Speaker 1: Hello everyone and welcome. In this video you are going to learn step by step and with all kinds of details how ChatGPT works. ChatGPT is a relatively easy-to-use tool, but very few people understand and take advantage of all its potential. The goal of this course is that you learn to use everything it offers, whether you start from scratch or if you already have experience using it. In the training we will see how ChatGPT works, how to interact using text, how to interact with files, advanced tools, and finally we will see how to make money using ChatGPT. This training is going to be taught by Ramón, a member of my team who has been using Artificial Intelligence since before ChatGPT came out. He used GPT-2 and GPT-3, which are the precursors of the famous ChatGPT. I can't think of a better person than Ramón to do this training, since he has experience, uses Artificial Intelligence in his day-to-day life, and he is going to explain everything perfectly so that anyone can understand it. So having said that, I leave you with Ramón and we start.

Speaker 2: Hello everyone and welcome to this course where we are going to learn to use ChatGPT. This course, as Adrián commented, we have focused it so that it is useful both for people who have never used this tool and for people who have been using it for a long time, but I am sure that if you watch this video you will be able to extract a lot more potential from it, since in most courses or in most videos people only scratch the surface of what this fantastic AI tool is capable of. Before getting into the subject, I am going to briefly introduce myself, my name is Ramón, I work on the team of Adrián helping him prepare some of his videos, and I consider that I am the right person to teach this course, mainly because I have been using this tool for a long time. In fact, before even, as Adrián commented, that the ChatGPT itself came out, since prior to this tool there was already a previous version that allowed you to interact with the artificial intelligence of this company. Much more rudimentary, a little more complex, but that still allowed you to make your first interactions. That said, and without further ado, let's get into the subject. We will start with a brief introduction. What is ChatGPT? ChatGPT is nothing more than an interface to interact with a much larger program than ChatGPT, that is, we are going to enter the ChatGPT website and we are going to move the mouse, we are going to type, we are going to touch buttons, and behind that there will be a much larger program on OpenAI servers, which is the company behind ChatGPT, which will be running what is known as an LLM. LLM are the acronym for Large Language Model, or in Spanish it would be a large-scale language model. Basically, this is a computer program that, as I say, normally would have to interact with it through code, through typing code, but thanks to this company OpenAI we can interact with it through a much simpler interface, as we will see later. In the case of ChatGPT, the language model we are talking about is, worth mentioning, GPT. And this language model has gone through several versions, let's say, through several updates, in a way. The power of each of these updates, or of these versions, is measured in parameters, that is, in the amount of data that this language model uses to generate its results. In the case of GPT-1, it had 117 million parameters. In the case of GPT-2, we are talking about 1,500 million, or what would be 1.5 billion, let's say, Americans, that is, about 10 times more parameters than the previous version. In the case of GPT-3, it would be about 10 times more, that is, we are talking about 175 billion Americans. And in the case of GPT-4, we are talking about, and in fact I had to put it like this because otherwise it would not come out of the screen, 1.76 trillion Americans, that is, again, about 10 times more than the previous version. And that is why we have represented it like this, with this fade here, because, as I said, this would not end here, but it would be much higher, it would be approximately 10 times the number of parameters that GPT-3 had. That said, here we are talking about a language model, that is, a model that from text generates more text, but there are many other types of models. We have, for example, mechanical models, that is, models of artificial intelligence that allow you to design a 3D piece and apply different forces as if you were in the real world, to see, for example, from what point that piece you have designed would break. We have, for example, sociological models that allow you to predict people's behavior in different scenarios. These are used, for example, in a pandemic situation. For example, the pandemic that we lived a few years ago was very important to predict people's movements and how those movements would affect the different infections. We also have meteorological models that also use artificial intelligence to predict time. Or, for example, economic models that use artificial intelligence to predict future crises or future economic cycles. Within the category of language models, there are other similar models or variations of the same that allow us to do very different things than we are used to, for example, with GPT-3. One of them would be a text-to-speech model, which basically what a text-to-speech model achieves is that from a text input we generate a voice. This is very famous, the Azure iSpeech, which is an artificial intelligence model from Microsoft. We also have Speech-to-Text, which is basically the opposite. From a spoken text, let's say, from a voice audio, you turn it into text. We also have Text-to-Image, which would be all those models that you introduce a text to it and it generates an image. Like, for example, Stable Diffusion, which is one of the most famous. In addition, this is open source, that is, it does not cost the user and you can run it directly from your computer. We also have, which I am showing here through a video, Text-to-Video models, which have basically become very popular recently from Sora, which is the new artificial intelligence of OpenAI, from the same company that has generated ChatGPT, which allows you to generate videos like this one here. That is, in this specific example, I put in the text A woman with style walks down a street in Tokyo, full of warm and bright earrings and posters animated by the city, wears a black leather jacket and a long red dress and black boots. And from this text, it has generated this video, which, as you can see, is amazing. And these are the things that are being achieved today with artificial intelligence. All these models that we have just seen, as well as the different versions of GPT, work with something called Deep Learning. Basically, it is a type of program, a type of model, since it would largely be the same, in which the human brain is tried to imitate. That is, nodes are created, which would be the equivalent of neurons, and connections are generated between these nodes, so that if a connection that was generated during training is considered positive, that is, it is considered to have been a good connection, this connection is strengthened. That is, all the connections that are going to give positive results are strengthened, and all those that are going to give negative results, that is, results that are not desirable, the weight is lowered, they are not given so much importance. And by and large, as I said, it would work a bit like a human brain. The goal of programming GPT, that is, the intelligence behind GPT chat, in this way, is to make it understand and produce what is known as natural language, that is, written language, like the one I'm using right now to record this video, like the one you would have in a conversation, or like the one you would write, for example, in an email. To make this program respond, let's say, like a human, respond with natural language, what has been done has been to train this program, train this model with a lot of data, specifically 570 GB. 570 GB at first glance may not seem like a lot of data, since most of us on our computer either have much more than that, or we get close to that, around 200-300 GB of data on our computer, but really, if we transfer it to text, since this program has been trained with text, we are talking about approximately 20 million books in text. Obviously, in this case, it does not contain only books, it contains all kinds of data, it contains newspaper articles, it contains social media posts, it contains all kinds of information, but to put it a little in perspective, we have decided to make the comparison with books, and those 20 million books would be approximately twice as many books as the British Library, which is the largest library in the world. In addition, ChatterPT does not use only that initial information, but it is fed back with the information that users generate when using it, that is, every time you are using ChatterPT, you are using that information to learn even more and offer better results for the future. As for the company behind ChatterPT, which we have already mentioned before, it is OpenAI. This company was created in 2015, Elon Musk was one of its founders, but the visible face and the most recognizable person within OpenAI is called Sam Allman, who is this person here. He is one of the most important members of the board, and the curious thing, let's say, at the business level of this company, is that it started as a non-profit foundation, that is, it started as a company with an open source code, that is, a code that was open to everyone and that everyone could use for free, but in 2019 a non-profit company was created that controls part of this non-profit company. So, to put it simply, today OpenAI