Chat GPT Webinar by Igor Pogany - 2 August 2023 - Organised by World YMCAs

Foreign.

Good morning, good afternoon, good evening to you all. Depending on where you are joining from I bring you greetings from World YMCA and my first disclaimer is to tell you I'm real, I'm not artillery as artificial intelligence I'm not produced by anything some real and I can drink. The AI does not drink. So this is very serious and very real. So welcome.

And today we want specifically to focus on ChatGPT. We are not discussing AI today because we want to go to the practical entry point Something that many of us we are trying to understand. So the purpose of today's mainly educational is about understanding how best if we wish to, we want to use this tool called Chat GPT I'm also aware that many of you are using it already and sometime I just realized the other day talking to people that I only use a very small part of it. So there are many opportunities that we could use to improve our work. How do we get this access, this knowledge and how can we learn more about it? So today is very educational we are aware of a few things that not all the country have access to ChatGPT as this is a reality we need to be aware of and we also are aware that there are a lot of ethical issues, a lot of access, a lot of many related topic we are not dealing with that today Is it ethical or not? This is not the point of today Today is those who are already using it, who are planning to use it, who just hearing it, don't know what it is is to be able to know what it is and then to decide for themselves where to go from here. If there are any subsequent question and conversation we need to have, we will have that another time. So in our quest we'll give this information to you. I have asked Plaud and I will head of Digital Transformation to source a capable person to help us to deliver this. We also know that different YMC have people in the YMC who have that expertise. We have some conversation and at the end we decided to go outside because someone would do that all the time. So this is my pleasure first to acknowledge about the high turn off of you. There are many of you here Today is one of the highest registration we have. I want to invite Claude and to introduce us to ego is going to talk to you about ChatGPT today. Claude, over to you.

So hello everybody. So today we will have Igor Pogany He's a young entrepreneur, he's 29 years old so according to the YMC standard he's still a young person so I'm very happy to have today a young person and a young entrepreneur. So this is the first thing I want to say. And Igor defined himself as an avid tech enthusiast. So this is something very important to be enthusiastic in life. And he's passionate about learning and exploring emerging technology. But one of the most important mission he has is to break down the complexity of AI and to make it practical to enhance our daily life. I think it's a nice mission and motto. So for doing this, he has a YouTube channel, AI Advantage with 146,000 subscriber and one of these most famous video ChatGPT in two minutes received more than 3 million view. So I think it's quite something, quite impressive. But behind AI and all these kind of things, he's also passionate with other things like snowboarding, kite surfing and carpentry. And he's multilingual and English, German and Slovak. And I will finish with the quote I like from him. For me, so I'm quoting him. For me, the key to success always was and still is embracing change rather than fighting it. Igor, you have the floor. Thank you.

Wow, thank you so much for that beautiful intro. I truly appreciate that and I'm honored to be speaking to such an international crowd. I mean, I'm just looking at the chat and it's truly impressive. I don't think it gets more global than this. So first of all, thank you everybody for showing up. And as has been mentioned, we'll be looking at the practical aspects here. And right off the bat, I want to start with a promise to you. Okay, so we're going to have a roughly a one hour session and then we're going to do a little Q and A in the end. So anything that is directly related to the things we talk about you can ask in the chat. You know, I'll keep an eye on it, but mostly I'll try to get through the material I prepared here. And in the end you can ask your other questions. But mainly I want to make a promise to you. So the promise is going to be you're going to walk away from this webinar with free use cases. Okay? You're going to have three things that you can put to work, no matter your technical expertise, no matter your previous experience. Okay? There's going to be three things you can use. And the beautiful thing about this is, yes, we're talking about ChatGPT, but you could substitute ChatGPT with large language model with.

Generative text AI. Okay? These skills, all the prompts we talk about Today, all the frameworks and a little bit of theory we're going to be talking about, don't worry, that theory is going to be very quickly applicable to what we're going to be doing. All of that applies to any other language model because at the end of the day, you might have realized it doesn't matter if you're looking at a image generator or a music generator, they all take human natural language as the input. So what we're going to be looking at today is making your inputs a little better or showing you some things that you might have not thought of by yourself. And also I crafted this in a way where it's going to be extremely accessible to people that have never used it before. But also the people that are quite experienced, I'm sure are going to find a trick or two in this little presentation I'm going to hold here today. So with that being said, keep that in mind, we'll dive into a little bit of theory first because you need to understand how, how it thinks and how to communicate with the system. But at the end of the day, you'll walk away with free use cases. And not just that, I also have like a. I'll give you a notion template and everything we talk about, you don't need to take notes. Okay. I would really like for you, I mean, if that's the way you learn, go ahead. Not going to stop you. But I would really like for you to kind of listen and to, to experience what can be done here rather than like meticulously note down every little thing. In the end, I'll provide you every single line I'll be using here. Okay. It's all to be there, ready for, for you. But essentially I want you to gain an experience today that is going to show you that, wow, these things are possible there, are extremely simple and I could be using them too. Why am I not using that? Well, that's that I want to address one more thing before we begin, and that is the concern that some countries might not have access. And that is just the nature of this. I mean, there's. IT has over 100 million active users, but not every country in the world has this. But again, these skills translate to other language models so quickly. In the end, I'll give you some recommendations for like, like a learning roadmap, so to say. So we'll talk about prompting a little bit and in the end I'll show you where to go next. But right off the bat, I can just tell you there's competitors that work in certain countries where ChatGPT doesn't. So one would be Claude 2. Claude, that might be your personal favorite. But that, that is the AI model by Anthropic which you can use in certain countries. So you can check that out. There's also Bard by Google. Right. And many more popping up. There's now llama2 by Facebook. Whatever. We're getting a little too concrete here. I want to start zoom, zooming out a little bit and I want to start introducing you to some of the base concepts here. Okay. Because at the end of the day, what we'll be doing over here is the following will be talking about this little interface for the foreseeable future here. So what is this about? I mean, at the end of the day, even for more experienced users, what's really going on here? Well, you get to interact with a so called large language model that has been trained upon the wider Internet and you get to interact with it the way we are best at interacting with things, with human language. Right. Doesn't matter how introverted you are, doesn't matter where what other personality traits you have. You have the ability to speak and to communicate with people. Right. Otherwise you wouldn't be understanding the words that are coming out of my mouth right here. So all of us have that ability. And this presentation, what is really going to be about, it's about making it clear to you that you already have the skills that it takes to get results here. Okay. Because most people are intimidated by it, right? They get into it and they're like, oh my God, there's this field of prompt engineering. And then, oh, they talk about these advanced topics and like, what is this essay that he just posted in there? Like, how did he come up with that? Like, that is, that is not the right way to approach it. The right way to approach it is like, hey, if I had a person come in here and they would like, their sole purpose would be to assist me in whatever I do and they would be paid for by somebody else. Right. I could probably find a way to explain to them how to get certain things done for me. Right? So that is what this is essentially. And that is my first big point here. Okay. The way to think about this is not as some extraterrestrial alien life form that now came up on Earth and now is here to help us or whatever. No, no, no. The way to think about this is as a helpful assistant. Matter of fact, that is not something I came up with. Right. If you look into the development side of ChatGPT, because there's also a way to kind of integrate it into your apps and everything. That's not going to be the topic of today, but if you look at that, there's a thing called system message there and that's essentially the personality of the system. And you know what the default system messages, it's very simple. It says you are a helpful assistant and that's exactly what you're getting here. Okay? So if you're using chat.OpenAI.com, which is the website that ChatGPT runs on top of, you're talking to a helpful assistant. And really that has been so helpful for not just myself, but also many students, because there's so many ways to conceptualize this and everyone comes in with kind of their own biases and everything into interacting with this. But every time you get stuck, if you just lean back and think about the fact that, okay, all I'm doing is asking a favor or giving a task to a helpful assistant, a lot of the question marks you might have will simply resolve themselves because we know how to talk to people already, right? So this is the framework that I want to present. Like number one, ChatGPT. You can see it as your helpful assistant. Matter of fact, if we want to get a little more concrete, you can think about it, okay? You own a company and you just hired a brand new assistant, very well read, right? Educated upon the vast knowledge of the, of humanity across the Internet. Right. They know all that, but they don't know anything about you and they don't know anything about your company. You don't have to pay them. Right, because it's free. Sure, there's a premium model, but basically like ChatGPT 3.5 here is free, right? So somebody is just this talented, all knowing assistant, showed up in your office and was like, I'm ready to work for you for free 24. 7. That happens. Most people will be like, whoa, what's the catch? Well, what's the catch? There's no real catch. They're kind of. Yes, they're using your inputs for research purposes. That might be one of the few. Yes, there's all these ethical considerations that we could talk about another time. But at the end of the day, those deal with things that might happen in three to five years, whatever that timeline might be, right now have a helpful assistant that is there to help you. So let's do that, let's take advantage of that, right? Why not? Now the question presents itself. Now what do you ask the helpful assistant? Well, it can't do Everything, right? If I tell it, get me a cup of coffee, won't be able to do that, but if I tell it, write me an email, it will be able to do that. And that's where we transition from, like the theory into the. Into the practical part, which we'll start in a second here. But before we do that, I want to. I want to give you one more. One more theoretical concept. Because thinking about this as a helpful assistant, really, it's such an unlock, trust me, like, every time when you get stuck and you just think back, okay, what would I tell an assistant that would be sitting next to me? Well, that will open up a lot of doors for you, but you need one more concept to really understand how to interact with that assistant. Because it's. It's really quite simple. So if you look at my screen right here.

