Creative Prompting & Communication with A.I. Explained | Get Your A.I. Advantage with Igor Pogany

Prompting is often like packaged in this kind of like mysterious way. And people say like prompt engineering. And then it gets certain people who are not that deep into it confused and they're like, sounds like a technical skill. At the end of the day, it's just communication. It's just good communication. The same things, same rules that apply to good communication apply to good prompting.

When I've taught marketing and copywriting, I always basically start out by saying, we're studying thinking, right? We're studying thinking and we're studying communication. But we're going to do it by practicing marketing and copywriting. And then now with AI, when I teach this stuff in prompting, I say the same thing because it really is about clarifying your thinking. If you can get your thinking clear, then the solution to the prompt is very, it's easy, it's right in front of you.

And the whole like, dystopian take that many people have, I think it's, you know, there's good reason to have that discussion. But I think in the near term we're in for a lot of, you know, quality of life upgrades with how we can interact with our phones and how we can use the health data from our Apple watches to like actively improve our life. And I'm excited about that. I think that's gonna make a lot of lives easy.

Today I'm speaking with Igor Pagani. He's in Lisbon. His passions are technology and AI and everything that's new and cool. He's the founder at AI Advantage. He's an AI educator, He's a tech enthusiast, video producer, self taught coder, marketing expert. Over 10 million views on his YouTube channel. Igor teaches people how to do things like build their own chatbots in 15 minutes and build custom GPTs. He has a multilingual background, speaks English, German, Slovakian, and he loves kite surfing. He said he has an irrational passion for kite surfing. So this is going to totally be an interesting interview. Welcome, Igor.

Hey, thank you so much. Good to be here. I'm looking forward to chatting about all these topics and more.

All right, so what's your, I mean, let's just dive right in. What's your favorite new way that you're using generative AI in your life or your work right now?

Yeah, that's a good question. I mean, generally speaking, I always try to experiment, you know, so if there's a new way, then a week from here on out, that's kind of, you know, the old news and I'm looking for a new one. So what I'm Experimenting with right now is actually GPT mentions and actually like changing up the instructions of the GPT so they can talk to each other more. And I really think that is kind of the direction OpenAI is taking their product into. So that's what excites me right now. That's what I'm experimenting with. And yeah, I mean, what an exciting time to be alive. There's always something new and this space is ripe with surprises.

So never bored AI's talking to each other. We're going to have to talk about that a little bit as we dive in here. So how did you become interested in generative AI? And I mean, what happened or what did you learn or what did you see that got you really interested, that grabbed your attention and had you say, whoa, I need to pay attention to this.

For me it was really the ChatGPT moment.

But not the moment in January, February 2023 where it went giga viral. But I actually discovered it one day after launch the product that was online just because I was involved in communities that followed this type of stuff independent of AI. I always kind of was interested in hardware, software, technology and always interested in kind of the newest technical papers and newest releases. And I always love to be early adopter. That's just kind of at the essence of me and always has been. So, you know, when I discovered the tool and spent a 12 hour day playing with it, I was just like, this is beyond anything that I've experienced so far. I've played with some other tools, you know, AI tools before that, but it never like really clicked like that. So got me enthusiastic. And then I woke up the next day and spent another 10 hours just playing with it. And all of a sudden after two weeks like that, I just like had some things where I figured that it would be helpful. So, you know, being also a video creator at heart, always been that too. It just made sense to kind of record some content and show people what I found and apparently they really liked that. So I just kept doing that. And by the time the kind of, you know, the mainstream attention, so to say, came around in like January, February, I had a library of videos up already. So yeah, was well positioned to kind of teach people more. And at the end of the day, you know, I don't consider myself an expert. I consider myself a student and a power user. And with some of these things it's simple as that. If you spend five times more time than anybody else in it going to be better. You're going to have, have more skills and you're going to have more things to share. And that's what I do. I just spend a lot of time with this tech and then I share what I find.

Yeah, it's kind of. It's really powerful. Once you realize that this is actually a digital mind, it's not a search engine or something. Once you kind of have that shift, this is a new kind of thing.

When you're teaching people about this or when you're talking about it, how do you think about the. Let's call it the AI, you know, or the digital mind. How do you hold it in your mind and how do you explain it? Right. Because I think it's really important that we get how to think about it.

