, we get through this, there's going to be a surprise. And no, it's not like some discount code sale or anything like that. It's something that I haven't seen anybody in the generative AI space do on YouTube yet. I believe you can correct me if I'm wrong, but I think I'm pretty well informed. This is going to be a first on YouTube and you can be here, and then we'll be building up on that. What I show you should be really cool. All right, excellent. Mods are in the chat. Apologies if I don't see every single message, especially if I kind of, you know, like, lecture and teach you things. It's kind of hard to have split attention. There's always going to be these phases where I ask. So, you know, if I have a question, you can. I'll be asking regularly. You can just post it in chat at any time. Mods will respond. Or then I'll open it up for Q and A and we can do this interactively. Okay, so what's up? What. Let's see. First things first, let me share my screen and let's see if ChatGPT is even up, because today has been a rough day. You guys might know if you're not familiar status.OpenAI.com will show you if your favorite tool on planet Earth is accessible or not. And it looks good. Looks good to me. Looks good to me. I mean, it's operational right now. Today we had major outages for five and a half hours. Yeah, that sounds about right. While preparing for this event, this was a bit of a pain. This is the new GPT builder. Okay. Last time I showed you the prompt, now we turned it into GPT. Now let's talk about it. Let's start talking about GPT creation. First, from the angle of me creating this quite complex GPT, and then from the angle you creating your first second. Or maybe just using this builder to enhance your GPT building experience. We'll cover all that. But let's start with this custom GPT builder. What is this? Well, you know, last time I shared the prompt behind it, the prompt behind it. You can find it in the description below. So if you check out the videos, the live streams description, there's multiple links. One of them is a link to this GPT builder. The second one would be a link to the challenge. Actually, I just realized I still have to move this to the public area. Now everybody can see this challenge. If you already want to check this out in the public area of the community, freely accessible to anybody, you can also check out that link in the description below. So that will bring you here. I'll make this a pin post. And here you have the full guide to this week's challenge. We'll be going through this together, but all the links are below. Okay, so, one, so three important links. One, the challenge. Two, the GPT we'll be looking at now. And three, if you want to look under the hood of this GPT, no need to prompt injected or whatever, you can simply look at the instructions at the current version of the instruction. We're on version 1.3. Last time I showed you, I think three weeks ago, we were on version 1.1. Okay. All right, so let's talk about this. So what changed from version 1.1 to 1.3? Well, we used it a whole lot more. And you guys used it a whole lot more. Okay. The AI Advantage community used it a whole lot more, and we got a bunch of feedback and implemented that backend. So what are some things that improved? One of them is some of the building blocks changed. So let me pull up the detailed instructions here so we can talk about, you know, the concrete prompt engineering aspects behind it. Let's get into the details here. So this link is in the Description below, it's just a notion page with the entire prompt. So this is the entire prompt. So as I said, you know, like going through all of this would take an hour, but let me give you kind of the overview and then talk about the changes. So, you know, one thing that changed is we changed up the building blocks here at the bottom. And this is like, this is where I want to start. This is what you need to understand. And you'll be hearing the word building blocks in today's livestream about a billion times. To be honest, I'll be saying building blocks. Building blocks, building blocks. So anybody you know who's about to make fun of that, you can do it now, you can do it later, it doesn't matter, I'll be doing it anyway. We'll be talking about building blocks all the time. Because this entire GPT is based on this idea of building blocks. Now what are building blocks? Well, they're different, they're this. Okay, the different modularities. It's just something I invented, something that we've been using, something that we've been using in like various tutorials and products we have. But it's an idea that I kind of, how to put it, it's different functionalities that a GPT can have formulated as concisely as possibly. That's what it is. Okay, so one of them, a very simple one would be if you want to reinforce the fact that it should use that the GPT should use the browsing tool. Well, for that we have the browser tool integration building block. Right. If you, you know, like basically the title is the browser tool Integration. The description of it is effectively use the browser tool for