It really comes down to this. Instructions plus context. Okay? And this is the formula for every single prompt out there. Okay. Every single prompt. Igor, really? Yes. It's always instructions and context. What does that mean concretely? So maybe the most basic prompt you'll see is write me an essay, right? And then there's a topic. So write me an essay about a plant, right? Doesn't matter, whatever topic you pick. But at the end of the day, there's going to be two components, right? There's going to be the instructions in the beginning. Write me an essay, and then there's going to be the contextual part about a plant, or write me an email to my boss about the coffee machine, about the grant proposal that we sent out yesterday. Right? But if you do that, you have to realize that the instructions are quite easy to arrive at. You know, a quick Google search, a quick look for YouTube or Twitter, you'll find a bunch of these prompts and instructions that people give out and market themselves with. That's not hard to get by, Right? And if you're. If you're a professional or whatever you might be doing, chances are you'll have a rough idea of the type of things that you would like this assistance to help you with. Right? So if you're, for example, responsible for the communications within your company and you maybe receive emails, well, just makes sense to kind of go with a prompt like write me an email, right? So the instruct. What I'm trying to say is the instructions part is pretty straightforward. That's no rocket science, right? Where it gets interesting and where it gets really deep and where a lot of people get lost is the second part of this equation. It's the it's the context. Right? Because again, instructions, just to name a few more examples, you know, write me an email, write me an essay, rewrite this text, or.

We'Ll actually look at a bunch more like over here, I'll give this document to you in the end. And we have some examples down here that we'll talk through one by one. But essentially all of these, even if I look at the entertainment prompts down here, right? Jokes. Can you provide a joke about, you know, xyz? The instructions here are, can you provide a joke? Right, but then the second part, this is the tricky one, the context. Right. Okay, so let's talk about this for a second, and then we'll dive right into the practical part. But we really need to talk about context, because the art of using this well and getting actually actual results that are useful to you and not just something generic that might be useful to a beginner in the domain you're navigating, but actually to you, it's all hidden within the context. Okay, so let's try and unpack that a little bit. What does context mean? Well, you know, let's just get very practical and let's talk about what is the context of the situation that you see right now that I'm in. Right? The context is, well, I'm holding a presentation through a webinar. The context is I have a laptop from the brand Apple on my. On my table. Right. What else is the context? Well, there's a plant over here. Well, there's some lighting going on. There's, you know, there's a board behind me and there's some, you know, all of these things are context. Another piece of context is we're communicating in English, right? Another piece of context is if you zoom way out, it's like all of us are on planet Earth. And yes, that might be relevant to many conversations, but not to all. When you're writing a sci fi story, you would want to mention that Igor that is holding the webinar on planet Earth. All of a sudden, that piece of context becomes relevant because you might have some other planets involved in your sci fi story. So it really depends on what you're doing, how much context you want to include. But there's an infinite amount of context, right? You could start counting the amount of fingers that I have. You could start describing the shirt and the texture. You could start describing the headphones. Like, you can get as detailed as you want to or as surface level as you want to with your context. So that presents the ultimate question. And this is where I try to arrive at. It's like, what is the right amount of context? Well, there's no definitive answer to that, but I can tell you two things. No context at all is not the way, and all of the context is also not the way. So what does that mean in practice? If I go in here and if I start replacing this with. Let's just do the most basic prompt, right? Write me an email.

And then what do we include as the context? Now let's say to Claude about the ChatGPT webinar. All right, that's not bad. That's a beginning, right? So we have some amount of context over here and then we have instructions over here. But what happens if I drop one of these? Why are these two so essential? Well, if I just say to Claude about the ChatGPT webinar, what is ChatGPT going to tell me?

It actually does it. How does it do it? Well, it assumes. Okay, what about the other case? Let's talk about this in a second. But what about the other case? What if I just say, write me an email, save and submit.

There you go. So now it's like super general, right? So, dear recipient's name. It doesn't know that I'm talking to Claude. I hope this email finds you in good health and high spirits as the organizer of the. Whatever. It doesn't know anything right here. Date, venue, start time. It just assumes we're doing some kind of event.

It's not going to do a good job if you do this. Because take a step back. What would the helpful assistant do that you just hired? If you told it, if you told him or her, write me an email, they would be confused, right? They would ask, okay, no problem, I'll write you an email. What is the email about? Right? But if it was a really, you know, really ambitious and hardworking assistant, they wouldn't want to bother their boss, which is you. When you're typing into this. They would just write you an email, right? They would do this, they would say like, okay, we don't know what the recipient is. We don't know what the, what the content of the email is. But I'll write a draft and I'll leave a bunch of the things open. And then, you know, my boss can review if that is what they. What he wanted. So when people talk about these, like, AI hallucinating, right? It only hallucinates because it's set up in a way where it tries to help you. And if you tell it, write me an email and you don't give it, give it any context. It's gonna do that. It's there to help you. Right. It's not gonna refuse. It's not gonna say, like, nope, not gonna write you an email. You need to give me more details. No, it's actually gonna do its best to be helpful, and it's gonna write some draft. And when people talk about hallucination, what they mean is they asked for something that they didn't provide enough context on, and then the AI, like the language model, has to start making things up. So I hope that makes sense as a beginning. Like, it's really this concept of instructions and context that is so important and will be important throughout the presentation. So, again, so if you put these together, then we'll get the ultimate result because, you know, we'll get an email to Claude about the ChatGPT webinar.