Yeah, that's a very good question. And it's actually one of the first things I talk about when I kind of like open up this topic with a person that is looking to learn about it.

I have a few frameworks that I go with, and this is the very first one you asked, the first question question, the correct question to ask when people are beginning and starting out. I think about it as an assistant, and in other cases I think about it as an employee. Those are really the two frameworks that help me. So both work employee is maybe more like action oriented. That's where we get into more the autonomous agent territory for some of the advanced folks. But assistant really is the one framework that works and helps people get over the first hurdle of how do I interact with this? Because if you interact with Google, you don't talk to it like you would to an assistant. You give it words like capital, France, population. Right. That's a different way of communicating than saying what is the capital of France? And if you talk to a human, if you talk to assistants that you hired and that you were paying for and that you would want to do the best job possible. You just talk differently. And it also allows you to tap into your social skills, which is huge and beautiful because all of us have them. Right? And I love it. Every time I said that, the introverts in the audience are like, well, I don't. But just by the fact that you understand the words coming out of my mouth, you have some social skills, right? You have language. It's just hardwired in us. So if you're able to activate those circuits and if you're able to talk to, you know, the AI or the large language model or the chatbot, whatever you want to call it, you're able to talk to it as it would to an Assistant. It's going to unlock so much intuitively because you're going to start having conversations and you're going to start asking follow up questions rather than, you know, just throwing up your hands in the air. Like this is not able to do it. Right. If you have a new assistant and you give her a task, him or her a task, then she might not be able to do it right away. She might need some context, just like the AI needs some context. So that is my tip number one. Think of it whenever you get stuck, think of it as an assistant and you're going to get unstuck and go places.

Yes. Yeah, like a super genius assistant that knows everything, that occasionally makes mistakes and sometimes gives you like surrealist answers or something, right?

Yeah, exactly. And that's the weird part about it, right, because it's like a super educated assistant that kind of like read all books in the world, but then at the end of the day might not be able to complete some basic things without like exact guidance. So it's a little weird like that, but also very powerful. I mean, who wouldn't want an assistant that has read all of the Internet?

Hey, quick question for you. What are you learning from this interview and what do you want to learn next? Well, you probably want to stay ahead of the curve with AI, just like we want to. And we want to be your top resource for interesting cutting edge AI tools and knowledge. Here on MetaMind, we're finding and interviewing interesting experts from different worlds, different industries. And then we're using this knowledge so we can keep up with AI and improve our success, improve our quality of life. So stay with us on this journey. So first click on the subscribe button which you'll find under the video so that YouTube puts this into your feed when we release new interviews with AI experts. And if you're enjoying this interview, please give me a like to let others know that this is the real deal. And also, if you know someone who's interested in what we're talking about here, then send this interview to them so that they can learn with you. And I also want to hear from you. So go down to the comments below and let me know. What's your favorite prompt or prompting technique? What are your favorite AI resources or tools? What are you learning from this conversation? Do you disagree with anything we're saying here? And also, what else would you like to learn? Who would you like me to interview here on metamind? All right, well, go and subscribe. Leave me a comment below. And back to our interview. Yeah, I'VE been into, you know, ray Kurzweil for 20 years or something. And it's interesting watching his, you know, prescient, visionary predictions, all kind of coming true over time. And so he talks about, you know, when AI is going to pass the Turing Test. And he's been talking about it like it's very matter of fact, like, you know, in the, whatever, late 2000s or something, 20, 29, it's going to pass the Turing Test. And now he's saying something that's totally fascinating, which is that AI, you can't really get it to pass the Turing Test right now because it knows way too much. So you'd have to ask it to dumb itself way down because no human knows as much as it right now. And so we run into all these fascinating paradoxes where when you get something that knows like a thousand times as much as the next most knowledgeable human, you, you know, but it'll do anything that you ask it to do. This is not a. It's not a normal type of relationship to have. We're not used to relating to that kind of vast intelligence and that much power that'll just do whatever you want.