research, information synthesis and citation. And then there's always an example of how it comes can be implemented. So the GPT builder uses all of these various building blocks. As you can see, there's many expertise projection, there's integration of links into it, and here's a few new ones. I said there were updates to these building blocks. Okay, what are these updates? One of them is the scenario based training I found. This is a fantastic capability that you can start adding to your GPTs. And that is, you know, the ability to provide examples and role play. You know, the GPT when it has this building block, when it has this inside of its instructions, it has the ability to offer you a role playing type scenario that might help you to, you know, prepare for an interview or you know, train some sort of skill. Like a good one is, you know, negotiation trainer. Right. So if we're building A negotiation trainer GPT. This prompt will intelligently, like, look at what a negotiation training GPT should be and then it will look at all of these building blocks and it will pick the appropriate ones, customize them for the GPT, and then give you the instructions and you can build your own GPT. Okay, so in the case of a, you know, negotiation trainer, it might be very useful to have an option in the GPT where it says, hey, like here are some tips. Would you like for me to role play with you as if I'm, you know, the person negotiating against you? That's a fantastic functionality. And now we have it in here. There's a few more new ones. I wanted to highlight one more which is I quickly have to find it. We changed the structure of some of these. So for example, AI personality definition follows this traditional role prompting style where you start with the role and then lead into the target audience. Yeah, which is like the target audience preferences. That's this one. And then, yeah, Interactive learning modules. This is the one. This is also super exciting one to me because what we found is that one of the best use cases for this GPT builder is actually creating tutors, creating an AI assistant that will simply assist you as you try to learn a new skill or try to achieve something that you haven't done before. This is going to become even more relevant once the, you know, new assistants, voice assistants are going to be rolling out, but as of now, we don't have them. Nevertheless, it's still super useful to add something like this. Look, if the chatbot is designed for educational purposes. So if it's a tutor, right. Provide interactive modules and quizzes to help users learn and test their knowledge on various topics. And this is the point here. If you're aware of the fact that ChatGPT is really good at creating interactive quizzes, for example, then you can add a building block here in this prompt or to your GPT that generates interactive quizzes for the user. Right? So all of a sudden your GPT cannot just tutor the people, but it can also go in and be like, okay, you just learned the skill. Would you like me to craft an interactive quiz for you? And now it can do that, but it can only do that if you specify it in the instructions. Right. Otherwise it won't come up with these ideas. There's a few more, but these are kind of the main updates. We improved it for learning based use cases. So that's what we'll also be looking at today. Those are some updates to the building block. So I hope you get the point. If you have any questions relating to this, please hit me with them. I'm here to help. I'm looking at chat throughout. But basically, but basically, this is the point. These building blocks make up the GPTs that we will build with this. Okay, what else is in here? Okay, there's quite a bit. This is free shot prompt, free examples of various GPTs for it to know what a GPT that it should build should look like. So they always look the same way. They start with an interaction sequence, they start with a communication sequence. This is basically how should the GPT interact with the user? Should it start with a welcome message or should it get right into generating tables and working with your data? Right, depends. You can customize that in here. And then if it starts with a welcome message, as this one does, the chatbot begins with a greeting like, welcome to Career Compass, your mental health ally in the professional world. Great. Then it proceeds to step number two. The chatbot invites users to share their current career challenges or aspirations, offering support and guidance tailored to their mental health and career goals. So this is a very simple one, right? Welcome message, and then proceeds to helping people. You can make this more complicated, but this is the basic setup we have in here. Always two messages for the communication sequence. And then, you know, I wouldn't say random. An appropriate selection of building blocks to go with the GPT. So if we're creating a mental health GPT for professionals, then these are useful. There you go. So this is what will spit out for you. These are free examples. And at the bottom you have the building blocks. What else do we