could be said to be a private company. And GPT-4, which is the latest version of the OpenAI code, can no longer be obtained publicly, for free, and the only way to interact with this artificial intelligence model is to go through OpenAI and pay the cost they charge you. This change is mainly due to the fact that Microsoft, the well-known company that produces the Windows operating system, among many other things, entered with an investment of 10 billion and forced a little to make this change, since today it controls 49% of the private company that is behind OpenAI. If we talk about competitors, that is, alternatives to GPT-4, which we remember is the artificial intelligence that is behind chatGPT, we have mainly two. We have Gemini, which belongs to Google, and we have Lama2, which belongs to Meta, that is, the company that used to be called Facebook, and which has the particularity that today Lama2 is still open source. That is, like OpenAI at its inception, and today it is no longer like that, Lama2 is open source, you can download it, you can run it on your computer, and you do not need to go through the ring or pay anyone or use any web page to be able to use this artificial intelligence. The problem is that you are going to need a fairly powerful computer, specifically for the most powerful version of Lama2, which would be the one that competes a bit with GPT-4, the artificial intelligence behind chatGPT. We are talking about that you need an RTX 4090, which is a fairly expensive graphics card, and you need 64 GB of RAM, that is, you need a higher-end computer, to be able to run it locally. Even so, as with chatGPT, there are portals, there are interfaces that allow you to interact with Lama2, but again here we are already talking about payment services. Anyway, today it does not make much sense to use any of these competitors, since chatGPT has a 60% market share, and this is very important, and in fact it is something that you can apply in any software, even though it hurts us, it is always better to use the software that has the most market share, that is, the software that uses more people, because if one day we have a problem, it will be much easier to find information on how to solve it, or find informative videos like this, if we are using the most popular software, rather than if we are using the competition. And considering that chatGPT has a 60% market share, not only of text models, but of artificial intelligence tools in general, we are talking about that, in principle, it is the best option with difference. That said, and after this reintroduction, we are going to get down to business, and we are going to start with the practical part, that is, we are going to see how we are going to create our account in chatGPT. To do this, we are going to go to this link, which we are going to leave you in a document in the description, and it will take you to this page here, where we can either log in, or create an account. We, in this case, since we do not have an account, we are going to want to give it here, to sign up, which is to create an account. It gives us several options, we can either create the account with an email and later with a password, or we can create it directly with our Google account. In my case, and I recommend you do the same, if you have a Google account, I'm going to do it with my Google account, since it is much more practical. We are going to click here, we are going to select the account we want, I am going to select this one here, and we are going to give it to continue. Here it gives us a brief introduction, we are going to give it to OK, and done this, we already have the account created. Now, if we go up to the left, we see that the version that comes out by default is chatGPT 3.5, which is not chatGPT 4, that is, the version is more powerful, the newest version that we are commenting. To be able to use chatGPT 4, as you can see here, it will force us to improve our account to a premium account. So, it is true that you can already start interacting with chatGPT using this more basic version, but we are going to see now a comparison of why I think it is essential to improve your account from the basic version to the premium version if you want to start using chatGPT, since the basic version is very limited. I want to emphasize that this is not a video sponsored by OpenAI, nor chatGPT, nor much less, but it is simply the version that I use and it is the version that I would recommend without a doubt if you are going to use chatGPT in your daily life. So, let's go back to the presentation and let's see the limitations that this free version offers that we are talking about. To start with the free version, we can simply use text, instead, with the paid version, we can use both text and images as well as voice. We will see this later. Then, what we commented, we cannot use chatGPT 4, that is, the latest version with the free version, instead, with the paid version, yes. We cannot analyze documents, this is also something that we will see later, but in the paid version, you can upload documents to chatGPT and analyze them and interact with them. We cannot generate images, we can generate images with the paid version, with DALI 3, which is OpenAI's artificial intelligence to generate images and that comes directly integrated, as we will see in chatGPT. ChatGPT 3.5, that is, the free version, is much less precise, as we mentioned before, chatGPT 4 is 10 times more advanced than chatGPT 3.5. The premium version also has more memory, that is, you can write more things to it and make it remember all the answers it will give you, and this is very important because, again, we are talking about approximately 10 times more memory in the case of the paid version than in the case of the free version. In addition, the paid version has internet access, that is, you can connect to the internet to find answers, unlike the free version, which simply has the data with which it was trained in 2021, that is, it will have very obsolete information. In addition, the paid version has something called custom GPTs, which are, by the way, specialized chatGPTs, which we will also see throughout this course. And finally, the paid version of chatGPT does have a negative part, and that is that, as its name indicates, it is paid. It costs 20 dollars a month, but if you stay until the end of this course, you will see that I think it is worth all and each of those 20 dollars. Having explained this, let's see how we would improve our free chatGPT version to the paid version. For this, let's go here, let's go here where it says Upgrade Plan. Let's click here on Upgrade Plus. This would be if you want to use it at a corporate level, let's say, use it with more people, share chats, etc. So, for this tutorial, in principle, we will use the plus version of a normal user, let's say, of a new company. And here, it would be simply to enter our payment data. And with all the data filled in, we click here to pay and subscribe. This is also important to take into account that it is a monthly subscription, that is, it is not a single payment, month by month they are going to charge us this amount, which would be the base rate of $20 plus the taxes corresponding to your country, and month by month they are going to charge us until we decide to cancel it, that we can cancel it at any time. Let's click here to continue. Now that we have the paid version, we can go up here and as you can see here, it already lets us change to GPT 4. And this already shows us that we have the paid version. If this does not appear, it is because maybe there has been some error in the payment or maybe something strange has happened and in that case it would be a matter of trying again. It would be very strange that they charge us and that the premium version was not activated. I do not know anyone who has had this happen and it is not a frequent error. So I said, if it does not appear here it is because probably they have not charged us and in that case it is a matter of trying again either with another card or with the same until we get to make the payment. That said, we are going to do a brief review of the interface so that you can see how each of these sections works. First of all, and the most important thing is the section here which is the section to write messages. This will be what we use to interact with ChatGPT. That is, here, for example, I can write what is YouTube and it will give me an explanation of what YouTube is. The funny thing about this and hence the name of ChatGPT is that beyond having asked him this, what I can do is from this text ask him another question or give him another indication. For example, from this text I can say OK, I like your explanation but make it shorter. And as you see here, you have written the same explanation that we have up here about YouTube but much shorter. That is, we can keep interacting about this. And for example, I can ask Can I upload video games to YouTube so that people can play them? Obviously, you can't. And as you see, it answers me here also based on that previous context. Another important thing when interacting with ChatGPT is that, for example, I can return a previous message like this one and tell him the opposite. For example, I want it to be longer and edit the message. And we have done this by clicking on this little button here which is like a pencil. Then, it will erase everything that was there until then and it will make the message much longer. I just finished generating and as you see, I have done it. We have the same explanation we had before but here we have three paragraphs and here we have one, two, three, four, five, six, seven, eight paragraphs. It has done exactly what we have asked. What is the cool thing about editing