Good. So, and now it will have to assume, like, okay, what is the ChatGPT webinar going to be about? And if I look into here, obviously it has some more information on it. So, you know, let's just have a brief look into this. The subject line, invitation to join our ChatGPT webinar, Unlocking the Power of AI Language Models. Okay, that's a good title, right? I hope this email finds you well. I'm writing to personally invite you to an exclusive webinar that we are hosting. And the name. So it assumed the name. And this is my next tip. Okay, so when we'll talk for the use cases in a second here, you'll see that I provide a certain prompt to you that you can certainly use. But the magic of doing this well really lies in understanding what type of context could help it. And the answer to that question often hides within the answer. Okay, what does that mean? So if I say this, write me an email to Claude about the ChatGPT webinar. If I start reading and I just use common sense, what is the first thing I noticed? Oh, it's, you know, it assumed the name of the webinar. So what could I do here? Well, I could extend the context. Well, that's a little theoretical, but what would I do if I talk to my helpful assistant? Right? Like, I would tell her, write me an email to Claude about the ChatGPT webinar titled, let's say in this case, you know, conquer ChatGPT. Right. And if I do that, it's going to do the same thing as before, but all of a sudden, there you go, it uses the appropriate title, right? And then you can do more. What the. Like all these things in square brackets. It doesn't know, right? It doesn't know the date, it doesn't know the time. So what could I do? I could just take all of these and just for demonstrational purposes, I'll, you know, I'll, I'll just say mark a second.

Let's say that. And then.

One hour platform zoom and then test link xyz. And my point is, it doesn't take any engineering skills, it doesn't take any advanced skill set to do this. You just got to pay attention what your assistant replied. And if it doesn't know certain things, well, go ahead and refine your prompts to include those. Right? We established a lot of concepts here in a short amount of time, right? We talked about how you're supposed to think about ChatGPT as a personal assistant. And what that also means is that you need to provide it with some of the information that it doesn't know by default. So a lot of our conversation here is going to be resolving around the fact that, okay, what's the information that it doesn't know? How can we really help it out? Because if you just say, write me an essay, it's not going to know what essay. And there's some crucial details that you might be leaving out. But the beautiful thing is, and I'll go to this next. So there's a bunch of prompts that I prepared here.

And I'll just show you one that has been really going viral across the Internet over the past months, because it's a really simple one. It's this one, okay. It's the context finder. And what we're going to do is we're going to send this little prompt and again, all of that is going to be in the notion template that I'll just give you a link to. It's free. Don't worry, you don't need to copy all this. It's going to be all in there. But basically you can see this, right? And this is really. And now we went from how to get started to quite advanced territory. But I think you guys will be able to handle it. Because if you understand this concept of the assistant and the assistant not really knowing what exactly you demand from him or her, then you're going to understand that following up with a question like, now ask me questions so you can gather all the context you need to answer the previous prompt. And if I just send this, well, there you go. It's going to give you a bit of a questionnaire. And what I can do now, certainly here are some questions to gather additional context and. Yeah, so I can just go ahead and I can start answering these one by one. And you know, if I answer these, it's going to readjust the answer above and it's going to be better. Okay. But I just wanted to give you a glimpse of that, of how we would move on. Let's not get too deep into that because let's return to kind of the basic level of use cases. So with that being said, let's get into use case number one here, shall we? And that one is not writing. The thing that I want you guys to realize is really that it's rewriting. That is the superpower here, okay? Because if you tell it, write me a grant proposal, it will do it, but it will miss all the little details of the context that you might want to have in there. And it will also miss your personal style and everything. Right? So what I propose to you is instead of going ahead and asking it to do all of the work, ask it to do the heavy lifting, right? And if you're writing something, what is the heavy lifting? Well, it might not be hard to come up with a sentence or two that describes what you're looking for, but it might be hard to get that to a form where you're ready to send the email, to polish it, to readjust it. So that's why I would propose this very, very basic prompt, right? And we can even start simpler. Rewrite the following text. And then you can always special, you can always say which style. But we'll just start with this. Rewrite the following text. And here I want to kind of demonstrate the abilities of the assistant, because this assistant that we're interacting with here is a very sophisticated one. It's very well read, very intelligent, and it really does its very best to understand what you're doing. And again, it's a helpful assistant, right, what that means. So it's always going to try and do its very best to serve you. Okay, so what would a helpful assistant do if you give it a message that is barely English, Right. Well, it would try to decipher it. So let's do this. Let's have a little fun here. And let's actually type in something that most people would frown upon. But we're interacting with this helpful assistant that doesn't judge, right? So we can just go ahead and we can completely write something very basic here. So rewrite the following text. Hello, I would like.

Donation.

Good day, sir. Okay, so look this is not a message you could send. Right. But you could go ahead and say rewrite the following text. And then let's see. And then let's specify instead of text we're going to say email. Okay. Because it's going to be an email. So let's just send this and see what it comes up with.

Oh, that's a whole lot better. Right, so we went from hello, I would like a donation. Good day, sir. Right, so we had a spelling mistake. In here we have the most basic of English, but it goes ahead and it fleshes it out, right? It expands it into something. And how long did that take? Like from copy pasting that prompt and writing hello, I would like a donation. Good day, sir. To hitting enter.

Five seconds, maybe, maybe 10. Maybe 10, right. But not 15. So it's like, now we have a first draft, right? That's not bad. So let's have a look at the draft. So it doesn't know the recipient's name. Fair enough. I hope this email finds well. I'm reaching out to humble requests or kind considerations for donation. It doesn't know what donation we're looking for. Right, so it doesn't. So it, so it kind of leaves that blank. But the point is you can rewrite things super effectively. And now if it the point, like if it works with something this basic, you bet that if you feed it a well written blog post that it will be able to rewrite that too. You bet that if you feed it some of the email communication you have already done or some other templates that you have used for emails that will be able to do and rewrite that too. Okay, so let's go a step further because now we didn't really specify what type of donation or we didn't specify and what the languages that it should use. Right? But here's one of the most beautiful things and one of the most. One of the biggest capabilities of ChatGPT, this thing, it's able to mimic styles. And because it has been trained upon the wider Internet, it has a whole lot of styles that you can choose from. Okay, so let's just go ahead and insert a part into this prompt here. And this is how it works, by the way. There's this little button where you can like edit the prompt. You could always just make a new conversation. Right. And send a new one. But I just find this to be nice and convenient. So we'll just edit the prompt and I'll just reinsert the thing from the notion template that you guys will be getting but essentially we're going to say in a formal and friendly style. Okay? So we're not just picking one style, we're picking two and we're kind of blending it because if you just say formal, it's going to be very, you know, it's going to be very strict and if you just say friendly, it's going to be too loose. So often I find it to be most effective when you combine two styles. Right. What styles can you use? Well, you'll find that in this notion, templates. There's actually a category here where you can find 25 different styles. If you just check this out over here, then you'll see you can do formal, informal, educational. You could write in the style of J.K. rowling, Charles Bukowski. You could write in the style of Nike. Right. You could imitate brands because it has been trained upon all the Nike marketing materials, so it knows exactly how Nike communicates. Might not be useful to writing an email, but might be very useful to creating a marketing campaign.

You can also do magazines or journals. So this is just here to kind of jog your brain to show you what's possible. These are all the styles you can write in and there's many, many more. Okay. So.

You can really go wild and you can really start getting creative and thinking about, okay, what publications do I enjoy, what celebrities do I do? I enjoy, like the way they speak. Right. You can, you can start mimicking that if ChatGPT is aware of them and if there are a list celebrities, chances are it will have been trained on some video transcripts, some articles, some movies from those celebrities. So it will be able to like mimic characters, personalities. Okay. Back to. Back to earth. Right? So rewrite the following email in a formal and friendly style. So we just use two styles. Hello, I would like a donation.

Okay, let's just add the D here because that's painful to look at, but good day, sir. And let's just submit this. So let's have a look at this. What does it do? What does it change?