have in here? If we go above the examples? These are the examples. What else is here? Well, these are various guidelines, and these are guidelines on how to interact with the user right here. So this is the communication sequence of the GPT. This is what you will see. The GPT will interact with you in three steps. First step is, hey, welcome. What kind of GPT do you want to build today? That's sort of the first message you will receive in different words. You know, start by clearly defining your chatbot's purpose, goal. And then a little tip, second, it's going to follow up with some clarifying questions to primarily find out more about the goal that you have. Okay, so you're gonna get five questions. Now, you can answer these or you can answer with auto, and it will come up with answers itself. That's the second step. Third step, it will generate all the instructions for you. So this is how the GPT Builder is going to be interacting with you in these three steps, and it's going to be using some of the examples down here and the building blocks all the way down here to build a GPT for you. That's what's happening here. Last thing we didn't touch on in this prompt is all the instructions here in the beginning, various elements to this that make it work. Basically, through a lot of trial and error, I arrived at most of these, and now we even implemented a more consistent image generation at the end of the process. Okay, enough talking about theory. Let's build a GPT, shall we? If you guys have an idea for one that we could build, feel free to leave it in chat. I'll happily do that, but with that being said, let's just dive right in and, you know, let's hope that ChatGPT is working. Let me refresh and see if this is up and running. Nice. So, as mentioned, prompt available in the description, but also linked to the GPT builder available in the description. What's the advantage to using the GPT builder? Well, you know, as we go and as we learn more and as we use this more, we're gonna keep updating it. So this is version 1.3. If you just pin this to your sidebar, if you start a chat with it, like I have it here, you're gonna know that this is always the newest and best version of it. And how do you start using it? Well, simple as this. You say, let's build a GPT. Now, we did cover this last time, but we didn't go this in depth on the prompt behind it. And we also didn't give additional tips on how to use it, which I will do now. So there you go. Welcome to the AI Advantage Custom GPT Builder. Start by clearly defining your chatbot's purpose goal. Again, if anybody's just tuning in, we're building GPTs with kind of this relatively foolproof system. All you need to do is you need to state the purpose and the goal. And you guys know I like to take things spontaneously. I did not prepare an example that I want to show you here. I know sort of what works well. Tutors really work well. Mental health coaches really work well. Right. Because they don't need any external actions or any additional knowledge base files to function well. So if we do something in the educational realm, I know that that's just gonna work fantastically. So, Chat, do you have any suggestions? What kind of GPT should we build? I'm open to any ideas. Feel free to Share it. What kind of educational GPT do you want to build? And feel free to, you know, go a little crazy, like, I don't know, let's do. You know, Cat? Behavioral analysts or whatever. You're a master already. Looking forward to version 1.4, my man. Always trying to be ahead of the curve. Yes, a law enforcement knowledge center. Well, you see, that is an example of a GPT where just instructions won't suffice. You need some sort of knowledge. You probably need some sort of localized knowledge to the region where you. Where you function. Probably some laws, probably some guidelines, maybe some cases that you've worked on, some details, maybe your approach to solving things. Let's see, what else? A jazz guitar tutorial. Oh, that's a good one. That's a good one. General relativity for laymen. You could do that. That's not a topic I know much about, so I would have a hard time customizing that. One for quantum mechanics and one for Greek mythology. My man has a widespread. Has very widespread interests. I like it, by the way. I love Greek mythology. I grew up on it. When I was like 4, my grandma was reading me all the Greek mythology and I couldn't get enough of it. I always wanted to reread it. Okay, so some good ones coming here. A machine learning tutor, cybersecurity instructor, alternate medicine guru, how to be a good screenwriter. Supply. What do I pick? So many good ones. I mean, it's going to be hard to vote on this. I think I'm just going to pick a random one. Curious about Guitar GPT Coach. For planning your life purpose, how to use ChatGPT Tutor. Oh, that's a good one, Marcus. Actually, that's a good idea. That's a good idea. A custom bunny slipper sizer and designer for most creative idea goes to seafood. Yeah, Seafood die. Great. Okay. All right. How to build a YouTube channel about ancient Egypt. That's a good one, too. Jazz chords