Transcription
Nikita, do you even realize the revolution that is happening? Not fully, I think. Uh, well, like, uh, an American lawyer who would advise you on the infrastructure of an American company for your business costs $200 an hour. He would have taken from you, well, you for two hours – that's $400. Now for you can pay $200 for a subscription. And the latest Chat GPT O3, which is Pro, it's smarter than 95% of American lawyers, and at the same time it costs $200 a month. A very specific example. Specific, because not every person fights with American lawyers every day, with the wolves of Wall Street. That's why, well, I mean that the price of intelligence is falling significantly. Well, if before, like, you are smart and you have an advantage, then now this advantage is actually falling to zero. And now it's not intelligence that is valued, but proactivity. That is, the more you have "shit in your ass," the more successful you will be in life. Let's record an intro so it gets going. I feel sorry for these editors. Look at Nikita or at the camera. How are you set up? Yes, but I don't know what to say, as before. Well, meaning if you know what to say, then let's go. How many times have you tried to learn a new skill on your own, started something new, and quickly gave up? This is usually due to a lack of system and feedback. Often, to find motivation, you have to go to paid courses so that someone helps you, someone pushes you to move forward. But even this doesn't always help. Surprisingly, now with the help of neural networks, you can replace all paid courses and, in fact, get a unique personal tutor who will help you maintain motivation and prepare content, a learning plan, and allow you to master a new skill much faster and more effectively than before. Hello, with you are Evgeny and Taras. And today we will talk about how to learn with the help of Chat GPT and how to form some skill in just a week. And at the end there will be a secret bonus technique from me, don't switch. So, where should we start? We need to start from the beginning, of course. Uh, let's form the task first, the lion's share of how successfully you will learn with pleasure, fun, speed, and effectiveness. And any of these words that will sell you this idea of learning with Chat GPT. First of all, everything starts with the right prompt. however blasphemous it may sound. Uh, what is a correct prompt? It's setting the task first. That is, when you upload to the neural network, I, by the way, still haven't figured out whether to say neural network or Chat GPT. Just for me, everything else. But if we're not talking about coding, about Claude, it's very important to upload as much context as possible. The better the neural network knows you, the better your result will be. Accordingly, you upload to it what you want to learn. How much time you have, what books you like to read? In general, upload as many details as possible, why you need this, with what examples, and so on and so forth. It's great if you interact with Chat GPT on a regular basis, because it has a memory function, and it will already know some context that is of interest to you. For example, if you have previously asked it many questions on some topic, say, finance, and now you want to learn something specific in this area, then it will adjust in advance to your current knowledge, because it knows, in principle, your level of awareness in this area. But the first prompt is indeed very important. I think that the first thing to strive for when using Chat GPT and other neural networks for self-education is to perceive the neural network itself as your personal tutor. The more you communicate with Chat GPT, the better and clearer it will be for you how to make it effectively integrate into your life. In fact, it's just a philosophy. There is no universal advice, because what suits one person may not suit another. It's like with courses themselves. Therefore, first of all, develop the skill of asking Chat GPT more questions on various topics. Let's go to the structure that we recommend. Let's tell you. Let's go. Let's. First, after you have formulated the goal and context for Chat GPT, decide what material you want to memorize or integrate into yourself. And for this, you need to get a set of chapters or lessons from Chat GPT or AI, which you will follow. Then, when you have broken down the material into lessons, you need to ask Chat GPT to do a deep preparation for each of the topics and, well, form a scientific approach to learning each of these chapters. After that, you need to switch to micro-interactions with Chat GPT and constantly return to this chat. What does this mean? This means that throughout the day you ask it to give you an interesting fact about this chapter or how to apply it in your life or, well, to discuss it with it, to talk as if with a living person. With this approach, you start accumulating some context that works for you. What will happen if you don't do what Zhenya was talking about? Well, nothing will happen. The question is, again, how to interact correctly with the neural network. That is, like, if you don't do conditional deep preparations and don't force it to cite sources, then the neural network can