messages? Well, even though I have lost the previous answer right now, I can click here and not only the previous message is preserved but the whole thread is preserved. That is, we can have different branches because from now on I could write a new message here. For example, I can upload something that is not going to be uploaded to YouTube. Physical documents such as notebooks or papers to YouTube. And as you can see here, what has been generated in a certain way is a sub-branch. That is, we would have two main branches, the one of ok, I like your explanation but make it shorter and the second branch ok, I like your explanation but make it longer with this message that we saw before with the longer explanation. But also, if we go here to the first branch, we have those two sub-branches, the one of ok, I can upload games to YouTube, I can upload videogames to YouTube so that people can play them and the other branch of the physical documents that comes later. They are nonsense, let's say, that I ask you here I have explained a little how the part of the chat works, we have to emphasize also that documents can be uploaded, that this is something that we will see later. They can also be dragged. Then we have here the different chats, that is, here we have a chat but now, for example, if I click here I'm going to create a new chat, for example, I say hello, how are you? It responds here and the new chat that I have generated appears. If, for example, I copy this text here that I have prepared here and I ask how much the Mount Everest measures, Max is creating a new chat and all the different chats that he has opened in the past have been saved here and this stays here forever unless you decide to delete them. For example, we have here the first of all, the one we talked about on YouTube, the one we asked in the chat how it is and here the one we just generated of the Mount Everest and they are always generated with the name, that is, the first phrase that you have passed is generally put as a name. Now it appears in a chat but if I reload, generally, let's see, maybe it has been bugged by this, but hey, I said it. In principle, it stays with the name that you have put in a first instance. This, anyway, is not very relevant since you can always create a new chat. Another thing, look, indeed, here it has already been put. Another thing we can do here is if, for example, we do not like the format in which he has made the response but we do not want to change the text as I have put here, we can give this answer here to regenerate and it will respond again to the same question and as you can see, again, as when we edit a message, a new thread is generated here and now we have the first thread and the second that is being generated right now. And another thing that we can do is that if, for example, now we ask him a longer explanation, explain it in a more extensive way, which would really waste a lot of time since now it will be a lot of time generating the text, what we can do is with this little button here, stop it and move on to the next, that is, we can write it or in a new message or modify what we had already written here. We have many options. In addition, here we have another tool that says Explore GPTs but we will leave this later since it is a little more complex and advanced. Before, other parts of the course come. Another important part to take into account is the issue of adjustments that we can enter here I'm going to leave it in the dark because I like it more. We can go there a double factor of authentication. Basically, this is to make your account safer. We can, for example, delete all the chats we have so far. If I click here Delete All, all the chats I had so far will be deleted and all this among many other adjustments but at the moment either we will leave them for later or they are not too relevant. Having reviewed the GPT chat interface, let's see now what are some of the most optimal ways to interact with this tool to offer you a complete vision of what are the possibilities it offers. For this, we are going to review what are the prompts that basically prompt in Spanish would mean something like indication, stimulus or instruction and it is a text to interact with the AI. There is a concept that is known as prompt engineering that would mean in Spanish ingeniería prompts which is not an engineering but it is usually called that and the objectives of this concept let's say of this philosophy is one is to use the optimal prompt that is, to use the optimal indication through the text that we will introduce to GPT chat in order to achieve a much more effective result, much more efficient and much more consistent. Above all, we also speak in this case of consistency that is, to obtain relatively predictable results for the format for example. So, in order to master this prompt engineering what we are going to try to do is to build a toolbox in order to be much faster and obtain better results when interacting with the tool. The first prompt that we are going to work on is to guide and develop and you have to put yourself in the following situation. Imagine that they ask you what is YouTube? The worst thing is that you do not know how to answer since in the first instance a thousand things come to your mind. This is a bit the same thing that happens to the GPT chat since, as we said, it works as a human brain but of course, it has the computational capacity to generate it instantly. Then it releases exactly what you are asking it. So, if we guide it and, for example, instead of asking what is YouTube, we ask it what is YouTube, explain its three fundamental aspects or what is the size of Mount Everest, give me the result in meters and feet above sea level. If we are much more precise we will get much closer to the result we expect. This is the same as if you talked to a human but applied in this case to the GPT chat. So, we are going to do an example of this and we are going to see the GPT chat and we are going to copy the first prompt we had which is what is YouTube. If we ask it directly this, we will get a similar result to what we have obtained before. And as you can see here, I have already finished writing the prompt and it is already appearing here that it was founded in February 2005, and so on. Imagine that all this does not interest me. As we said, I only want the three fundamental aspects. Instead of asking it directly what is YouTube, we are going to open a new chat and we are going to see the three fundamental aspects. Let's see. And as you can see, we would have it here. The same information has been written to us, but this time it has already been reduced here to its three fundamental aspects. So, yes, because if we were in a hurry or we simply want some information, it would be much easier to locate. And a little bit in the same line, in the other cases, for example, if I want to know how good the mount is, I have to specify to what extent I want the result or with respect to what level. Or, for example, if I want to know what is the difference between drop shipping and drop service, if what I really want to know is which one has more profit margin, the question is given. For example, we are going to do the example with this one. If, for example, I ask this here, what is the difference between drop service and drop shipping? Ok, it has finished generating. Well, as I said, it is a much more extensive explanation. But if I, for example, ask the second part of the question, tell me which one has more profit margin. Here, by the way, there is a missing one. Well, it would have already finished generating and the explanation is already 100% focused on what we are looking for, which is with respect to the profit margin. What we have to keep with this is that it is important that whenever we can, we give all the context and all the details that we can to ChatGPT, since in this way it will be able to get a much closer result to what we are looking for. That is, not to simply stay in a short question without too much context. Another quite useful thing, especially in order to look for answers to questions, because generally whenever we ask ChatGPT, it makes a fairly long answer, is to add things like this at the end of our question or as an answer to its answer. Here, in this case, for example, make it shorter or write it in 100 words, which would be approximately about 30 seconds of reading. It is also important to emphasize that they will never be 100 exact words. Sometimes yes, sometimes not, but generally there is always a little variation there. Or, for example, you can ask for this. Without extending the text, you just explain and there you introduce the part you want to be explained in less than 100 words. In this way, it is clear exactly which part you want to be explained. In this case, the suspended points would represent the part you want to be explained. Throughout the video, whenever you see this annotation of two brackets and something in the middle, it will be because you have to replace it with your text or your word or what you are looking for. For example, in this case, if we go back here again to the text that ChatGPT had written to us about DropService and DropShipping, we can ask for the same, we can tell it to make it shorter. Here, all this text that we had here will be summarized in a single paragraph much more digestible. DropServicing really offers more profit margins sorry, offers higher profit margins than DropShipping because the services have a perceived value higher, all this paragraph, well, all this text rather, would be summarized here. Again, taking back what we had said at the beginning of the video, in case we want to put it in 100 words, we can use the edit function and replace this with this. And we will have the answer