would be complicated and interesting. Soccer coach for kids. Jesus. Okay, I'm a bit overwhelmed. So many good ideas here. Let me just go with I like the ChatGPT tutor idea. There's so many good ones. The jazz guitar one was first. Let's just go with that. Jazz guitar tutor. The thing is, I really don't know much about chords. I used to play drums. You know, I'm a little bit like, simplistic like that. Absolutely loved it. Played drums for four years. Machine learning tutor. How about that? That's a good one, too. Okay, so let's do that. GPT. That creates GPTs that. Create GPTs. Yeah, that's true. We could do that. Okay, let's just do machine learning tutor or Stock analysis, a study guide. Okay. Okay, let's do a machine learning tutor, simple as that. Okay, so any single one of these messages in chat that you guys just sent, you can just go to the link and you can just type it yourself and you could go through this experience yourself. Okay, so let's go into this as somebody, you know, who. What perspective are we going to take here? Let's just do it. Let's just do a generalized one. Let's say, you know, we're like an 18 year old that is like really getting into kind of maybe, maybe we started learning Python a few months ago. We got the basics down and we're really interested in kind of the entire artificial intelligence space. But we're not just interested in Gen AI, we're interested in and kind of the inner workings of it and we want to really, really get technical with it and maybe we're thinking about studying it. So this is really a spot where this tool shines. You're going to be able to build a little tutor that's going to help you along your way. Now why I'm pointing that out, why am I fleshing the example out? Because I think if you're, you know, a seasoned veteran and you have a decade worth of experience, then a tutor like this, this is not going to do much for you. Right? It's going to be, you know, basic to intermediate, beginner to intermediate. But if you're getting into a new topic, it's a fantastic, fantastic tool to use. Okay, so as I said, second step will always be these five questions. How do we answer them? Simply like this. Let's refine the details of your machine learning tutor GPT. Okay, Would you like the GPT to act as a beginner's guide, an advanced tutor, or something else? Complete beginner's guide. And as I pointed out, you could also go ahead and just type auto and hit enter and it will auto answer all the questions if you're lazy. I do it most of the time. Okay, number two, response format. Should the response be detailed explanations, step by step guides or quick tips? I like step by step guides and I like quick tips and quick tips. How about that? Now you could also, you know, go beyond what this suggests. You could say, hey, I want every response to be interactive quiz and it will do that for you. Okay, Number three, creativity level. Do you prefer the GPT to provide creative and engaging explanations or Strictly, strictly factual and concise. You know what, let's go a little. Let's do something a little unconventional here. I want creative explanations in the voice of. Do you guys like spongebob could do spongebob here? Drum tutor. Oh, we actually just had this. Like this just came up in the community we were talking about. Philip brought up Terence Fletcher, the teacher in Whiplash. You know, like this, this. If you've. I've seen the movie for the first time a week ago. But he's the super strict teacher that, you know, wants to get the best out of people by crushing them, essentially by being super strict. We could do that. Or drum tutor Yoda. Yoda, I like it. Tarun. Fantastic suggestion in the voice of Yoda. Yeah, that's a good one. That's going to be funny. So we have we kind of creative explanations from Yoda. Number four, what should be emphasized or avoided in the responses? I don't know. I'll just say not sure. Five, any specific guidelines or key topics that should be included in the tutoring session? I have no idea. You tell me. So you can talk to natural language just like you talk to a human, right? And this is the thing. Like I'll be, you know, throwing in tips here and there for all levels. Some people are completely new here and are just getting into this. And some people have been following every Single upload since 1 1/2 years. You guys, you guys know this. You guys know that you can talk to it as an assistant. You guys know all the tricks. You might be just here for the entertainment, but that's fine because what it did here while I was blabbering, it generated everything for us. Now here's one inconsistency that we haven't managed to fix yet. Six out of 10 times, I would say, or seven out of 10 times, it generates an image. In the end, not every time we set it up, that it also gives you the image. So as you can see here you have the title of the GPT. Here you have the description, and these are all the instructions. So let's build it, shall we? What I'll do is this. I'll throw this over and then we can go to explore GPTs. Create a GPT. Configure. Okay, perfect. So. Does it want to cooperate? Ah, there's