Yeah, it's true. And it's funny, if that came in a truly human form, if a person had a mind like that, a lot of people would get intimidated by that. Right? So it is just a new paradigm and it takes some getting used to. And that's why I'm such a big of a. Such a big advocate of people learning it as soon as possible, because this isn't going anywhere. It is the next evolution of software and these skills, they stack, right? So you're not going to be, you know, even a year ago, I was telling people, like, hey, no worries. Like, just catch up whenever you feel comfortable. But at the pace that the space is evolving, like, we're getting into these topics and what I outlined at the beginning of the interview, like what I'm experimenting with right now, you kind of need a few prerequisites, requisites to even have that conversation properly. If you haven't internalized how to have a basic conversation with a large language model, you're not going to be able to wrap your head around the fact that there might be four of them and they might act independently and there might be a manager that assigns tasks and everything that is too far out. You kind of need the base version of ChatGPT that came out end of November 2022. You need to understand that interface first before you get into you know, kind of the PhD level of using it. Right. So yeah, I always recommend people get started and yeah, you can tell yourself whatever stories you want, but this isn't going anywhere.

So yeah, let's talk about prompting. So first of all, how do you think about prompting and do you have approaches or mindsets that you use when you're actually prompting and maybe you could even talk about particular situations or examples?

Yeah, so generally speaking, I think there's a lot to teach there and there's a lot to explore. But at the very base level, even when I use it on a day to day basis, I find myself just talking to it like a human. And there's this very, very basic question and thing that I needed to clarify with myself early on. And that is because you see all these kind of like fancy prompts and approaches and Personas and context and priming it with information and feeding it five examples or databases in advance, whatever. But at the end of the day, only thing that matters is that you get a result that you're happy with. Right? So that is one of the first things that, that one needs to realize what makes a good prompt, right? That's kind of a big question. And the answer is very simple. It's when you get a result that you're satisfied with your prompt is good enough. So often a single sentence might just do it. Right. But when that doesn't yield the results that it could be, there's that component of. Yeah, you also need to be aware of what it could do. Right. That comes with education. But, but at the very base level, when I try to get something very simple done, just ask a very simple question like I would an assistant. And then beyond that there's, you know, I have certain frameworks and ways of kind of fleshing it out and adding the context it needs and specific techniques to get more advanced results can help in certain situations. But in my day to day, whenever I encounter a problem, like for example, doing my counting, and I ran into a little problem with a calendar with different conversion rates in it being transferred to Excel file. And I didn't just want to type it, I just told it like can you turn this into Excel file? And it just did it, right. Like there's no need for the super advanced techniques if the simple one does it.

That's the first thing I would say. And honestly that for most people that unlocks like 60 to 70% of the use cases already just talking it. And at the end of the day I'll just add this one more point. Very important point is that prompting is often packaged in this kind of mysterious way and people say prompt engineering. And then it gets certain people who are not that deep into it confused and they're like, sounds like a technical skill. At the end of the day, it's just communication. It's just good communication. The same things, same rules that apply to good communication apply to good prompting. Be concise, avoid ambiguity at all costs. Right.

These principles will go a long way if your sentences are convoluted, and it's not, not even a human would clearly understand what you're trying to communicate. There large language model is going to have a problem with that too, because all it did was get trained on things that humans said and human communication. So if you're good at communication, chances are you're quite good at this. And that's why a lot of people who have natural communication skills really excel at prompting too.

Yeah, that's a really nice definition. Right. A good prompt is one that gives you a good result. It really gets you back to the basics. Yeah, well said and agreed about the communication. When I've taught marketing and copywriting, I always basically start out by saying, we're studying thinking, we're studying thinking, and we're studying communication, but we're going to do it by practicing marketing and copywriting. And then now with AI, when I teach this stuff in prompting, I say the same thing because it really is about clarifying your thinking. If you can get your thinking clear, then the solution to the prompt is very. It's easy, it's right in front of you, you understand.

Exactly. A funny example there is. Philosophers are extremely good prompt engineers. From what I've seen, people who dabble in philosophy, and it's so confusing to most people, why would they be good? Because just what you said, their thinking is clear, they understand whatever that worldview might be, but they understand their own worldview and how it's built and the axioms that it's built upon. And then they're able to call upon those right in their communication, which makes it so much simpler if you're, if you're able to really boil it down and just communicate the essence of the thing or, or communicate at different layers. That's what, that's what makes it really good. And I've seen some incredibly gifted prompt engineers that had a philosophical background.