here and if we copy this and go to a character and word counter, you see that it tells us 108 words here. How about we comment that it will never be exactly 100 words but approximately it will be quite close. Finally, the last prompt that we comment that would be this here, without extending the text, explains only this part in less than 100 words. We could do the same, imagine, we come back here, we edit the text again, we paste this prompt, which by the way you will have them all in the document in the description, and for example, I want you to explain the issue that has lower variable costs. It is not even necessary to write all the text, but you can say directly, without extending the text, explains only section 2 in less than 100 words. And as it already has the previous context, since it is a little what we commented, the grace of chatgpt that takes into account the previous texts in order to generate new texts, if I write this now, it already directly explains the entire text in less than 100 words. And if I write this here, it explains and directly I'm going to talk, well, that. They are related with time, with the ability of providers, such and such, and it is directly extending this explanation in 100 words. In this case, it would be extended that if I'm not wrong, these would be less than 100. Correct. And nothing, it is doing this same explanation in less than 100 words. Another thing we can do through prompts is fill in gaps within a text or replace the text. Imagine that I have this text, YouTube is a video exchange platform, such and such, basically a text writing YouTube. And I have this paragraph that talks about the feature of YouTube Premium, and I really do not care that this is talked about. What I want is to talk, for example, about monetization. Well, what I can do is grab this text that we originally had on YouTube, fill in the space between brackets, which would be that paragraph that we had in the monetization system through ads. Then it would be as simple as grabbing this prompt. Then we would go here to ChatGPT, open a new chat, paste this prompt. It is very important that the original text is in brackets, because that way we help ChatGPT to distinguish it better, even sometimes to distinguish, let's say, our instruction, which would be this here, from the original text, what we can also do here is jump a line. And this will help ChatGPT understand that this is the text we are referring to, and this down here is the same thing we have here. And now, if I hit enter, and as you can see here, we have already generated a text that would go directly into this space that we had pending to fill in. If we do not want to do it manually, because we are very lazy and we want it to directly integrate into the text itself, then we directly tell it to rewrite the entire text with that new change. And here, as you can see, it has rewritten this text we had here, adding this paragraph, exactly what we asked for, this paragraph that we have added here, which basically what it has done has been to extend the explanation about advertising. Another thing we can do is refer to a paragraph to replace it, but with ordinal numbers. That is, if, for example, our text has five paragraphs, we can ask ChatGPT the following, replace the fourth paragraph with a new paragraph of approximately 60 words that deepens the monetization system through ads. So, in this case, we are going to take back this same text, which already has the paragraph in question, and we are going to ask it for a slightly different prompt. It would be to copy this prompt, but instead of asking it to deepen the fourth paragraph, we are going to ask it the second paragraph with a new paragraph and instead of asking it what it is already treating this paragraph, we are going to combine it with the previous section and we are going to ask it to make it much shorter. So replace the second paragraph with a new paragraph much shorter. We can even ask it the same thing we asked here. Rewrite the entire text with that change and in this way we save asking it later. And here it has rewritten the same text, that is, the first paragraph is the same, the last paragraph is the same, but the middle paragraph, which would be this one here, as you can see, is much shorter than the previous paragraph. While preparing the next explanation, I got this error message and I thought it was important to show it since it is an important limitation that has today to be using ChatGPT 4. Basically, what this message tells us is that we have reached a limit because despite being paying a premium version, even so, this has a cost, let's say, at the level of computer processing for the company that is behind ChatGPT, for OpenAI and they put a limit that basically will be removed in a few hours and right now what it is telling me is that I have to wait until 5.46 pm, which if I'm not mistaken is effectively in two hours before being able to continue using GPT 4. Even so, it tells you that you can use GPT 3. What I'm going to do now to continue the video with GPT 4 is to change to my other account and so we can continue with the video. Ok, I'm already in my other account so let's continue with the course. The next thing we can do through Prompt Engineering is refactorize. Refactorize would mean changing the format, that is, from a format that can be, for example, without spelling errors. Really, what we have explained before cutting would also go into this category but as it is a very useful category I decided to give it its own section. But well, let's go back to the repository where we are and let's start with the first example which would be correcting texts. This is very useful because many times online correctors do not take into account the context of the sentences and, for example, they can make corrections of spelling errors but they cannot make corrections of the structure. Oops, so let's see how the example of one of these corrections would be. Let's go here. I have a text here that I have prepared before which is basically the same text as before but with spelling errors. Here, for example, we get platform with two M's exchange with V that you see is already marking the word itself. So, let's grab this text, we are going to paste it here, we are going to put it in quotation marks, very important, this is always that we paste something, a text with which we want to interact, we put it in quotation marks and we are going to paste this to improve the structure. Then we will grab here and we are going to paste this here. And this we would remove because this is just to indicate that it would be the place where your text would go. We are going to hit enter and, as you see here, we are already correcting all the errors. Here, platform to platform, exchange to exchange, everything is already correcting and there are no errors in the text. In addition, we asked to improve the structure, I do not know it is doing it but maybe I have detected that the structure was already good that ChatGPT had generated for us, that is, it already has a fairly coherent structure, so I think that simply what it has done has been to correct the lack of spelling. Anyway, if we passed him a text written by us and that has a margin of use, surely he would also do it. Another option that we would have within this category of prompts is the fact of simplifying the explanation or explaining it, let's say, for another profile of people. For example, we can tell him to explain it so that a 12-year-old boy can understand it. This is very useful if we want an explanation to be very simple and understood by everyone. So it would be as simple as, for example, in this same text that we just passed before, we are going to paste this prompt and here it is already giving us a much simpler explanation. Here it told us that YouTube is a platform for video exchange online, where the user can upload, share and visualize videos. And here, instead, it tells us, imagine that YouTube is like a giant library, but instead of books, it is full of videos that you can see about anything you can imagine. You see, it is a much more adapted platform to what a smaller child could expect. And in this way, you can get an explanation adapted, so to speak, to all levels, not only to a 12-year-old boy, since if a 12-year-old boy understands it, practically anyone will understand it. Even so, maybe a 12-year-old boy has technological knowledge that an older person does not have. What do we do then? Well, we can ask him, for example, to explain it so that a person who is not a technological native understands it and in this way, it is applied to different profiles of people, depending on what the text is and what our goal is. Within refactoring, we can also ask for, for example, translations. Again, here, ChatRPT is very useful because, despite the fact that the Google translator works very well, again, sometimes the subject of context does not take it into account. And instead, an artificial intelligence does take into account the context. So, this same text that we have here, we can ask it in English. Translate it to English. English. Well, remember that the spelling or typos don't matter because it will understand it anyway, as you can see here. And here it is already translating it to English. Imagine YouTube is like a giant library, but instead of books, it's filled with videos that you can watch about anything you can think of. And it is basically exactly the same text, but translated and, above all, very important, with context, contextually, that is, not word by word, as many internet translators would do. Within what would be translation, we can also do contextual translation of words. For this, you have to give it a little context of where that word would fit. I have used this