like some content policy issue with the image. Let's see. What is it trying to generate? Anyway, so let's begin with the title, right? And this is really how you do it. Like one description, two. And then all the instructions up here, starting with communication sequence, ending with the last building block down here free and you're ready to go. This is the first version of your GPT. Now you know what I'm going to do? I'm actually going to, we're going to edit it like this. This is not the best overview. What I like to do is I like to open up a document just like this one. We can delete this and we're going to edit it in here just because it's a bit more. Oof, that's. There you go. Okay, so just because this is a bit. Well, it's not perfect, isn't it? How about, I mean, okay, then let's look at it in here. Basically, this is our GPT. Okay, so let's review what was generated here. As I look at the chat and what you guys are thinking right now, Yoda is planned for GPT5. Yeah, I'm getting ahead of the curve here with the Yoda voice. All right, so let's look at this. Yeah, so this should be fun with the Yoda voice. So what did it generate here for us? First of all, there's the communication sequence with two steps. First of all, it's gonna begin with a greeting like welcome to the machine learning mentor, your guide to the mysterious mysteries of machine learning. I am. Gosh, now I should have like a Yoda voice filter. That would be great. Maybe next time. But that's how it starts. Welcome message. Okay, following up, the chatbot invites users to share their current understanding or questions about machine learning. Offering step by step guidance and quick tips in a creative and engaging manner. Perfect. Very simple, you know, welcome message. One, Help the user out. Two. Moving on instructions. This is always the same. This is just a great way to initiate the GPT. And then here is where all the magic happens. This is where it gets its personality, its behavior, all of that. Quite simplistic, but very effective. Now of course you can extend this as much as you want. Not for the challenge though. Stick around for that there. I have some guidelines, but basically various building blocks. What did it pick? Okay, personality definition. You're an expert in machine learning, providing step by step guides and quick tips with creative explanations in the voice of Yoda. Amazing. Looking forward to that scenario based training. See, it used this building block that we highlighted earlier. Guide users through basic concepts such as supervised and unsupervised learning, feature selection, model evaluation, and more using creative and engaging examples. Nice. Creative and engaging examples. It's echoing our decisions in the, in the questions, in the clarifying questions. Remember that? Create a chatgpt. I will. I like that comment. Okay. Structure, response and tone. User structure. You know what? I actually like that comment so much. Well, there's this feature that I don't use enough where I can throw comments on screen. Fahita, you're with me on stream. Good to have you here, man. I believe this is going to fade soon, but anyway, cool. Okay. What else did it put in here? Structured response and tone. Use a structured format for responses, maintaining a creative and engaging tone mimicking Yoda's speech patterns. Okay, so you see this echoed in like multiple parts where it matters. Interactive learning modules offer exercises and quizzes to reinforce learning and help users practice machine learning concepts. There you go. That's what I was saying. If you're building a tutor, it's going to use this. If you're building something that is not educational related, it's going to leave it out. It's smart like that. It's going to make these decisions for you, but you can make them by yourself by understanding how the prompt works. That's why I studied that way. Okay. Professional resource recommendations. It suggests articles, books, online courses based on the user's current understandings. It adapts the files you upload. These are basically toolings that we include. I found it's really good to include these building blocks to reinforce that behavior. Sure, you can always call up on the browser tool and if you type in something concrete like hey, use the browser tool to find me free articles on this topic and then you give it a topic, it will pull up the browser tool. But I found that if you include it in the building blocks, it will autonomously do that more often, if that makes sense. You know, you don't have to specifically ask it to do that. It will just keep considering in its responses that should I use the browsing tool? Should I use the image processing tool here? I mean, sure, the image processing tool maybe not as like not as essential as the browser one, but I really found this to help. Actually it used the browser tool more often. More often. If you have this in here. And that is really it. Simple as pie. I believe everybody can customize this, right? If you want to customize the welcome message, you can customize it here. If you want to delete one of the building blocks, you