So, you know, that's a good point. I hadn't thought of that, but I agree with you. I, I interviewed someone who had a philosophy background. He was great at his promptings, really Clear. And I'll just add to this too. If you've always thought about learning philosophy, but you were like, oh, that's just too weird and too abstract. Try using ChatGPT to teach you philosophy because it can start with you right where you're at. It can explain philosophy, you know, in a word, in a sentence, in a paragraph. It can explain it for your situation. You can really pick it up and learn such an incredible teaching tool. How do, how do you use it for learning?

For learning? Good question. So I really, I'm a visual learner and I love to learn through long form audio and video. That is kind of my preferred form. So, you know, courses, podcasts, interviews. That is really my audio books. That is what I kind of default to. And it's a bit of a godsend in that, in that regard, because you can just like feed it a link or a transcript and then like talk to it and follow up as if you had the author in the room. So I really like doing that. If I listen to a long form podcast, I love to pull a transcript, feed that into ChatGPT and have a little conversation around the topics that might have not been clear to me or the ones where I had some ideas and I wanted to maybe contrast it with, you know, the knowledge of the wider Internet. What a, what a luxury to have that, that we're able to do that. So that's what I like about that.

Give me an example. Give me a couple of examples. When have you done this recently? And then what did you, you know, how did you do it? What did you ask it?

Yeah, so one example is I even did it in my own content. I love to pull video transcripts and we have a little internal workflow where all the videos we create get automatically transcripted. And then you're able to do things like feed it your 10 latest videos and then talk about your next one and see how that fits in, or if you covered some of those topics or if any of the topics from the latest videos. If it might be worth resurfacing a topic in there in the newest one to enhance it. So things like this, you can get really creative with it. But essentially pulling transcripts of your own videos or pulling transcripts of long form podcasts, you can pull the whole thing and you can just summarize it, which gives you kind of a deeper look into it. And you don't have to spend two to three hours if it's not right for you. I don't think it's a replacement for spending Two hours on watching the entire podcast at all. But I think it's a nice little peek into what content is covered in there. And it allows you to make a know more educated decision if that is, is right for you and if you want to spend those free hours. So those are, those are two that I like to do. I like to prompt on top of like my own videos, and I like to summarize podcasts and talk to them in advance. That's something I do regularly.

You know, I remember reading, oh, 20 or 25 years ago, I became interested in how do I learn quickly from a book or how do I learn quickly when I'm, I mean, learning anything. And I read, you know, Mortimer Adler's how to Read a Book and, you know, speed reading stuff and whatever. And the big takeaway that I got from it was when you get a book, instead of sitting down and like starting on page one and then sequentially reading the entire book all the way through, start by taking a few minutes to flip through the whole book, see if there are bulleted lists, see if there are summaries, see if there are visuals, and, you know, see if in 10 minutes or something you can just extract the essence and, and then figure out the composition of the book. So that, because there's a very good chance probably with like 80% of books that you don't need to read the whole thing to actually get what you need from it. And that taught me to look at a book from like a higher dimension, in a way. And I'm, I'm using that as an example. That's what's coming to mind as you're talking. Because most people don't. Yeah, most people don't realize that AI, the generative AI is it literally looks at knowledge from a higher dimension. So you can give it a piece of knowledge and then you can ask it to help you find where in the knowledge. Right. Is the relevant stuff for you, and then apply it to whatever you're doing.

Yeah, yeah. And that's the thing. And that's, I think, where it comes back to kind of the philosophical piece. Because philosophers love to think at different levels of abstraction at the same time, right. They talk about, you know, what constitutes humanity, but then they also talk about the concrete situations and everything. And in the same way, these large language models that Underlie products like ChatGPT.