Well, you will find that this is not the same output. And here you can always toggle before and after. This is a great little trick on how to compare. That's why I also wanted to show you guys this feature. But essentially you're going to find that it's a different style, it's definitely a different length, it picked different words, but it's not that different. Right. And why is it not? It's not day and night. Right. If you look at the first paragraph, I hope this message finds it. Well, my name is XYZ and I'm writing to inquire about the possibility of receiving a donation for a noble cause. Okay, what was the first one about? I hope this email finds it. Well, exactly the same. Right. I'm reaching out to humbly request your kind consideration for a donation. One could say that is essentially the same. Sure. A little different wording, but you know what else will be worded differently? If I just resubmit the first prompt without the style, it will word it differently. Because what your helpful assistant does is there's a certain level of thinking going on when it generates these and when you think, you kind of arrive at different outcomes. Right. Especially with something like an email. So every time I regenerate this, I can just press regenerate again and again. You will see, you know, this time I hope this email finds you in good health and high spirits. Regenerate. I hope this email finds a good health and high spirits. Same thing. But let's do it one more time, one message at a time. Okay, just reload.

But you'll see that it just varies slightly every single time. Now high spirits are gone. And it does that because at the end of the day, this assistant, all it does is always predict the next. For our purpose, you can just say word. In reality, it's token, but let's not get into that. But it just predicts the next word. So every time you'll regenerate this, a new set of probabilities will be assigned to the next most probable word. Okay, might be going a little too deep, but I kind of want to touch on these topics that surround this mysterious topic of AI, because I feel like that can be helpful to understanding all of this. Because as you saw, we ran two different prompts and the results were quite similar. Right. The first paragraph, arguably the same. Like slightly different wording, but the same. Why is that? Well, if you remember what I told you earlier, that helpful assistant is there to help you at all costs and it's not going to tell you that it can't do the work unless it goes against OpenAI's policies and it's something illegal. Then it's going to tell you, hey, I can't do this, but otherwise it's going to do its best to help you. So what would a helpful assistant like that do? If you tell it to write like this first prompt? Right. If you tell it rewrites the following email. Hello, I would like a donation. Good day, sir. Well, it would make certain assumptions. It would assume that you're not trying to be very funny in that email, it will assume that you're. You might just like when you would present the intelligent person with the question of, okay, what tone and what style should we write an email that requests a donation in? Well, intelligent person would probably answer. It should be formal, right? It should be also friendly. It shouldn't be hostile, right? So when I run this prompt, rewrite the following email in a formal and friendly style, it produces the same result because the assistant assumed that it should be friendly and formal. Now, what happens if we change this, right? What happens if I go in here and we say, you know, in a informal and unfriendly. Let's not even go with hostile, because that's even stronger. But an unfriendly style. Save and submit.

There you go. That's it. That's the email. Hey, listen up. I need a donation from you. Make it snappy, buddy. Right? Very different story. And ChatGPT is smart enough to realize that, hey, usually if people, like, request donations, they're formal and they're friendly, right? They're not going to be writing things like this. But the point is, you got to start realizing these assumptions. You got to start realizing what context is involved. Because if you tell it to write you an email that requests a donation, well, there's a whole lot of context that comes with that, right? Like, asking for donation is asking for free money for a good cause. That means there's some goodwill involved. That means there's a certain tonality that is appropriate. All these things that we just take for granted as humans in our everyday, right? All of these, all of this, like, context that we kind of, you know, we put on the blinders and we just phase it out because most of the time it's irrelevant, right? As I mentioned earlier, it's kind of irrelevant to this conversation to mention that we are, you know, all on planet Earth, right? Now, if I was to bring that up like 15 times in this webinar, it would be a little weird, right? Me just returning to, oh, remember, we're on planet Earth. We're on planet Earth. It's like, yes, but like, let's get to the point here, right? You don't always have to, like, explore everything that is out there because. Because there's just. There's an infinite amount of context around us, and ChatGPT will assume it for you if you don't do it yourself. And that's why it's so important to assume it yourself. So that kind of rounds out my first Use case. And before I move on to the second one, which is going to be expanding on this notion of it writing in a certain style, we'll actually make it write in exactly your style. Okay, that's going to be the next one. Really exciting. Use case here. But before we do that, I just want to talk about how I actually use this. So the way I actually use this prompt is first of all in combination with the second one, which kind of defines your personal style. We'll talk about that in a second. But also, most importantly, I write drafts, okay? So if I'm trying to write an email that is supposed to communicate something for the Advantage brand, maybe, you know, we produce a new product or we have a new tutorial video, then usually I go ahead and I start with something very basic, right? I just say, rewrite the following email and then I insert my style. But maybe we could just start with this and. Or you know what, we'll keep the informal and we'll keep the formal and friendly style.

We'll just say friendly style. And what I do here is, this is. This is how I draft, right? And it's all. Let me tell you, after years of writing and content creation, it's so beautiful to be able to write a draft without anybody judging you over your shoulder, without anybody kind of criticizing what you came up with. You can just. You can literally transfer your thoughts into this textbook, how they flow. You don't need to respect punctuation. You don't need to respect capitalization. You don't need to respect heck, instead of good we wrote, right? Or what it was. You can make mistakes. It just needs to communicate the core ideas here so quickly. Just as an example, let's say I created a new video and I want to send out an email to my email list that basically says, here's this new video. It would probably benefit you. Go check it out. Right? So I'll just do that. Right? So let's just get started. So unfiltered. Just drafting, right? And this is how I actually use it. And this is how you should consider using it, too, because it's just so powerful at this. So rewrite the following email in a friendly style. Hi. I just created a new video on how to use ChatGPT to replicate your very own style. Check it out.

This is a really valuable one, right? Would this be an email that I would send out? No way, Right? I would put more care and consideration into it. I would think about my target audience. How can it benefit their life? Kind of underline that but instead of that, I can just do this and then I'll just hit enter. Right, Subject. Hey there. And I have a nice first draft. And this is really, like, this is how I use it. Okay, so you have this. And there you go. That's the first draft. Took like a few seconds to generate. This. Took like 10 seconds to write. Pretty efficient at this point. So what else? It might be a little too long for my liking. So what would I do if my new assistant would, like, walk in here and be like, hey, Mr. Pagani, like, I prepared your. Prepared the draft of your email you asked for. Here you go. I'm like, oh, thank you. That's really good work. But make it more concise, right? I like to be more concise in my emails. And she'd be like, yeah, yes, of course. No problem. I'll make it more concise. Right, well, same thing here. If I feel like this is supposed to be more concise, just tell it, make it more concise.

And guess what? We'll get an output that is more concise. There you go. Ah, that's too much. Little less concise.

There you go. Nice middle ground. And this is the way it works. So you can really go back and forth with it. And that's why I started, like, think about it as a human, think about it as an assistant. Because if an assistant comes with a draft that doesn't meet your expectations, no problem. Just, you know, let the assistant know. So that this is really your task here. It's just about. It's about communication. It's not about engineering, it's not about coding. It's not about going super in depth into some technical solution. No, it's about communicating your thoughts. And if you're a good communicator, chances are you'll have a very easy time with this. If you're not a very good communicator. Well, guess what? A lot of your lack of communication skills might stem from the social pressure that is present when you communicate with humans. Right. None of that is present here. It's just you and chatgpt and nobody else's opinion. Well, in this case, with the webinar, it's a little different. Right? A lot of opinions in this room. But that brings me to my next point. Every single one of you is going to write differently, right? They're not going to write in this style. Chances are most people in this room are not going to be writing the way ChatGPT wrote this. What do we do about that? How do we solve for that? Well, there's also a way. Now, is this the end all, be all solution on how to apply your style to AIs? Not entirely. You can certainly go even more advanced and even more complex. But when it comes to the quality of output you get for a certain, for a certain amount of work, this prompt that I'm about to share you, which is going to be use case number two here today, is really the winner here. Okay, so if I go down here, you're going to find this one. Analyze the text below. There you go. We're just going to open a new chat. Analyze the text below for style, voice and tone. Create a prompt to write a new paragraph in the same style, voice and tone. All right, so here we're asking it to create a prompt basically for anybody who's like very, very new to this. Prompts are just the things we input here. It's just the thing we ask of our helpful assistant. And see how I reframe that. Don't even refer to it as AI anymore. It's just my helpful assistant, right? So it makes everything easier. Okay, so we're just going to ask it this. And now as you can see, all of the materials and the prompts I always provide, I like to provide them in a so called formula form, which means there's a certain variable that you can exchange for your own, right? So in this case, insert your text. Guess what? Going to be inserting my very own text. So what I'll just briefly do here is I'll just head on over to the AI Advantage website. There's this little like text, this about paragraph and I'll just copy this, right? This is a quick little bio about me. It's similar to what Claude read at the beginning because guess what? I rephrased and then customized this bio on here so you can do that too. And what I'll do now is I'll just copy that text. And look, this might be any piece of writing you have ever done, right? So if you have a thesis you can feel free to copy from there and analyze the thesis like so if you have just an email that you know, you use as a template or like an email that you wrote yesterday to a colleague, you can copy that, head in here, paste it like so.