can delete them here. If you want to add another one, come up with one of your own. We're going to talk about that soon. Then you can do that down here too. And basically we copied over all of this into here and I edited nothing. Right. So all I'm going to do is say I like to always have some sort of conversation starter to make the GPT more user friendly. And then. Do we need dall e here? Probably not. Do we need code interpreter? Maybe. So I'm just gonna do this. You can just turn all three on by default. That's fine. That should be just fine. Ooh, one more thing. Our image. Did it give us an image? It didn't. There's some problem with following the content policy, so. So we can just generate one up here and then we're ready to go. So we're going to test run this and then we're going to talk about customizing the building blocks. Okay. For your to your own needs. And then we're going to be. We're going to go further. Is the broadcast recorded? Absolutely. Absolutely. Rusty, what's up saying ask to output it in markdown to copy it easily. Exactly right. So these are the kind of the extensions we can make. We'll talk about that in a second. By the way, shout out last year. Great to have you in the community. Great to have co worked on some of the products we sell. And also thanks for some of the feedback on structuring this thing. Yeah, some of the. Some of the formatting Markdown is king, ladies and gents. Markdown formatting works super well. Okay. Okay. Are we ready to go? Do we have an image? Hey, what's up chatgpt? Where's my image? Use Dali. Come on, give me my image. These wonderful people want to see the GPT at work. What's up with that? Is Dall e down? Usually this works. Is there some issue again with them operational? It doesn't seem like it isn't down. Okay, you know what, we'll just ask it for the image prompt. That's a good suggestion. That's a good suggestion. We'll do that. Okay, you know what, we can just skip this step. We don't necessarily need an image. The image gen has been stupidly slow lately. Ah, makes sense. So this is what it did. What's wrong with this? Resembling Yoda holding a book. Ah, it probably has a problem with Yoda. Huh? I see. Yeah, it's like, hey, this is a copyrighted character. That makes sense. Why I wouldn't want to do that. Let's do. You know, now I need to visually describe Yoda. Oh God, this one's tricky. Can you guys help me out here? What do we. How do you describe Yoda? Or a child sized alien with green skin? And pointy ears. I don't know. I need a reference image of Yoda. What else? With. With a white hair growing out of his ear, I suppose. What else? What do I describe? How do you describe Yoda in, you know, in an outfit that a homeless person would wear, I guess. A wise and old character. Okay, let's see if we get this. That looks like Yoda. You think that works? We can try that too. Okay. Funny looking, child size. Jack Russell lives in a swamp. A lizard like child alien character. Short with green skin. Yeah, Tara, good example. Good suggestion. Yeah, just ask ChatGPT what Yoda would look like. Oh, there you go. Okay, just some intuitive prompting here. Let's go. Absolutely roasting Yoda right now. My bad. Please don't. You know, sir Yoda. Sorry, I am. Okay, okay, that's. There you go. A wise old character. Let's see. Let's see if this prompt is better than, you know, the homeless person with. With air. With gray hair growing out of his ears that I suggested. It is close enough. I mean, I gotta say, this is. This is how you get around these things. Okay, good. Let's get back on track. Let's complete this GPT, shall we? So let me just upload this photo. Oh, oh, okay. ChatGPT wins. I gotta, you know, gotta keep it real. Yeah, yeah, yeah. This is. This is better. Nice job. This is how you get around, like these silly, silly restrictions. Oh, I can't generate an image of Yoda. Describe Yoda and then generate it. Okay, no problem. Please. I mean, to be fair, it doesn't look entirely like Yoda. Still like this guy. Okay, let's get into this. So let me full screen this again. And here we shall. We shall test it out. Okay, hit create. I'm gonna make this, you know, public anyone with this link update and then I'll share this in chat with you guys if you want to, like, talk to Yoda and test this out yourself. But let's give it a spin. Let's see how this GPT that we built with, you know, minimal effort. I mean, you know, I'm yapping on here, but this would have took two minutes if I would have been efficient. Okay, copy link. I'll share this in chat with you. You guys can check this out also, like one. One sec. Okay, so now we can preview this and see how the interaction looks like and then we'll get into customizing it. Okay. View GPT. Okay, so all I'll do is say, let's begin. Right? Let's have a look welcome to the machine learning. I don't know. I can't do Yora voice. I have a bunch of voices here but not Yora voice. I could do. I could do, you know. Small robot. Welcome to the machine learning mentor. Your guide to the mysteries of machine learning. I am share your understanding or questions about machine