Have that understanding of the world, Right. They have a very deep understanding of human culture that was derived from the written words that it consumed. Right. But if there's a sentence, there's many interpretations to that sentence. Because if you ask somebody, how are you doing? And the answer is, I'm doing good, you could take that at face value and accept the fact that they're going good. Or that's one of the contrasts between Europe and US. If you ask that in Europe, you're going to get kind of an honest not doing so well. And in the US it's more of a greeting, right? So then there's that cultural layer, but then you could, like, ask, what was the intention of that person? Right? And that's a whole different layer. And, like, maybe he's playing into one of the insecurities of the person, because last time they had a bad interaction, now he's trying to move over it. That's a whole different layer, Right? So there's many layers to communication. And the funny thing is that these large language models are aware of many of them, so you can also talk to it at these different levels just to round that thought out. One thing that came to mind when you brought up how to read the book and that you kind of first inhale the framework of it and see how it's structured and what are the chapters and everything about. One thing that I learned over the course of my adult life is that actually, just before this podcast, you see these little two shelves? I constructed these, like, half an hour ago. I got them off Amazon. And one technique that I applied that I learned by, first of all, moving around and constructing furniture, but second of all, also prompt engineering, is the same thing that. That I said with the podcast and the same thing that you said with the books. When I go into the manual and there's 15 steps, I don't start with step one and take out the screws and start, like, I start by looking at the whole process, right? I look at the whole thing holistically, and then I go back to step one. I'm like, all right, like, first I'm going to, like, create a shelf, and then the box around it, attach the legs and connect the two. It's done, right? And once you have that roadmap, it really helps get into the nitty gritty. And again, it's these levels of abstractions. Understand how the thing is made and then get into the first step and you're going to ask different questions, and maybe you're going to lay out the screws in a more effective way and already skip ahead and things like that. So, yeah, it really helps to have a deeper understanding of the topics that you're handling. And that's what I use it for.

Well, really well said. And.

The example is really beautiful. When you're learning something new or when you're working on anything for that matter, you can use this to get that multi kind of higher dimensional perspective to orient yourself. It's like having a map of a city before you start trying to navigate around and you just kind of like see where you are. You situate yourself so that when you're navigating you feel more comfortable and it really increases your efficiency. That's a really great example.

Yeah, yeah, I like that one too. Rather than just like taking the turn to the right and like where am I? Like having that context, having. Yeah, having multiple layers of understanding.

So Igor, what are your favorite trends you're paying attention to lately? What's happening where you feel like you're watching, you've got a vector and it's going to be a big deal.

All right, so there's a few and I'll try to go with the ones that are a little less speculative and more concrete. One of them would be the fact that all of this technology really happened at the software layer over the course of the last few years. These large language models are usually hosted in a website or you get to download a local model. And I think we're really entering in 2024 concretely an era where all of that will be integrated into hardware. We've seen the first previews of that, you know, with the meta Ray Ban sunglasses and certain phones adding features here and there. But I think there's going to be a big shift because at the end of the day, what all of this is and what all of this technology is is just the next evolution of software. And software powers most of our hardware these days, namely smartphones. Right. And I think with players like Apple entering the arena this year, most probably.

It will change a lot and it will change people's perception of what their phones can do and how they function. And that gets me really excited because a lot of new startups are going to enter the space and try to disrupt the bigger players that obviously have the data and the hardware and they can just integrate it. So it's going to be really exciting. But I think one of the big trends that we're going to see is just this incredibly powerful software being integrated in things that we already have and into the ways we already live and just enhancing it. Right. And the whole like dystopian take that many people have. I think it's, you know, there's good reason to have that discussion. But I think in the near term we're in for a lot of, you know, quality of life upgrades with how we can interact with our phones and, and how we can use the health data from our Apple watches to like actively improve our life. And I'm excited about that. I think that's going to make a lot of lives easier and matter of fact, even save a lot of lives. I think being able to talk to health data in an educated way at a fraction of the cost what a doctor would charge is a huge win for, you know, for old world countries, for example, it's just not reasonable to, to have doctors anywhere and just living in the western world, we kind of take certain, certain things for granted like, like having access to health care. Right. But I'm excited about what this could do and I've seen some applications with like remote diagnosis or just like AI models that, you know, on very specific medical use cases. And I'm excited for what that is going to change in the world. And I think the upside is way larger to the downside, at least in a short to midterm, the long term discussion, that's a different one. But yeah, I think hardware is going to be huge and we're going to see more and more of that.