throughout the preparation of this video, for example, specifically for this slide here. Large language model. The literal translation will be something like a large language model, which does not reflect exactly what it is. So, I asked him, look, I want a translation, but within this context. So, if we go, for example, back to the slide where we are, which is this one here, let's imagine that I want to translate, as you said, the word prompt. If I put prompt to the Google translator and I translate it, well, in English to Spanish, it will tell me immediate. And this does not make sense in the context that we are looking for. We are talking about a prompt in artificial intelligence. So, I'm going here, I'm going to see a new chat, I'm going to copy the command that we are seeing a moment ago. How would you translate the word, and we are going to put it in quotation marks, prompt, in the context of an artificial intelligence tool, artificial, based on a large-scale language model. And here, as you can see, well, I have not yet finished the general, but we can already see it here. It has given us three translations, which are much more in line with what a prompt is, much more than immediate, which would not make sense. Fast, prompt, punctual warning, none of these translations fit, since the Google translator does not have the context. On the other hand, with this, being an artificial intelligence and understanding what you are asking for, you can make a much better translation than you would get with a page like this. Another very useful option, and in fact it is one of the ones that I use the most when refactoring, is this prompt here, which basically allows us to get a much shorter and much easier to read explanation from a long explanation, that is, imagine that I have a text in which they explain one thing to me and I do not have time to read it whole or I do not want to read it whole. Well, I asked for this here, basically I asked for an adaptation of your explanation in a bullet points format, which basically the bullet points is, well, look, I think that in this same PowerPoint we can find an example of what bullet points are, it would be basically this, short phrases that are sometimes not complete by points and that make reading easier and explain a concept in a much shorter and much easier way to read. So, what we are asking for here is that you adapt the explanation in a bullet points format with the essential information, that is, removing everything that is not essential, using incomplete sentences because we really do not need the sentence to be complete to understand what I said, to make it very fast to read. And here we are explaining a little what our goal is, we want it to be easy to read. We are going to apply this prompt to see a little what the result is, following, for example, with the same example that we had before, we are going to resume the Google explanation and we are going to edit this text here to ask for the same. So, we copy this, we paste it, we remove this here because this is in case we want to pass a text to it, and we hit enter. And as you can see here, from this text, which was much longer, it has only taken us the essential points. Online video exchange platform, founded in February 2005, allows you to upload, share and view videos. From a fairly long text, it has taken the essential information from us and has put it in such a way that there are no unnecessary words and that simply with this we can understand the entire previous text. The next type of prompt with which we can talk to ChatGPT is to ask for tables, that is, there are situations in which we may be interested, but a table of a specific topic, either to compare, to order information, etc. So, I said, for example, we can ask it to classify something with its negative points and its positive points. For example, in this case we have this text here, the one that we have just generated, from YouTube. Well, I can ask for the same. We go here, we are going to copy the prompt, we are going to modify it a little so that it makes sense with the previous context and we are going to say, classify these aspects of YouTube in a table with its positive points and its counterpart negative. So, here we are also going to give it a small indication, which is that, for example, the fact that it is founded in 2005 is not a positive or negative point, so we are going to say, omit the points where it does not apply. And this is very important that we ask for a table. We are going to hit enter and at the beginning something weird is going to appear, but as you can see here, the table is already being generated. We already have it generated here and it has done exactly what we asked for. First, the aspect, varied content, positive points, a wide variety of topics and available genres. Counterpart, it can include low-quality or inappropriate content, quite accurate. Aspect, educational tool, access to tutorials and free knowledge, wrong or unverified information, it can confuse. And as you can see here, we have it classified in a table. And this goes further, we can do many more things apart from positive points and negative points. For example, if now I go here, I can say this prompt here, this one that I had prepared here, in which we basically ask for a table with more things, apart from positive points or negative points, which can be a thousand things. In this example, we are going to do it with cars. So, we are going to copy this prompt, we are going to paste it here, and we are going to say, make a list of the 10 most popular cars in history, classified in a table, in order of year of release, for example, with three columns. One for the manufacturer, another for the model, another for the power, and a row for each, in this case, for each car. And we remove these quotes here. Here the text has already started, here in the first instance, it is simply putting us to the list, and we are going to wait a bit for it to finish generating this. Now it has finished generating the list, as you can see, here we have all the cars, and it will start generating the table. This here, very important, many times, when we go to ask ChatGPT for a prompt, this circle that is filled up will appear. This is because ChatGPT has integrated a code interpreter. A code interpreter is basically capable of running simple programs, specifically in Python in this case, and basically what it is doing here is doing code functions. For example, I understand that maybe here what it is doing is analyzing the years and sorting them in the table, because maybe it is not able to sort them with its knowledge, let's say, with the context, maybe it is not able to sort them numerically by year, and what it is doing, I understand, will be to use this information to generate the table and generate it correctly. And as you can see here, as a result of having used this code interpreter, the years are in order as they should be. And we already have the table here and we have exactly the information that we have asked for 10 cars, the ones that ChatGPT has considered to be the most popular, with the manufacturer in the first column, the model in the second, the year in the third, and the power in the fourth. And as we have asked for them, they are sorted by their year of release. Another quite important thing that we can do with these tables is that, directly by dragging and copying them with CTRL-C, we can go to an Excel and directly paste them. This is quite useful if, for example, we wanted to reorder the table based on the year, so that instead of in increasing order, it is in decreasing order. Well, we could operate with this data directly with an Excel, as if we had created the Excel from the beginning and everything is kept in order. Even here, these texts that were in black in the table itself, have also been kept in black in the table that we have pasted. And in the case that we do not like the design, we would simply select it, we would continue here and we would put the design that we want. But these are secondary things from Excel that they would give for their own course. Another thing that we can do with ChatGPT, that is, other types of prompts, would be these here to sort information. And a very strong point of ChatGPT is that its ability to interpret information, in most cases, is far superior to that of a human. And we are going to show it now with an example. Imagine that now I go here to CoinMarketCap, and I copy all this information about Ethereum. I copy all this information about Ethereum, I copy, I copy, I copy, I copy, I copy, I copy. We could even copy all this from here, for example, and all the news here from Ethereum. I copy it all, up to here. Even, look, we are going to download more, we are going to copy it, literally everything, badly copied. I copy all this, I open a Word and I paste it here. Let's see what we are pasting, here it is. As you can see here, the information is crazy, that is, we have giant images, we have weird texts here, all this is not understood very well, it is difficult to know very well what is happening here. On the other hand, if this same information that I have just copied, I go to ChatGPT, and I go with this prompt here, I am going to pass you disordered information, order it, and we paste the information that we are talking about. We see here all the information that we have pasted, which is what we commented, that there are a thousand different data and none of them is very ordered. And ChatGPT is automatically interpreting each of those data and is organizing it for us, let's say, so that it makes sense, so that a normal person can read it. This, obviously, here we already have it in this way, we have it well ordered, this would not be necessary, but let's imagine that, for