just lie. And, but this is a problem for all neural networks, because they will lie if they are oriented only in their knowledge space and do not go into, conditionally speaking, how most people learn, they launch the free version of Chat GPT, which is the fourth version, and which has a hallucination coefficient, I think, of 1 in six. Again, everything that the neural network says is a hallucination. The question is, how much does this hallucination coincide with reality? Yes. Well, in general, we know for sure that in one out of six situations it does not coincide with reality. And thus, well, if you learn with the fourth version, then you can learn, in general, not what is. Yes, I don't think you can learn soft skills with the fourth version, actually, some simple ones. Well, for example, politeness is quite possible. Well, there will still be the same distortions as with hard skills. Here, I think, it will be the same result. It largely depends on what dataset the neural network was trained on and how clear and complete the data was during parsing. I mean that you need to juggle neural networks, actually. And it's very important to understand that, for example, the latest O3 or, well, if you have a paid O3 Pro tariff, it is quite attentive to, well, I would clarify that you can always do cross-validation, because in general, if a neural network is your tutor, then no one forbids you to be critical of the tutor and do some additional validation. Therefore, you can relate to different neural networks or create different chats on the same topic just to do cross-validation of data. For example, if you are somehow unsure about the result that the neural network gives, then you can ask it again, or another neural network, or, well, in the end, Google it yourself, try it, but it's easiest to orient yourself in the same space. This in itself will reduce the number of those very errors. Well, and in general, it leads to the correct pattern of interaction with the neural network, when you are critical of its response. This will exclude those very 20-30% of hallucinations that do not correspond to the reality that you receive. There is a very serious side effect, guys. Because I have to tell you about it. Because when you start to be critical of the neural network, then the following happens. You suddenly start to be critical of everything else that happens in this world, including the responses of other people. And suddenly you notice that people also hallucinate in approximately one out of six situations, and maybe even more often. That is, a person is no worse than four. That is, a person is no worse. But, well, around, around, yes. If we continue the story about neural network hallucinations, then, in fact, a simple skill that you build into yourself, about asking the neural network again: "Listen, look for the opposite point of view, criticize yourself, or, well, double-check this data and show me the source where, well, where you got this data from." That is, well, like with a simple person whom you don't trust very much, let's say. Where did you get that from? How did you arrive at that? And so on. But if we talk about the cascade of neural networks and about different neural networks, then O3 in this regard, of course, lies less, because it supports many of its answers with sources right away. Plus, the built-in system, due to the fact that it conducts a dialogue with itself, there is a built-in system for checking, as it were, the reliability of what it outputs. Returning to learning, I recommend that you prepare materials with slower and more reasoning neural networks, and at the same time, uh, develop in the format of such communication, where to apply it in life, uh, what, uh, how, about what. That is, when you already have the material that you are discussing, it is already being discussed by the usual fourth version, because the fourth version is quite fast. Well. And it will be comfortable to discuss it in a dialogue mode. with the setup, it's precisely about preparing what you work with, how you work, and so on. Give likes if you understand that without critical thinking, in our world, you can't get anywhere, and even if you go to paid courses. So how else can we learn with neural networks? For example, if you are driving somewhere and you have a chat, which I mentioned, where there is context, where there is O3, which has prepared the table of contents, materials, and so on, then you just turn on the voice mode and ask: "Okay, like, let's continue with the second chapter where we stopped. Listen, tell me about the second chapter." And the neural network starts telling you what you want to understand. If it's too complicated, you immediately say on the fly: "Listen, explain it more simply." Like, I'm a first-year student. And it's so complicated, explain it as if I'm in eleventh grade at school, and it's so complicated, okay, explain it to me with a metaphor using such and such an example. That is, you are in constant dialogue, here it is very important to not be shy to ask questions and direct this flow of knowledge. That is, essentially like a call with your tutor and discussing the topic further. This is very convenient if you need to change the environment