learning. You will step by step guidance and quick tips. Offer you. I shall. That's pretty good. Okay. Yeah, let's just start. I'm completely new. Where should I start? Let's see. Let's just interact with it as a user. Would you to machine machine learning. You are. This Yoda suggestion actually kind of made my day. This is great. Begin at the basics. We shall. Sorry, I'm not going to try doing a Yoda voice. I'm not good at it. Understand the key concepts we must learn. You will. What is machine learning? Okay, here you get the basics. So let's have a look. Okay. It gives you some basic concepts and then look at this next steps. Practice with simple data sets. You should try tutorials and courses I recommend. Ready to dive into a specific topic or a concept. Are you? Ask more. You can. Hmm. So there you go. Now let me just pull up a quick comparison. So for anybody who's new here. So you understand if you simply do this in ChatGPT, what would you get? I'm completely new. Let's give it a little bit of context, you know, to machine learning. Where should I start? And let's compare these two answers, shall we? It's giving you kind of the steps which is fair. I think this is slightly better representation for a beginner. So I would prefer this. But here he just goes all over the place. So this is more tailored to, you know, a beginner, a beginning student rather where this is just hey, here's all the things you can do. And then this is kind of the essential part, right? Because like you're talking to an assistant, you're talking to a digital assistant here. You want it to interact with you. So this I always have like have my eyes on the last part of the prompt. Like how does it, how does it ask me to continue in a conversation? This is the big problem with like this used to be the big problem with 3.5 by the way. People got into it, they sent it a basic prompt and the result was mediocre and it didn't prompt you to continue. So people gave up on it quickly. Right. You really have to get into this mindset of this conversational mindset. I always say like think of it as A super highly qualified and expensive assistant that you hired. Right. If you don't like something, give it feedback, go back and forth with it. You'll figure it out. So let's see, how did it fare in the end? By following these steps, you'll build a strong foundation in machine learning and be well equipped to tackle more advanced topics and projects. Good luck on your journey. Okay, so essentially it's like, yep, good luck out there. Have a good one. See? And a typical user, if you give it this experience, they would be like, oh yeah, okay. ChatGPT is all right, but it makes all the difference if it goes ahead and says practice with simple data sets. You should try tutorials and courses I recommend. Okay. What to reels or courses? Not sure how to continue here. Okay, Okay. Legally distinct Yoda. That would have been a good one. Yeah, yeah, yeah. Lawyer Yoda. Hmm. It's my new catchphrase. I like it. Yeah. Whenever it's such, everybody should actually replace their ums and ahs or their weight, their like with. I might start doing that. Let's see. So one thing that it did here, it gave me links. Now I'm not sure if these are going to work. Sometimes they do. Sometimes they work. Look at that. Remember in the GPT instructions we told it, hey, provide me with links to resources. It just did that. Whereas the original version didn't do that. Right. There's no, there's no links to the books. But here it's like, yep, here's. Here are a few books and here's the link to it. Now as I mentioned, like some of these, they don't always work, but often they do, as you can see. So that's great. That's a great little feature. Okay, next steps start with a course. You should practice on Kaggle or Collab. You can specific questions or challenges. Bring them to me. You may guide you further. I will. Ready to begin with a course? Are you or have a specific topic on mind? Ask away. You should. Not sure. Can you help me practice? Let's see. And let's do the same here. Not sure. Can you help me practice? Let's see the difference. So here it tells you. Okay, let's start with a simple machine learning project. Step by step guidance I provide. Excellent. See what we're predicting? Housing prices. Same thing. Predicting housing prices. Good stuff. Both cases there's a summary. Run the code in your environment. Ask. I think you get the point. See, it just does these modules at the end, it's built for user interaction rather than being built to answer a question. And that's kind of it. Right. This makes all the difference in a tutor. Tutor, I think. Right. It's this interactive aspect. It's this aspect where it actually prompts people to do more. Okay, perfect. So, good news. The little surprise demo thing we have is looking like it's working. Excellent, excellent. So I'm going to be a little surprised here for you soonish. Yoda doesn't know the meaning of verbal predicates. Yeah, fair enough. Fair enough. That's probably true. So I think you now get the difference between just using ChatGPT and using a GPT build with this builder.