Like so. And then what's going to happen is it's going to go ahead and everything after this little sign, it's going to, it's going to go ahead and analyze, right? Because that's what we asked it to do. This is not, it's very like strictly separated. So you can always use.

You can always use a column to kind of separate your prompt from kind of the external information. So again, what's it going to do once I hit enter? It's going to analyze the text below this text for style, voice and tone. Create a prompt to write a new paragraph in the same style. Voice and tone. Okay, so this should give us a prompt that we can then reuse to replicate my very own style. Right.

There you go. Let's have a look at this. Let's have a look at this. So there it is. Style. The style of the text is informal and conversational. The author uses personal anecdotes, colloquially language, colloquial language, and emojis to engage with the reader. The writing is also motivational and enthusiastic, aiming to inspire and encourage the reader to embrace AI and its advantages.

That's very accurate, I gotta say. Like, that pretty much encapsulates the way I write. Yep, that's, that's spot on. Let's see, what else? Voice. The voice is first person and the author speaks directly to the reader, sharing their personal experiences and beliefs. The use of I, me and my establishes a strong sense of individuality and authenticity. The tone is positive, energetic and encouraging. I mean, look, just think about the presentation so far. Is this accurate? Is my tone positive, energetic and encouraging? I guess so. To a certain extent it is. So, you know, this did its work. The author is passionate about AI and its potential and they want to share their knowledge and experiences to help others succeed in the AI dominated world. The tone also conveys a sense of approachability, making the author seem friendly and eager to assist. I don't know you be the judge of this, but I feel like this is pretty accurate. This pretty accurately reflects the way I speak and the way I communicate. So now that it created that, you know, here it just does an example for me. So the prompt is write a new paragraph with the same style, voice and tone, and then it just writes something. But the trick here is really this. You just take this part, copy, right, and then you can save it. So we'll get to this in the end. But essentially this is like the way I organize my prompts and it's just, it's so beautiful because Notion allows you to like manage databases and it's free. So you know, I just save it here and then I can infinitely expand my personal little prompt database down here. Just a very useful way of managing this. But essentially now that I kind of, I encapsulated the style that is here For.

I have to say this, but for all the Pokemon fans out there, it's kind of, you know, when you're like running into a wild Pokemon and the trainer throws the ball at it. It's the same thing here, right? The Pokeball just caught a wild Igor. And now I'm going to be able to use that in any situation with the AI that I might desire. So let's take a step back and let's look at the last prompt we ran, right? Rewrite the following email in a friendly style. Even the one before that, right here, this very basic one. Rewrite the following email. Hello, I would like a donation. Good day, sir. So let's take this. Let's actually use this. So let's put this to work. What we learned so far, right? So first of all, I'm going to write this prompt that consists of instructions and then context, okay?

And then we're going to do a new line by hitting Command enter. And we're going to paste the second part that we produced, right? So all we did is we pasted in the style, the voice, and the tone, right? So rewrite the following email. Then it's the email, and then it's style, then it's the style tone. And as you can see, it just recognizes this formatting. My recommendation would always keep it clean. So if you always just do a new piece of information, just hit shift enter, put it into a new line, and if that's consistent, ChatGPT is just gonna have clarity on what your attention is. So in this case, it's gonna use my style, my voice, and my tone, and it's going to rewrite the following email. Let's have a look at how well it performs. And we're doing all of this on a free version, right? I haven't even talked about GPT4 yet. It's in one sentence. It does about 80% of the things you can do even better. And that's what you pay $20 a month for. But essentially, look, free version works perfectly fine. There we go. And look at this email. Okay, so already you're getting these emojis and. Okay, that might be too much and I might not want to use that. Actually, if you look at the article here or whatever, I only use them as bullet points. So, you know, just my intuition is telling me I don't want this, like, request for a donation to include emojis, right? So what do I do? I alter the prompt, right? I alter the question that I ask my assistant. So in here, I believe it was in the third paragraph, right? No, in the first one. And emojis to engage with the reader.

Yeah, let's just remove end emojis and let's submit this again. I don't even want to review that with that many emojis. There you go. We just got rid of all the emojis by removing the word and emojis. Okay, so let's have a look at this. This is very different from our previous results. Right, so we can compare this in a second, but let's have a look. I hope this email finds you in high spirits and ready to embrace the fantastic world of AI. Okay, so it's taking this like, AI angle because obviously that's kind of the bio. That's the style we copied here, right? It's like an AI enthusiast style. So of course he's always going to be relating the email that you're writing to that topic. But yeah, okay, so I'm reaching out because I've had some incredible experiences with AI and I can't help but share my enthusiasm with you. You know, a while back I was just like, you curious about AI. There you go. It goes into a story, right? Something that it didn't go before. And that's because it was specified in here that I like to share my experiences and stories. You know, if you review this in detail, you realize that everything that has been specified in the prompt above is reflected in this answer and you can now do the same thing, right? So if you previously took, you know, a month to write some, some thesis or, you know, some document you're very proud of, or you took multiple days to write some email and it really, you're really proud of that email. Well, guess what, you just go ahead. You use the simple prompt that I provided here. You analyze it for style, voice and tone, and then you take the outputs and then you can reapply that as a template to anything you might be writing moving forward, right? Whatever might be a grant proposal, but it might also be an email to your family, right? You can save a piece of you, a piece of the style you write in by using this prompt, and then you can reapply it isn't that powerful. So look, if I just quickly, if I just quickly contrast it to the informal and unfriendly style. There you go. That's output right there. And then this is the outputs in Igor style. Right? So one more thing, while we're at it and while we move, before we move to the third use case is, you know, let's briefly touch on GPT4. So you switch over by just pressing this button and now we're interacting with a different language model, a more advanced one. Right. So the main difference is GPT4 is it's better at reasoning, it's more creative and you're just going to find that it's better at assisting you with coding if that is your jam. But you're just going to find it's just better across the board. So if we send the same prompt.

Then you're see, let's just review it.

You see, all of a sudden it kind of went away from the formulation it had last time. It's way slower. See that? So that's a disadvantage. It costs money. It's also a disadvantage, but as you can see, it kind of takes a more nuanced approach to its output, especially when it comes to.

Creative tasks. I really prefer GPT4. So whenever I'm generating, you know, titles for blog posts or for emails or something like that, like always GPT4, whenever I'm really trying to produce the highest quality writing, I go towards GPT4 just because it's better at that. It doesn't always refrain to these, these phrases like you've seen before. Like if you paid close attention, you will have realized that the first emails have always been written with.