Okay, quick action step here. So the key is learning how to prompt better. You start out prompting by just asking questions and talking to AI like it's a helper, but you quickly realize that getting better at prompting is a huge accelerator and game changer for your own success in life. So these prompts will not only help you get a lot more done faster and accelerate your success, but they'll also help you really get prompting and get better at it. You can get all of these as a gift from me just by going to metamind Co Gift. That's metamind co gift and I'll put that link below and then just opt in to get it. I've also created a powerful video for you called AI in the Future of Business and Work and Success that demonstrates some of the key mindsets and prompting strategies to help you get up to speed and get a lot more out of generative AI. And it's also included as a free gift when you register. So just head over to metamind Co Gift to get your AI accelerator prompt pack and the other free goodies. All right, back to our interview. You mentioned having multiple GPTs and these things talking to each other and now we're starting to talk about agents and so forth. So what are you doing there first of all, what is an agent? Do you believe at some point that they wake up and they start being conscious? I mean, I just kind of want to go in that direction a little bit.

Okay, this is a great question too. There's a lot to unpack there. But first of all, maybe let's start with the definition. So basically, as we talked about, it's good to think about these products and let's just call them AIs, although a lot of people have problem with that. They're not really intelligent. It's really a matter of definition. But these AIs kind of act as an assistant and help you out. But when you make the shift to agents, it's really when you give them autonomy, when you give them the ability to act. So maybe we could compare this with James Bond having the license to kill and having the rights to make his own decisions and to go out there and to take action and just to function autonomously, independent of his creators or bosses. Right. Back in, back in the uk, just goes out there and he gets stuff done and he does it his own way. And that's kind of the same idea here. You kind of unleash the intelligence that is there and you give it tools to actually take actions. Right. So right now these tools are very primitive and these actions and agents don't really function at a level where you would be missing out if you're not using them in your day to day at all. It's very much in this kind of incubation phase and thought experiment phase. But yeah, the idea goes something like you have a ChatGPT that can also click buttons and use credit cards and actually do stuff rather than just assist you with information.

And you were speaking about having a whole team of these GPTs or these digital whatever they are, minds collaborating with each other. So how are you doing that? What's happening there?

Yeah, so there's this concept in prompting that is called Persona modeling, where you essentially set up the, you know, ChatGPT or the AI to follow a specific character profile. It has certain characteristics, it has certain preferences, maybe even a backstory, a way it acts. Right. A marketer is going to have a different set of preferences than Preacher, for example. Right. It's just two different, like you ask them the same question, you're going to get a different type of answer. And you can set up ChatGPT to follow a certain Persona like that. And then they become like better at certain tasks and worse at others. Right. And the whole idea here is it's great to have One agent or one marketer. But wouldn't it be better to have a product designer, a researcher, a research assistant, a marketer, a designer, a web designer, and then a project manager that kind of oversees all those people. And then all you kind of do as the user here is say, you know, create me the best website for xyz. And then the project manager goes ahead and it kind of creates a to do list, it assigns tasks, and then for a specific task, it might be using the researcher, right when it's about like doing market research and figuring out what works. And like, maybe for another one it will use the data analyst that you might also include, which like, looks at the data behind it. And then the marketer maybe thinks about the psychological component of it and the web designer puts the website together and they're like domain experts at multiple things, but together they're just stronger than a single generalist that kind of tries through it all. So the idea is instead of giving it one profile, you give it multiple and you spawn multiple instances of it. And then you kind of have a project manager that oversees that and makes all that happen. And then there's many terms for that. So you could call it like swarms of agents or, or the first version of it that kind of went a bit viral last year was called AutoGPT, where it's this concept. You don't have one ChatGPT, but you have eight. And each one of them is specialized and they kind of like work together to get things done that would be too hard for one person. And that is essentially the whole concept. Important to note that, you know, at this date and time, February 1, 2024, it's not at a stage where like, it would be very capable. It's very much in the experimentative phase. But from my read on the entire space and on what OpenAI is doing, they have clearly set a roadmap where that is the end goal and all the decisions and all the baby steps that they're doing in launching new models and new features. The release of the GPT Store mentions now, but I've been saying this since months, the end goal is this auto GPT like product. This product where you have a group of AIs that actually have agency to perform tasks and that work together and work for you. So eventually how I see the product developing is you'll be, you'll have GPTs that are multiple times more capable than what we see today right now. It's nice, but it's not incredible. And then you're going to be able to have an interface where you kind of select multiple and let them work together on a task.