whatever reason, you have a code, that is, you do not have this, but you have this mess of information, you could go to ChatGPT and do exactly the same. In that same line, another thing that we can do is, instead of asking it to order the information in this format here, which has been putting one thing after the other, we can tell it to order it in a table format, that is, as we had seen before. For this, we are going to use this other prompt that we have here, I am going to pass you the disordered information and order it in table format, omitting the incomplete information. Well, we are going to go here, we are going to open a new chat, we are going to ask for this prompt here, and this time, instead of ethereum, we are going to do the same, but we are going to do it with bitcoin, we are going to copy all this, let's see that I am leaving this here, we are going to copy the same thing that we have copied for ethereum, but this time for bitcoin. We are going to paste it, we are going to go down, as you can see here, we have all the same information, the same as with ethereum, and below everything, the table is already being generated for us, with all the well-ordered and easy-to-visualize information. Another interesting type of prompt that we can use are the perspective prompts. Basically, what we are going to do is ask ChatGPT to behave like an expert in a certain field, and to assist us in different tasks based on the expertise that we have asked for. That is, imagine the situation that I would like to get out of prison. Well, there are different ways to address the problem. You can consult with a nutritionist, with a psychologist even, if maybe I have nightmares at night that I can't control. You can consult with a personal trainer, if, for example, I am already eating well but I need to do sports, with an endocrinologist, that is, with a hormone doctor, let's say, or even with a coach. If my problem is that I really do not get the motivation and I already know exactly what my problem is, but what I lack is motivation, I could consult it with a coach. Well, we can talk to ChatGPT and ask him to help us based on different criteria from different specialists. So we are going to do the test with this prompt, we are going to go back to our reliable ChatGPT, we can close this already. We are going to open a new chat and we are going to paste the same thing that we are seeing here. Guide me on how to lose weight from the perspective of a, and we are going to put, nutritionist. Well, I put it in capital letters, but it doesn't matter. Well, I've been waiting here for a while, but as you can see, it's generating a very long text, so let's stop it. Ah, look, it's just ending now. But basically, as you can see here, all the tips he has given us are related to nutrition, to diet, etc. Energy balance, calories, that is, everything goes in line with the field of nutrition. If, for example, instead of that, we open a new chat and I ask him the same thing, that is, guide me on how to lose weight, but from the perspective, for example, of an endocrinologist, again with capital letters because I was wrong. Well, here he is already giving us much more related tips with the issue of hormones, let's say, with the tips that an endocrinologist could give us. Well, here, reduce sugar intake, glucose levels, portion control, evaluate your hormonal health, etc. Something important, I'm going to stop here now the generation that I've seen before here in the ... no, in this chat it wasn't, let's see, let's reload that it doesn't appear to me. Here it is, okay. Something important that is happening here is that in point number 10 of the first chat that we have opened regarding the subject of perspective, he is telling us to look for support from a professional. This is an important point to emphasize regarding chatGPT and it has come to my mind regarding this and it is that chatGPT can be very useful and can be very intelligent in a way, but it has its flaws. So there is not a person supervising it and the advice that I give you in some case can be wrong, it does not usually happen, but it can happen and it can have consequences. Obviously in the real world, imagine that now suddenly they give me bad advice, I follow them blindly because I trust chatGPT and no one is correcting me. So that is why it is always important, if important decisions are what you are going to make regarding chatGPT, you can also consult it before with a professional and not blindly trust everything he says. This, as we will see later with the subject of investments and such, is very important. Another option we have is to ask chatGPT to explain a specific topic from the perspective of a specific person. For example, explain bitcoin from the perspective of an economist or explain bitcoin from the perspective of a telecommunications engineer. Then, depending on the perspective we ask him, he will give us an explanation or another. If we ask him, for example, from the perspective of an economist, maybe he will explain the monetary part of bitcoin, the monetary policy part, the inflation it has, etc. If we ask him from the perspective of a telecommunications engineer, it would be more of a computer engineer, then maybe he will explain more the aspect of the blockchain, how nodes work, etc. So it is also something quite important. Also in the area of ​​what we talk about, guiding chatGPT towards the goal that we are looking for. Another quite useful type of prompt is that of self-criticism. This comes in handy in the case that you have written a text, that you have written manually, and you want to know if it is a valid text, if it is a good text, or if there are things to improve. Then you can use this prompt, which I repeat, you will have it all in a document that we will leave you in the description, which will tell us, I want you to act as an expert in, and here we would put the field in which we want him to be an expert, and that you make a constructive criticism of the following text. And here we would pay him our text, our image, our document. This can be different types of documents. Find the errors in it and correct them if you think it is necessary. Provide suggestions that can improve the result. If you do not find errors or improvements to do, I want you to communicate it. If a specific section is well developed, I want you to let me know too. This is basically what we do to see, because if not, sometimes chatGPT what it will do, if you ask for self-criticism, if you ask it to criticize a text that you have written, sometimes it will criticize because yes, because it is the instruction that you have asked for, and no matter how good the text is, maybe it will criticize because yes, because it has understood that you expect a criticism yes or yes. On the other hand, if you are telling it that there is a possibility that there are no errors, then it will consider it and it will give you a much more constructive opinion, I want, and that can help you a lot more. So this would be another type of prompts that we can use. In fact, we can even test it with the example that we had before YouTube, to see if I find the chat now. Here, for example, it is already looking for the text. We have the text this, which is the text that we had before YouTube with spelling errors, and we are going to see if we pass this prompt here. Then we are going to remove this and we are going to say I want you to act as an expert in written expression and dissemination. And make a constructive criticism of the following, because in this case it is a text, so we are going to put the following text. And here we would put the text that we want, in quotes, in this case because I have put the example that we had done before for the correction of errors. I have passed a text with errors directly and let's see what it tells us. Here, to begin with, it is already warning us that we have made errors in spelling and typography. Then here it is asking us to divide the text into more paragraphs, so that it is easier to read. Then it is telling us that we can include more details, since this is a very short text, and the expert in dissemination and written expression, which is what we have asked ChatGPT to be in this case, considers that more details should be included. It tells us to use logical errors to improve the cohesion of the text. And here it is already writing to us directly the improved version of that same text. And as you can see, here we already have the corrected text and basically it has applied all these tips that it was giving us before, taking the role that we discussed of expert in dissemination and written expression. So I think it is very useful if it is a text that you have reacted to yourself and you want someone to improve it, since generally the texts generated by ChatGPT themselves will be fine. It makes no sense to ask someone to improve it, but if, for example, you have written your own email, you have written it quickly because you were in a hurry, you can quickly pass this prompt to ChatGPT and have it improved as an expert in email response. The next type of prompt with which we can work with ChatGPT is very useful when we have very long texts that we do not have in a document. That is, if, for example, we have a website that we cannot download and pass as a document, and yet we want to work on that text from that website, we can use this type of prompt here. Basically, what we are going to do is pass a series of sections of the text to ChatGPT. Basically, this is done like this because ChatGPT has a limit of around 800 words per message, so we cannot pass everything at once, but we are going to pass it section by section, and once we have passed all the sections, we are going to ask the question that we want to ask