Just need to pull this up. I hope this email finds you in high spirits. I think it repeated that like five times out of the ten times we reran this prompt, right in GPT 4. It's going to be more creative because it has more options because it just, it's just deeper. So, okay, so it writes this very nicely and you'll again, you'll find that everything above has been reflected in it. This is just, this is, I would say it's just a more. It's if you, if you use it. Okay, so this is the ultimate question, right? When is GPT4 worth it? I By now I've been asked this like a thousand times and my go to answer at this point is if you use it for your work more than twice a week, it's probably worth it. If you just use it like once a week or sometimes just don't use it at all for a week, it's probably not worth it. But if you use it regularly, meaning, you know, two, three times a week, then it's probably going to be worth it for you. If you use it like 10 times a week, just, you know, stop procrastinating and just, just get to it. Bite the bullet, pay the $20 and you're gonna. You're gonna find that this unlocks a whole new universe for. You know, there's other reasons too. Like there's the plugins thing here, which we're not gonna get into, but essentially there's like a whole app store that allows you to kind of. For example, this one allows you to paste a YouTube link, and then it takes the video transcript of the entire YouTube video, and then you can talk to the YouTube video. Right? So you could take like a podcast and summarize it and then rewrite and then use the rewrite prompt and you could rewrite the podcast into an email. Right? You could do things like this. But that's. Again, that's, that's. I just wanted to touch on that. That's possible. Back to GPT 3.5. Okay, so what did we learn? Quick recap before we get into the last one here, because we learned a lot. We learned that think of it as your assistant, not as some foreign AI number one. Secondly, every prompt consists of instructions and context. Okay. How much context you want to provide, it really depends on your use case. But no context is the wrong answer. And very little context most often is also not the right answer. Okay, I'm getting some drawings on the screen here.

Then we looked into use cases, right? We looked into use case number one, which was right here, rewriting anything. Then use case number two was extracting your very personal style. And one last tip that I want to mention on that is you want to extract the style of a type of writing that is relevant to the thing you're writing. Meaning if you're writing an email, don't use your bachelor of thesis style. If you're writing, you know, a draft of a thesis, maybe just the first outline or something, or like you're using it for a chapter. Don't use your email. You kind of want to match it as closely as possible. Then it's going to perform the best, as you saw, like this, my profile was an AI enthusiast. All of a sudden, that grant writing email is going to be more geared towards the writing of AI enthusiast, and it's going to infuse that. Okay, so those are two use cases, but here's the third one. And this is one that is really underappreciated. And even the people that stuck with us that might have well thought like, oh, I've seen that and I've. And I know that, or I saw Igor's videos. This is no new, no news to me. Well, I hope to catch you off guard with this one. Because this one really, and we use GPT4 for this just because of its, of its more advanced reasoning capabilities and just because it will show the full potential of this prompt. But this is so powerful. And let me tell you, like a lot of life is about communicating, right? It's about communicating what you think. It's about communicating your best ideas to others and through that adding value to people. But the problem is people are usually pretty set in their ways, right? People are usually not very open to new, to new ideas or new technologies and everything. And the best way to get around that is to come in a little prepared, right? You come in and you think in advance. Like anybody who has been engaged in any sort of debate will know the best way to come out on top, if that is your goal is to think in advance about the things that your counterparty about the points that your counterparty will bring up. Well, what if I told you there is a way to do this with ChatGPT and it's extremely simple and it will not just give you one counter argument or one answer. No, it will give you multiple. So have a look at this prompt. Can you write an argument for. And then, you know, take the, the issue that you care about the most. But we're just going to use universal basic income here. So can you write an argument for universal basic income from multiple diverse perspectives? And do you see how carefully these prompts are crafted? Like there's no superfluous words in here, there's no fluff. It's extremely precisely to the point. Right? So we're saying from multiple diverse perspectives. Because if you just say diverse perspectives, it might take two. Right? But multiple implies do more than that. But it's also, this word is also open enough to not do. Like If I said 20 diverse perspectives, that's too much, right? Nobody wants to read that. Multiple does this just. Right. Okay, then we want diverse perspectives, of course, because we want different viewpoints. And then before you do so, state the characteristics of the various characters. So what this will do is I'll just show you, I'll just show you. Let's run this.

And it starts ideating various characters with various beliefs and it's going to start arguing for universal basic income. Okay, so let's review this briefly and then I'll show you the ultimate trick, the follow up prompt that we'll use after this. Okay, so there go. Sure. I can create four characters with diverse perspectives to discuss the topic of universal basic income. First of all, you have the economist, then you have the small Business owner, social worker, technologist. Now if you don't want those four, this is just crafted to be very convenient. If you maybe want something way more specific, just include it in the prompt. I could just say four various characters, including. And then you just name your character, right? But we'll just go with this. And then you have like descriptions here. So the economist is a progressive economist who specializes in labor markets and income inequality. They believe in the power of economic policy to address social issues. And then you kind of just get these profiles so you know what's happening. And then down here you get their arguments. And I mean, tell me, is there a lower effort way to prepare for a heated debate, whatever you might be doing, you know, negotiating with your spouse, negotiating your salary. Like life is just a series of negotiations, right? Everybody has their own interest. And then people come together to kind of like help each other. And hopefully at the end of that interaction, it's a net positive but essential part of that whole process of how the world comes to be is the negotiation. And this just helps you out so much because it shows you diverse perspectives. It shows you the arguments for Universal Basic income right here. So let's have a look. What is the argument of a technologist for Universal basic income? The future of work is changing rapidly due to technological advancements. Automation and AI are replacing jobs faster than we can create them. UBI could be a solution to this impending crisis. Providing income security as we navigate this transition could also spur innovation as people, as people freed from survival concerns might pursue entrepreneurial or creative endeavors. So a very optimistic point of view here makes sense, right? This is the point of view that a technologist would take. But luckily you also have a small business owner which talks about how this incentivizes work and how there's going to be labor shortages. Everything that, like, touches a small business owner. Right? So first of all, this informs you, and this is all nice and well, but here's really the unlock. Okay, so if you look at this, this little notion template is structured in a way where there's also a second column called follow up and you can fill in whatever you like. But right here, I kind of preset it. And we're just going to use this follow up prompt. So again, if you told your assistant, craft some characters and bring me their arguments. Yes, sir. Here you go. Here are the arguments. And then you will be like, all right, now, next task. What are three highly effective counter arguments that I could present to each character.

And look at it, doing all the hard work for you.

And this is really a way to just become a more effective human being. Whatever you might be pursuing, you can create a variety of characters, create the arguments that they would be using, and then with this follow up prompt, you can look at ways to disarm them.

Right?

And this just makes you more well versed because what you'll find is that a lot of like negotiations or.

A lot of just life in general, the person that is more of a chess player and that thinks ahead and that tries to like think more than one step ahead is more often than not going to come out on top in those situations. And with this prompt, you can use even the free version of ChatGPT to prepare yourself in a way that frankly hasn't been possible with such ease. Right. This would have took hours without AI. Right? You need kind of a group of intelligent people to kind of like present their viewpoints. And right there, not one, but three counter arguments. So let's look at the small business owner's counterargument. So he was talking about how UBI would create labor shortages.

And there you go, some counter arguments to that. While higher taxes might be a concern, UBI could also stimulate the economy by increasing consumer spending, which could benefit businesses. Okay, so that was his first point probably that we skipped was taxes. Right here. Higher taxes to fund ubi. Okay, so there you go. You have a counterargument to taxes. UBI could potentially lead to inflation. If everyone has more money to spend, prices might rise, negating the benefits of increased income. All right, UBI might disincentivize work. So that's argument. If people can meet their basic needs without employment, some might choose not to work, which could impact productivity. Okay, so what, what if I want to go even deeper? What if these are too surface level? Well.