Sorry, finish.

Yeah, so the task might be. So the task might be something like get me to inbox zero. So it's more of a goal. And that's like a different type of prompting than this, like, prompting an assistant. Because if you prompt an assistant, you kind of, like, lay out the different steps, especially if it's a new assistant. Right. But if you have a whole team of people that might already know about your preferences and how you like to get things done, then you can just set a goal, like, get me to inbox zero. And then they'll figure out what that takes. Well, first we need to look at the emails. Then we need to kind of create different to do lists on how to complete and how to respond to those emails. And then we might need different agents to do that and to come up with the data. And then in the end, we'll have to send the emails and done. You're at inbox zero. So, yeah, it's about setting a goal, and then rather than one generalist completing it, you have a team of specialists working together. That's the idea.

You have an example of how you've used this in a way that was, you know, productive for you.

I have to be clear that it's not an extremely productive use case. It's not something that I would go out there and, like, recommend. Need to use today.

Yeah, no, we're at the cutting edge here. We're experimenting, you know.

Yeah, yeah. One thing that I do, I have a little bot set up in a discord, in a private discord channel, a little bot that interacts with an agent framework like this. And what it does is YouTube research. So it goes out and it searches for certain keywords and it looks for the data on top of it. So it compares views with likes and comments and how much views the channel gets usually. And then it kind of does that for multiple channels and pulls the data together and presents with kind of a overview of outliers. That's something it can do quite well. So anything where there's. Where there's kind of like hard data that it can rely upon.

It does quite okay with that. Now, by no means is that, like, waterproof or, like, complete. And a lot of times just feels like it's kind of like just throwing darts and it brings you back the parts that it hits. So it's not like, oh, my God, I have, like, a godlike YouTube analyst at my disposal now. At all. But it's certainly interesting. And yeah, that is one that I experiment with.

Yeah, that's very interesting. Well, I mean, one thing about this is.

If it generates a hundred ideas and one of them was the insight that I needed right now to be able to move forward on something where I was stopped, that's a miracle. I mean.

It'S an absolute miracle. So I kind of laugh when, you know, I demo it and, you know, I ask it to generate 10 ideas for something and someone says, well, those three ideas are stupid, you know, and then they say, but that one, well, that's a really good idea. And I'm like, well, just tell it to generate 10 more or 100 more. And most people don't realize the value of an insight. They don't actually understand how incredibly profoundly valuable it can be to be able to go. Anytime that you have a question, you're stuck. You need to have, you know, something that bridges two concepts or whatever and be able to get ideas, even if it's a, you know, small chance that the ideas are going to be useful. And the reality is that there's a high probability that the ideas are going to be useful. Right?

Yeah. And isn't that how creative people work? Some of the most people. Creative people that I've seen, like, function this way. It's not like they sit there and they ruminate and then they're like, this is how we're going to do it. No, it's usually like, well, we could do this or we could do this, or how about this? Like, or we could take this angle and like, wait a minute, like, sit down for this one, because this is going to be wild. And then they say something. Right, but. And then like, 95% of that is usually, you know, like trash. But, like, the gem in there changes everything. So I don't know, maybe I just haven't met the right type of people. But, like, some of the most creative people that I've, like, had the pleasure of interacting with functioned in a very similar way where it's just throw stuff out there. And then it takes sometimes maybe a more grounded person to kind of.

Get through all that and filter out the ones that work. Or it takes some reflection or some writing, whatever. But, yeah, that's how Dear Generation functions. Yeah, I've seen that argument, and I don't really get it. It's like, okay, here's 10 ways to solve the problem. And then the argument is like, yeah, but eight of those don't work. So, like, okay, but like, that's not the point.

Yeah, yeah. Should entrepreneurs, CEOs, should they learn prompting themselves or should they get someone to do the prompting for them?

Well, I would rephrase that question, as in should entrepreneurs and CEOs be good communicators?

I guess you could argue that they don't have to. But you're going to have a very hard time in a world that relies upon communication. Right. Especially if your job is to kind of guide people and to give them the vision and direct them into so they go down the correct route. I would say communication is quite important. And if this is going to be the communication layer for all of most of the technology moving forward.