ChatGPT. Imagine that we have a text of five sections. We would say, I am going to pass you a text of five sections, answer only OK until you have passed all the sections. Here, for example, instead of an X, it would be a 5. Section 1D, instead of X, 5. And your section. Section 2D, 5. Section 3D, 5. And when we have passed all the sections, which here would be section 5 of 5, there we can ask him things. We can say, passing through the sections of the previous text, and your question. And then we would say, very important, quote the specific phrase where you got your answer. This has to be done like this, basically, because sometimes ChatGPT invents things. So you have to control it a bit in that sense, and if we ask him to quote the specific phrase, then we can corroborate that, indeed, it was not invented. So let's make an example. For this, we are going to grab this text here, which is basically a Lido whitepaper, which is a decentralized cryptocurrency protocol. We are not going to go into what it is, but it is something about cryptocurrencies, okay? So we are copying the text. We are going to copy it whole. We are going to copy all this. Let's see, OK. We copy. We are going to open a word counter and we are going to paste the text. And as you can see, here it tells us that there are 3,133 words. If we open a calculator and take into account that the text that ChatGPT can support in a single message is around 800 words, what we are going to calculate is how many sections we need to divide it. So we are going to open, we are going to write 3,133, the number of words, divided by the maximum number of words, which would be 3.9. So we are going to have to divide it into 3.9 sections or more. As we cannot do 3.9 sections, we are going to divide it into 4 sections or more. How do we divide those sections? Well, simply what we are going to do is duplicate this tab. We are going to duplicate it 4 times. Now what we are going to do is grab all this, more or less, until we have a quarter copied. Let's see, around here. We are going to copy a little more. Let's see, and around here. Let's see, 972. I believe that with this we can play it so that there are no problems with ChatGPT. We are going to open this other window and we are going to paste what we have left over, that is, the 2,161 that are left over. And we are going to do the same. We start at the bottom, we copy everything, up to more or less a third, around here. We cut, ok, 642. We go to a new window, paste, and now we copy up to half, more or less, around here, more or less. We cut and paste in this last window. So, once we have it divided into those four sections, which would be around 800 words, it really does not have to be exact, but 800 is a good point, let's say, orientative. Maybe we are already playing now because it is 972 and maybe it rejects the text, but we are going to try. Once we have it divided, what we are going to have to do is copy the prompt that we are showing before, this one here, and we are going to paste this prompt in chatGPT. Here, we replace this with the text, very important, as always, put it in quotation marks. And here we are going to put the first section, which would be this one that we have here. We copy, paste, and enter. And if we go down below everything, you see that here it has already put us directly, ok. Once we have done this, we go to the next section, we copy the prompt and we say, in fact, I forget here to put how many sections it was going to be. It's going to be four. Ok, now yes, we are going to give it again here to save the submit. Perfect, and now we go to the next section and we are going to put it here, section 2 of 4. We go to the second section, we copy and paste. Here it tells us ok, perfect, we go to the next section, section 3 of 4. We go to the third section, we copy this and paste it. By the way, what I'm doing to select all the text at once is done by pressing the CTRL key and the A key at the same time. And so we select everything. And then to copy, as with CTRL-C and CTRL-V of all life. If you don't know how to do it, we can also do it like this, copy. So now we have it copied, which I don't make a mistake, it's this one, yes, perfect. We hit enter. As you can see, ChatGP at all times remembers what we asked for before and is always answering us ok. That is to say, here we have already said ... where was it? I have lost it. Here, sorry, it only responds ok until all the text has passed you. And that's what it's doing, that is, we are passing the text, ok, we are passing the text, ok, we are passing the text, ok. That is, it remembers perfectly what we have told it at the beginning. And finally, we are going to pass the last section. So we are going to copy the prompt, section 4 of 4, we copy the last section, we paste it. And as you can see here, it has already detected that it was the last section and it tells us, thanks for sharing the complete information. How can I assist you with this text about Lido and its Staking Liquid Barycerium service? Ok, perfect. So now here we can ask our question. And to ask our question, well, I said it, we are going to use this prompt here that we have commented on before. Based on the sections of the previous text, our question quotes the specific phrase where you got the answer. We go back here to the text that we had already pasted, we copy the prompt, we paste it and we are going to ask it what is the governance token of this platform? And here directly it is already telling us exactly what we wanted, which is the answer. The governance token of this platform is called LDO and it tells us where it got that answer. So if we want to make a responsible use of ChatGPT and we want the information to be contrasted, what we would do here is copy the supposed quote where it tells us that it has taken the information, since I said sometimes ChatGPT invents things and you have to contrast it. We are going to copy it, we are going to go to the original document and with CTRL F to make a search, we are going to paste the text to the quote and indeed this quote is part of the original document and if we read it, it tells us exactly what ChatGPT has answered us, that the governance token of this platform, of the IDO, is the LDO. So on that side, perfect. Another very interesting prompt that we can use for ChatGPT is that of comparison, that is, from two texts, two elements, two tables, whatever, two images even, we can compare them and ask for an opinion about it or simply that we have it, for example in this case we have said in a table, in order to draw a conclusion. So the prompt for it would be something like this, I'm going to pass you two texts on the subject, answer only ok until you have passed both texts, that is, a bit the same as we did here with the subject of context. Then we will pass text 1, text 2, based on these two previous texts, compare both what we are comparing, elaborating a table, omit the information that is not present in one of the two. We basically tell each other this in case one of the texts contains information that is not compatible with the other, that it simply omits it, but we can ask for another different indication. So in this case, I am going to take up the example that we had done before of Bitcoin, in which we had passed disordered information and we had asked for a table and since we had Bitcoin and Ethereum, we are going to compare both. In this case, it is not necessary to pass the texts since one already has it here, we are going to continue on this chat, the information of Bitcoin would already have it, so we are going to ask it now to compare it with the text of Ethereum. So we are going to start writing the text, it would be ... first of all we are going to copy the text of Ethereum that I do not forget, we copy all this from Ethereum, ok? And now that we have the text of Ethereum, now we pass the prompt, we are going to use the same prompt that we had here, but we are going to modify it a bit, so I am going to pass you a text on it and we are going to put Bitcoin. Oh no, not on Bitcoin, not on Ethereum, Bitcoin is what we already have here in the chat. Ethereum, the response is only ok and with this at the moment it would be enough for us. And we put text, two points and now we pass the text on Ethereum. And as you can see here, it has responded only ok as we had asked it. And now that we have both texts, that is, both information more than texts, since here we have it in text format and here in table format, we can ask it for what we had said here, with this prompt, based on these two previous texts. Based on these two ... well, we are going to put in these two documents, because one is text and the other is table. Compare both, we are going to put, compare both cryptocurrencies, the one flying in a table, omits the information that is not present in one of the two. And we are going to hit enter. I made a mistake, the same one I had with the other account, so since this is a relatively simple operation, we are going to change to ChatsPT 3.5 in order to be able to execute, let's say, this prompt. So we are going to give it here to UseDefaultModel and here it is already doing the table. And as you can see, in this case, since it does not have too much information density and it is not very complex operations, what we are asking for, we can use the 3.5 model without problems. If what we were going to ask for involved some calculations or implied that the Python code interpreter we have seen before comes into play , in that case, maybe we should wait and use ChatsPT 4 in order to get the best possible result. And as you can see here, despite being ChatsPT 3.5, which as we mentioned is much less powerful, for this operation, which is relatively simple for ChatsPT, we have obtained the result we