Expand on the second counter argument of the economist.

You can just do the same thing that you would do with your assistant again, right? You can just say, okay, I really like the second one. Please expand on it. And now that I'm reading it, let's have a look. Let's have a look. What does. So this is their counter argument.

And there you go. It went from one sentence to.

We lost the sound.

Yeah, I guess we would lost your microphone.

How about now?

No, it's okay.

Okay, okay, okay. So just with this, you're gonna be able to get ahead of so many situations. And you know.

Really, if you think about it for a second, like this just unlocks a thousand more opportunities. Because I just showed you free concrete use cases. Right, but the way we kind of played with the context that goes into it. The way we kind of expanded the conversations by talking to our assistant and not just taking the first results at face value, but really pushing back right the way we did that. Well, it's gonna unlock so much more for you. And what concretely. Well, you can go ahead and I'll share this. I'll share this in the chat. Now you guys can just get this template for free. I give out a singular, a similar version to everybody that like signs up to my weekly newsletter. But there you go. This one is a little more expansive. We have a few prompts here at the bottom. And.

Really what you're going to be able to do here is you're going to be. And can you guys see the Notion page right here? That visible?

Well, yes, yes.

Perfect, perfect. So really up here you see a bunch of different problems. So if you're writing something, you can just open this up and you could be like, okay, you can write songs, you can explore email subjects, or you could just write anything. Right. And the way this works is very briefly is here you have a bunch of examples down here. So this is my favorite example. And then there's a bunch of other examples. And above you have the formula. And the formula, you really edit this by replacing everything in brackets. So this helps you just make this prompt your very own. But you could totally just go ahead and copy this and, you know, send it in a new chat and you're good, and then it's going to write this exact type of text and you can do this. There's a total of two hundred and fifty examples here. Right here. You have explored different arguments. There's other examples of that. But essentially you can use it to improve your marketing. You can use it to improve your very own health. So for the ones of you that are familiar with ChatGPT and been on social media at any point over the six months, you'll probably have a lot of prompts already. This is just like my attempt of, of kind of putting it into a form that is infinitely expandable because Notion works with databases and I can always just create a new line. And it's not, it's never going to overflow. Right. I can just infinitely expand this. I can create my own prompts, I can save them down here. And then the beautiful thing here is you can always, like, if you click these, you can always hit this little favorite checkbox. And if I do that, it's going to show up in my favorites. Which are, which are up here. Yep. Exactly. So these are all the favorites, these I preset up for this particular audience. So if you get this notion template, these are the ones I would recommend to start off. So if free use cases are not enough, then you know, here's a bunch more that I would recommend. And here's even more. You have a bunch of styles, you have many follow up prompts here that just work, right? So for example, you could always just say, can you clarify what you mean by and then a certain concept and it will go deeper on that. Or you could just tell it this. Could you elaborate on how the latest response provided does align with my previous instructions? After clarifying that, please make another attempt at following the instructions I gave earlier. Okay, so all of this is for you to explore. We went in on two theoretical concepts of how to work with this. We went in on three concrete use cases that you can start using today. Okay, they're right here. You can just copy paste them. There's no reason not to do it. There's no reason not, not to play with it. You can start rewriting your own text, your emails, your grant proposals, whatever it might be. You can start extracting your custom style from your writing and then reusing that in other prompts and you can explore different perspectives. So what's next? Okay, what's next is this like right here is a Persona. This is a social entrepreneur. And this is really like we just went from like the most beginner friendly stuff to kind of the most advanced topic that you could be looking at. But nevertheless, I wanted to show you this very briefly. If you open up a new chat and you start the conversation with a so called Persona like this, and this is directly extracted from a new feature by the way, that is accessible to people outside of Europe where you can preset ChatGPT to have all these characteristics. So what's your profession? What's your current projects, challenges, What I would encourage you to do if you are a little more advanced and you know, if you knew, just ignore this part. But you could take all of this and you could customize it to your very own needs. So you just fill this out for yourself. What is your profession? What are your projects and challenges? What are your specific interests? It's going to take a while, but then you're going to have a personalized profile that you can just copy and send us the first message inside of ChatGPT.

Okay, it's going to come back with a bunch of text here.

We could, yeah, we could extend this in the end to make, to make it not do this. But anyway, now it's set up and now if I say, you know, write me an email to my boss, all of a sudden the answer is going to be the answer of a social entrepreneur with all of these concrete characteristics. And look at that. I want to provide you with an update on our plastic recycling project and discuss some potential directions for its future. All of this information, all of this context it got from this Persona where I specifically defined it. So this is like something very advanced, something I would not worry if you have, you know, less than, I don't know, 10, 20, 50 hours in ChatGPT. But this is just kind of the deep end of the pool. You can get this specific and yeah, you just have this social entrepreneur just to copy paste right away. And with that being said, I think that's enough for the theoretical and practical concepts. I would love to get into some questions here. And one last thing before we do that is what can you do beyond ChatGPT? Right? Cause this is kind of like this might be the core of everything like how do you talk to your helpful assistant? But.

It certainly goes further. So one thing you could look into is what we just talked about, it's the Persona modeling, right? Another thing you could explore is.

Automations. Right? So you could automate all of this and you could say when a new email shows up in my Gmail account, then please summarize the email and then write a response in my style, here's my style and save that result to my drafts and it just does it. Every email that will come in there will be a draft ready to go as a response for you. Right? So you can build automations like that, right? That's a separate topic but I want you to know that it exists.

And one last one, also very advanced is so called multi shot prompting. But that's essentially when you tell it, you know, write me an email and here's free example emails that I really enjoy. Please mimic these and then you provide the free emails and then it's going to have the information from within those three emails to get you a better answer. So those are kind of the things that I wanted to point out. Of course the AI space is much wider and we could keep talking for hours here. And there's AI art and there's AI video and there's AI voice generators. But at the end of the day it always comes down to being able to communicate with the system effectively. And I hope that this little presentation helped you do that. I hope you could take away some tips and tricks. And with that being said, let's have a look at the chat.

Thank you, Igor. To give you time to brief, one minute. You want to say thank you. Thank you for this very passionate presentation. And for many of us there are many new things we are learning give us the motivation to go and search more. So thank you. For me, there's one important question that came out of this. I want to put it for you to start with answering that question. I think that true, there are some people to say, where do we find the link? Where do we download this app? What is the basic, where do we start? There are many people asking that. The second important question I saw on the charts is that based on the fake news, this misinformation, we know that the AI go and search all over. What is the risk of getting something that is not real? Those information that are fake, news that are hate, that's all those things part of the content. How do we make sure that we don't run the risk of saying something that is not true?

Hmm. That is, that is a good point. It's, it's a bit of a, it's a bit of a separate, separate conversation. But I think concretely with ChatGPT, the biggest risk is it's coming up with information that is just not true. Right. It hallucinating. So the biggest workaround for that is just being very specific with, with your input. Right. If you don't leave it room for hallucination, it will not hallucinate. But if you just tell it, you know, if you give it a one liner and you ask it to write a bachelor thesis, well, it's going to tempt it, but it's just going to make a lot of things up along the way. So I think that is step number one. Step number two is we're simply not at a point where you can blindly trust this and where you can like certainly anybody who wants to like build. I think the, one of the worst things you could do today is try and like build your academic career upon this. You just got to get debunked so fast. It's, it's just like, there's just no way to do it. Take it as a starting point. Take it as an assistant, right? Like those helpful assistant, those two words encapsulate everything you need to know. Know. Because you wouldn't hire a new assistant and just take the drafts she, she just provided to you and just be like, okay, I'm done. Right? No, you would, you would go in, you would edit, you would, you would use some critical thinking when dealing with that. So unfortunately there's no automated way of, of kind of getting past that obstacle right now. Just have to use some good old thinking while editing and always, you know, double check the sources. With that being said, in the Persona that I provided, there's a line that says I don't, I don't know it all by heart, but there's a line that kind of tells it to always cite references if they're available. And there it will at least attempt to tell you where it took that information from. Ah, that's from this blog post or that's from this type of research. It's not bulletproof. A lot of times it's going to just make a certain type of research up because again, you asked it to site research. So it will do its very best to do that. Doesn't mean that it's going to be very good at doing that. But there's certainly things where you can push it into the right direction. But at the end of the day, you know, trust but verify.