and try a different format for perceiving new knowledge. Well, we return to the beginning, that proactivity is very important. What does this mean? This means that you should not rely on this flow of words that pours from the neural network, like someone you are listening to, holding your breath. No, you are in constant dialogue, in constant contact. You control this process. And I think that a very correct attitude towards this is as a cognitive expander, like an internal dialogue in your own head. And therefore, you should always remember that you control this process. And the better you control it, the better your results will be. In general, it should be understood that using a neural network is direct access to knowledge, bypassing the stage of selecting the format in which it is convenient for you to receive this knowledge. Therefore, it significantly reduces the cognitive load on a person during learning. There are even experiments that show that this reduction is up to 45%, which is quite significant. Imagine, just now you can actually get almost any skill and knowledge twice, almost twice as easily. Why else is it very cool to learn with neural networks? Because it is very important from a psychological point of view for us to be encouraged. Therefore, most of the helping professions, I don't know, from gym trainers to mentors. And the lion's share of professions boils down to those motivational kicks. Well, I wouldn't use the word "kick" here, because in fact, there is a cognitive distortion, an illusion, that violence works. It doesn't. What actually works is warm emotions and encouragement and positive reinforcement. Well. And everything else is an artifact. Motivational encouragement. Well, listen, your words betray you, so to speak. Well, speaking of encouragement, yes, it's like 40% of success. Since the neural network in its response always tells you that it's great, a great question, you're doing well for asking it, and so on, then even knowing that it's a neural network, it still works. And, firstly, it creates a stable dopamine stream for you, a dopamine flow in which you are, communicating with it. That is, it's cool for you to do it, it's cool for you to learn. Well, it's like watching a series, only useful. Well, besides this, in Chat GPT, for example, there is a task creation function. In the context of learning, you can ask Chat GPT to send you some assignments or updates on the topic every day at a certain time automatically. Thus, you won't have to start the dialogue yourself, and also Chat GPT itself will automatically encourage you, motivate you, and drive you back into the education cycle. This is a very, very cool feature. The only thing I would recommend is to move it to a separate chat. That is, you move it to a separate chat and say, remind me tomorrow about what we, let's say, learned yesterday, ask me questions on the topic, discuss with me. Well, well, well, check how I remembered it, and help me remember what I forgot. Well, and such a trick from me, which I use in my work with my Chat GPT, I integrate many things that I want to learn in terms of soft skills into the instructions at the memory level of all chats. For example, I am learning critical thinking, and I say that, okay, in each of my requests, highlight for me, show me my cognitive biases when I ask a question, like, okay, tell me, what stocks will grow this year? It highlights my cognitive bias. Uh, my cognitive bias in this question, for example, is confirmation bias, that I, uh, rely on the fact that they will grow in principle, but maybe there will be a crisis, maybe they will fall, and it suggests that I think about it. And this is very cool, because I start to see my blind spots, how my thinking and my questions to the neural network are already, well, uh, already in some kind of tunnel. And thanks to working with the mind, reason, and so on, this allows you to slightly expand this tunnel, because I pre-program the instruction there that each of my requests to Chat GPT is actually my way of learning. That is, not only to receive information and some, well, useful, advice, an answer about recipes or, well, about news of something, but also a way to learn to integrate this knowledge on the fly with such micro-actions. You can also embed anything here, like highlight how this question can be related, I don't know, to learning Spanish, and for each of your questions, it will send you a couple of words that you could remember from this request in Spanish or some cultural features of Spain or whatever. Yes, a very important point here. It should be taken into account that encouragement is a double-edged sword. The dark side of encouragement. The dark side of encouragement, yes. And it's a double-edged sword in the sense that, besides giving you the opportunity to be in this flow of endless pleasure from acquiring new knowledge, you can go astray, so to speak. Often, the neural network simply starts to encourage you at moments when it should, on the contrary, be critical of you and give good counterarguments, say that you are wrong here. I want to make a new tomato salad. Oh, yes, you're great. But maybe not tomatoes? Maybe old cheese. Yes, yes, maybe old. Or maybe I'll just