Yeah, I think it's quite important now. Yeah. You could make the argument that, okay, it's, it's a temporary skill that over time will not be needed anymore. But I would push back on that and compare it to the example of building websites. 20 years ago. We didn't have no code website builders where everybody could just easily throw up a website. So at that point there was a different level of value to being able to code in HTML and CSS and maybe some basic Java. But even though we got those tools and even though with AI, we're getting things like Copilot Pro, where you get GPT integrated into Your Excel, your PowerPoint and it gives you little suggestions on what you could be prompting. Right. There's these three cute little suggestions that are very relevant and very good and even the prompts are solid that it works with. But are you going to settle for that? Are you going to settle for free little suggestions of what Excel thinks? If there's a world of opportunities, like the little conversation we had in the beginning around kind of like philosophy and multiple layers of kind of abstraction and thinking.

I wouldn't want to pass up on that if I'm building something, especially in the hyper competitive environment.

That today's world and economy is. So I certainly think it is valuable and at least understanding the basics and getting the basics right, understanding what good communication with a large language model is will bring you a long way because then if you look at some of the content and some of the news, you will interpret it in a whole different way and you'll be able to direct not just yourself, but your entire team and you'll be able to point them in the right direction because you understand the basics of communication and when something new and interesting is communicated, you will be able to understand it and implement it. So I certainly think it's valuable and at the very least getting the basics down, not completely turning your head into a different direction and pretending like this is just a temporary skill. No, we're going to be talking in natural language to these models. And just because companies are going to try and make it as simple as possible with suggestions, prompt suggestions, because that's all it is, right? Like this layer on top that is, hey, do you maybe want to do this or this or this? Right. They're just prompt suggestions, but you could always go out and flesh them out more. And there are certain use cases that it won't suggest because they might not make sense because it doesn't have enough context. And you need to understand that to get the most out of it. So I certainly think there's value in that.

Just like the website example, I think it's going to remain relevant. Right. Just because there's no code, website builders, I think there's still value to being able to build websites or at least to understand how they're built. So you can talk to a person who does that professionally.

So, Igor, where's AI going? Where's it going in the next few years and how's that going to impact business, our personal lives? Is it going to wake up? Is it going to say, hello, I'm self aware, what's going to happen?

Yeah. So I think it's important to understand that at the current state of things, how the underlying architecture works, this is just not, it's just not feasible. It's an interesting discussion to be had and maybe there's going to be some technological breakthrough that will allow that and fulfill all these dystopian nightmares that we have from a lot of movies and literature and stuff. But right now it's just important to understand that what we call AI today is at its core a transformer based upon a paper published by Google a few years ago. And all you need to do, all people need to understand there, is that a transformer at its essence just recognizes patterns and reproduces them. There's no level of sentience there, there's no level of autonomy there. It just reads a bunch of stuff, gains a deep understanding of that and then reproduces it. And that is very powerful. But that is very different from, from a self governing and autonomously thinking being that a lot of people are afraid of. But always new discoveries can be made, right? And new technologies can be discovered. But at the current state of things, yeah, I would be afraid of that. I think it's a very useful tool that recognizes patterns and reproduces them. And if you understand that statement deeply you're going to be able to take advantage of that in your work and in your personal life. And yeah, I don't think we're going like, at least at the current state of things, there is a 0% chance of it being sentient. Even if you give it access to a computer, it just recreates what has been there. But who knows what the future holds and what researchers come up with over time. But at this current stage, the technology, at least what's publicly available.

That is not a fear I have. But yeah, just an advocate of making the best of what we have. And it's a powerful tool for pattern recognition and reproduction. And that's all it is.

Igor, thank you for coming and spending time with me today. I really appreciate it. We'll put some links to some of your work down below in the comments. So thanks everyone. And thank you, Igor.

I appreciate that this was fun.

So thank you for watching another MetaMind AI interview. Remember to keep practicing prompting. Keep learning. Get our prompt pack and quickstart video. Watch our other videos to learn from other AI experts and stay tuned to us here on MetaMind. Stay subscribed because this is going to get really interesting.