Where do you get the link or how do you download the app? How do we know we are in 3.5 or 4.0?

Yeah, okay. Okay. So you simply go to chat.OpenAI.com so I can post that in the description too. But basically all you need to do is create an account which takes about. You know, I actually have a great two minute tutorial in YouTube that shows you how to create an account and run your first prompt. So honestly that's the best I can do.

I'll post a link in the, in the chat. Oh, I sent it as a direct message. That's.

Yep, there we go. So you can just go there, you can create an account. If you need assistance, just go to YouTube and look for how to use Igor.

There is a question that I think I know the answer, but can it extract from recent events ChatGPT? As far as I know, it's until September 2021, isn't it?

Good question. So yes, the base model is cut off in September 2021. There's ways to work around it. Right. If you just copy the article and you paste it into the conversation. Well, you just inserted the context that is necessary so that's one way to work with it. A second one is with plugins, so you might be familiar. Like Bard has a competitor, that one is directly plugged into the Internet so it can like use search and use that data. What you can do with plugins is use a plugin with ChatGPT plus that's the paid plan. There's a plugin called Link Reader. There's another one called Web Pilot. If you use those two plugins, you can just insert the URL and it will read the entire page and include that as the context to what you're asking for. So it's possible, but just the default free version. You're going to have to copy the external data that you're trying to interact with.

Before I hand over to Claude to maneuver other questions, I want to make sure that the video of this webinar, the link will be sent to all the participants, including the page that Igor was showing with all the prompt. So because you have your email, you will receive in your email the link to the video and that list of things that is part of you being registered. So you have this to play with for the coming days. Khloe, now you can raise those key questions that are.

I don't see many, many questions. People are more asking for reference or these kind of.

Reference as in where to find more or.

No, but we will, after the. We have been out with the recording, we can send also the, the reference of the, the sheet stink. Yeah, the, the. How do we call it? The, the notion template. The notion, I call it whether you have all the prompt.

I call it the Chat GPT advantage.

But yeah, certainly, certainly. Yeah. I mean, I think this is a good point. Like how do you, how do you go further than this? How do you educate yourself? Well, certainly, you know, this is. I made this my full time calling to educate people and to help you understand this. So we have our own materials, as you saw in chat. You can just join the free newsletter and then once a week you're going to get an update. And also the notion template, we have some products, but generally speaking, I think the best way to educate yourself is first of all YouTube and secondly Twitter, those are really the two best sources. And on YouTube you can certainly check out AI Advantage. But another recommendation I would make is Matt Wolf. He's the biggest AI news guy. So he always posts one video a week. It's like a 20 minute video where he summarizes all the different happenings in the AI space. So you can watch that to get all the latest news. There's a bunch of newsletters. So we have a very editorialized one where it's just like one app, one prompt and one news story that matters. But there's other ones that come out daily. Two notable ones would be Ben's Bytes. That's one. Ben's Bytes. And another one is.

Rundown. So Rundown is a daily newsletter where it gives you kind of all the AI headlines. This is all free, by the way, right? So the YouTube channels are free. This is free. Beyond that, I think being on Twitter is kind of the best way to stay up to date with these things. That's really kind of the main source of information for me day to day. Like if you spend an hour a day on Twitter and you follow the accounts that I follow, AI Advantage on Twitter, then you'll kind of get the same news flow that I do. And yeah, I think that should actually set you up for at least a few weeks. And then if you want to go even further than, as I said, the topics you want to explore are prompt engineering, automation and yeah, I think those are the two main ones that will kind of unlock a lot for you. And prompt engineering really is what we did here. It's, it's finding ways to talk to the model that get you good results. And maybe this, like this is before. Let's, let's, let's wrap this up. But.

No, sorry, I think there are.

Okay, okay. One really important thing, right. Maybe I should have said this right in the beginning. Like, what makes a good output? What makes a quality output? I think this is the people always like, they overthink it. At the end of the day, a quality output is something that is useful to you if you consider it useful. If it writes that email draft that would have took you 10 minutes, that's a good output, right? So at the end of the day, just find these little wins for yourself and you'll find that over time they will stack on top of each other and you'll be like, oh, write a draft now rewrite it in my style. Okay, now write me tweets from that. Okay, now write me a blog post from that. And all of a sudden you'll have like drafts for all these things that would have took you a day. So yeah, quality output is something that is useful to you.

Sorry, if, if I can add something personal, I will say that there is also a course or it's a paid course of Igor on with a advantage. It's a eight hours course. Where will you learn many, many things? Because it was only a minimum of all this knowledge that we share today. It can be worse. I mean to, to do it for people who really want to go in depth. But first go to his YouTube channel. There are free video, tons of free video. And it's really. I went through many, many people presenting ChatGPT and Igor is really, I, I have, I have no business partnership with him, but he's really one of the best one I saw on Internet. And it's very didactical and there are many things that was not put in consideration today. But it's really, if you go through the notion template, you will already go very far and I all encourage to test it. And one of the motto is try, try, try and try and get rid of your false assumption. On ChatGPT, you will discover what yourself, what is not very good, what is bad, but also what is helpful. Thank you very much.

Thank you.

Thank you. Thank you very much, Claude. And yeah, it's wonderful to have you as I think that's kind of the first point of contact you had. You signed up for the course, you're a student of the course and then you reached out like, we need to bring this knowledge to more people. How can we do this? That's how we arrived at today's event. Thank you so much for the invite.

Igor, there is a question before we close. There was a question. Someone said, if I'm writing in Portuguese, am I writing in French or Mandarin? Do the ChatGPT identify the language and answer or do we need to set in the language at the beginning?

Short answer is yes, it identifies it. It knows all those languages. I'll, I'll say this though, it's, it's really good at translating from, let's say Portuguese to English. It's above average at translating from English to Portuguese. So again, because it's, it's like deepest knowledge is in English. Like if you're going to be going from another language to English, it doesn't get better if you're going to be going from English to another language.

Average translator, maybe, but not excellent. But yes, it can do it.

Thank you. I like to say we won't stop here.

We want to all acquire these helpful assistance to see how you can help us in a different field. So the way forward will depend on your feedback, on your reaction and your request to say, how do we take it further? But we want to leave the change the world is going through. We want to be able to go through those change. And next time we might come with another angle of this. So please stay tuned. We need to make sure that we don't bother you all the time with the same thing. But our reaction, our offer to you will depend on your question and your requests. So once again I want to thank Igor for his time for the knowledge and for the passion. Thank you, Claudalaine. And thank you for identifying Igor and for making this happen today. And thanks to the technical team behind this And I have my president on the line and with all the members of my executive committee. This is how we are playing with our time at the World ymca Play with chatgpt. So my next report to you will be written by ChatGPT.

And I appreciate all the positive feedback in the chat. It really warms my heart. I, I hope, you know, most of you will walk away and actually try some of the things we talk talked about. I try to make it as practical as possible. I really think there's something for everybody in there and just give it a shot. You know, it, it's, it's free. Try it out and yeah.

Thank you. Thank you so much for joining. Thank you.

Bye.

Bye.

Thank you everybody.

Bye.

Bye. See you.

Thank you. Bye.