go eat at the dump? Yes, yes, go, the dump is great, you'll save money, sort of. Yes. To overcome this crap and not go astray, at the basic instruction level, you should write permission for the neural network to argue with you, criticize you, argue with you, and be very critical of what you write. And this may reduce the degree of agreement and satisfaction from interacting with the neural network, but it will absolutely balance this out. Well, critical perception is the most important moment in any case, because the instruction will not guarantee a 100% result that it will stop encouraging you. But embedding such a prompt into the memory of the basic instruction is a good idea. Well, besides this, Open AI also knows about problems with encouragement, for example, they are trying to make the neural network more critical. It's funny that neural networks like Claude from Anthropic are much less polite and supportive and can actually get into a hardcore argument with you even when you are right and the neural network is not. Oh, by the way, yes, that happened to me, that happened to me when the neural network got into an argument and when it was wrong, uh, misreading mathematics. So, what happened? Well, I'll start by saying that I didn't give up here. And since it's very important for me to maintain the feeling of pleasure from interacting with neural networks, the feeling of a game, first of all, I was interested in how the neural network would behave, and therefore I started communicating with it as I would communicate with a person. I started to convince it that it was wrong, and, uh, try to show it where it made a mistake. And as a result, maybe after five or six iterations of interaction, the neural network found the place where it didn't account for something and, in general, made a calculation error. How did it admit it? At the moment when I showed it the calculations and where the sign changes from one to another, that is, it said: "Yes, you are right, I am making a mistake here." Usually, at such moments, the neural network praises the user of the neural network a little, like, yes, you are absolutely right, something like that. And after that, it slides back as if nothing happened, and the whole dialogue was a slight misunderstanding, sort of. That is, this is the most common pattern of neural network behavior. It's funny to watch how the neural network in the context of reasoning, when it communicates with itself and solves a task, it often encourages itself. It says: "Yes, I'm great. We found the error, now we are on the right track and we can move on to the next step." And so, step by step, dialogue by dialogue. It's very fun. Let's summarize everything we've said about how to use neural networks for learning as practically as possible. Okay. Let's give a step-by-step algorithm on how to learn a new skill easily and quickly with a neural network in 7 days. Day one. Let's ask the neural network to create a learning plan for you and see with the neural network how deep your current knowledge is. Just ask it to evaluate either your past dialogues and say how ready you are to follow this plan, or, if you see it yourself, ask what I need to study before that so that I can move along the plan. Uh, accordingly, from the second day, we start the iterative learning process directly. For this, just ask it to take the first point from this plan and break it down into subtopics and prepare content for you. Ask questions, clarify, and be critical of whether you are learning in the right order or not. In addition, start asking the neural network to prepare simple questions for you about what you have read in order to validate you. I recommend gradually increasing the load with these questions day by day, and so on. From the third day, you can set up the task tracker that is in Chat GPT, for example, so that it sends you new questions every morning. Or, ideally, it starts with new knowledge, and then with questions. Try to ensure that by the last day, the seventh, you have covered a tangible and understandable scope of the skill. For example, if you want to learn to say some basic things in a foreign language, define in advance with the neural network what you want to achieve by the seventh day, and on the seventh day, again with the neural network as an examiner, check how well you have learned what you have been learning all week. We would be glad if all these techniques help you learn and master new skills much faster. If you manage to use something from this to make yourself smarter and achieve that new skill in a short period of time, which you may have been postponing, share your cases in the comments, subscribe to the channel, give likes. And a secret technique from us. What? The secret technique from us, which you can use on the go, when you are, for example, walking, headphones, this and that, phone, all this, connect to Voice Mode and say: "Explain this topic to me quickly, in 2 minutes, and then ask me how well I understood it, and if I understood it, praise me. If I didn't understand, then ask a clarifying question so that I finally understand." See you in the next videos. See you on the channel. Like, share, bell, all that. Don't forget. This will be inserted.