expect without any problem. But obviously, we will always prefer, whenever possible, to use ChatsPT 4. The next type of prompts that we are going to see are very different from everything we have seen so far. That is, so far, ChatsPT is generating text. Now we are going to see a type of prompts that are going to generate images. And you can tell me, but how can we generate images if ChatsPT is really a language module? Shouldn't it be prepared to generate images? The thing is that, as we mentioned at the beginning of this video, OpenAI has a second artificial intelligence called DALI3. And this artificial intelligence is connected to ChatsPT. That is, ChatsPT itself cannot generate images, but it will ask this second artificial intelligence to generate them. And all this will be done internally from the chat, without it being fluid for the user. That is the way to say it. So, the prompts to generate images are these here. While you ask it to generate an image, you can ask it whatever you want. These are some of the ideas, but simply ask it to generate an image. So, for example, I can ask it to generate an image of a Ferrari. And as you can see, this icon will appear here, which is a circle that is being filled in. And here it is already indicating that it is creating an image. This takes a little longer than a text, because it requires more computational power to ChatsPT. Well, ChatsPT specifically to DALI, which is the artificial intelligence that is really being executed. But as you can see, we already have it generated here. Even the logo has done quite well. Why did I comment that the logo has done quite well? Because now we will see that it has certain limitations when generating images. This is frankly good, but maybe it was not exactly what I was looking for. Maybe I wanted it to be in a different photography style, or even in an animated drawing style, or in another angle. That is why we are going to be able to give it a lot of details, so that the photo looks much more like what we are looking for. For this, I have prepared a list of tips, of things that you can ask ChatsPT, so that the photo is more of the style that you like, of what you are looking for. First of all, what we say, be specific and detailed. You can specify things like the environment, like objects that are around, or the object you want to be in the foreground. The colors, the environment, if you want it to be in a festive environment, in a dark environment. All these are things that you can ask ChatsPT. Then the atmosphere, you can also ask it a little about the feeling you want the image to convey. If, for example, it is an image for an ad that has to convey tranquility, then you can put, I want it to convey tranquility, or chaotic, or mystical, or futuristic. This depending on the result you want. Then it helps a lot to add descriptive adjectives. For example, in this case, instead of putting a dog, you can put a small, soft, brown dog. And in this way, it will get much closer to the result that you want. Then it also helps to give it a perspective and a composition. For this, we have put this link here, which is basically a website where they explain the different types of frames, that is, the different types of framing that a photo can have. And in this way, based on this information, you can ask it exactly how you want the composition to be, and how you want the framing of the photo to be, so that it gets closer to what you want. Other details that we can ask it when generating this image are the lighting of the time of day. This helps basically to specify that exactly. If we want it to be day or night, if we want it to be sunny, if we want it to rain, if we want it to be cloudy. The light, for example, if we want it to be natural sunlight, light from candles, neon lights. Another thing that we can also specify is ... Oops, my microphone is moving. Another thing that we can specify is the movement. This, for example, in the previous case, in the case of a car, if we want it to be stopped, we can ask it to be parked. If we want this movement, we can ask it to go 100 kilometers per hour, etc. Or, for example, in this case, in this example, a cat versus a cat jumping a fence. And in this way, by putting a lot more detail, we get a much more complete result and much closer to what we are looking for. The more detail, the better. In that same line, we can ask it for analogies, we can ask it, for example, to make a stereo from paint to oil, we can ask it to do it in a diagram style, although in this case we will see that it fails a bit, later we will see how it can be done to get, let's say, correct diagrams and that it does not make a mistake, because we will see that it makes a mistake in these things. We can even ask it to do it in the style of a specific camera. For example, here I have put the Canon EOS R10, but if you have, for example, a camera, you can ask it to do it in the style of a specific camera, even with a full lens. And the most important thing is that the cat understands it. Then we can also ask it for a style, as well as something more abstract, cyberpunk, art deco, minimalist, etc., in case we are not talking about a photograph. And finally, as I said, we can guide it a bit in the sense that we can iterate on the same image. That is, I can generate an image and then ask it to change some aspect. Well, taking the example here of the Ferrari photo, if I go here, I can ask it, ok, well, imagine that I don't want it to be the Ferrari, I don't want it to be a frontal photo and I also want the Ferrari to be blue. Well, I tell it, I want it to be on the side, well, that it looks, rather, on the side and that the color of the car is blue. And as you can see here, it has already finished generating the image and indeed it has made us a blue car and has made it on the side. And despite the fact that the model is not exactly the same, since it still costs a little those things, and you see here the headlights are a bit different in this photo, that in this photo the shape is quite similar. So, in fact, if I'm not mistaken, I think the model it has made is the Ferrari Superfast and if you look at it, well, the design is similar in the three photos, in the first one that we have ordered and in this second one. And all this without specifying the model later, that is, here, let's say, it has taken as a reference both the text that I have passed through here and the reference image that I had of the previous prompt, that is, it is based on both the image and the text, and that is why it is cool when iterating on the same image, since we are not only contemplating our text, but we are also contemplating the previous image. And to close with this the tips, I have also prepared a prompt here so that you can see the difference of what would be a more worked prompt based on all those tips. For example, I do not want a modern Ferrari, but I want a pink head Ferrari, so I have prepared this prompt here. Imagine a pink head Ferrari from 1984, one is impeccable, with its iconic red color, standing out under the warm sun of the sunset. This classic vehicle is parked near a cobbled street, and so on. Basically, it prepares a much more complex prompt, also with the help of the chatGPT itself, and if we now hit enter, it will also generate another image. We would already have it around here, and as you can see, it has respected perfectly everything that we have asked for. We have raised a cobbled street, here we have it, with cobblestones, it is not an asphalt street, we have raised a pink head Ferrari, it has done it perfect, in a red color, in the sunset, basically it has done everything as we have asked for in the prompt. So, as I said, by detailing our prompt much more, based on these 10 tips that we have given you here, we can get the usual, a much more precise result, and much more concrete, and much more in line with what we are looking for. Then also, something very important to emphasize, regarding chatGPT and regarding the images that it is going to generate through DALI3, which is this artificial intelligence of images that is connected to chatGPT, is that the texts that it is going to generate are generally going to have faults, it is going to have problems, as there are applied letters, etc. A bit like what happens in this image that I have put here as an example. It is something in which to this day chatGPT still fails, and in fact almost all the artificial intelligences that generate images fail, and it is that many times they are going to generate things that look like text, but that are not text, that have errors. And sometimes, if the text is quite large, it can more or less manage to generate it well, but the smaller the text, the more characters there are, the more chances there are that they will fail, and this is something in which it fails a lot. Then the consistency between images also fails, that is, two images with the same prompt will be radically different. This can be seen very clearly if now this same prompt that I asked it to generate an image of a Ferrari, I ask it again. So we already have it here, and as you can see, this image looks like a video game, in fact it has generated two of us. Sometimes this happens, that chatGPT decides to generate more than one of us, and that seems to have much less detail, the colors are more vivid, it looks a bit like plasticine, and that the prompt was exactly the same as we used at the beginning, to generate an image of a Ferrari. But if we go back to the original image, we see that it is much more detailed, it would look a bit like a real photo, very edited, but a real photo for an ad, for a magazine, for something like that. Then another thing in which chatGPT fails, and a bit also in the same line as the texts, is what we mentioned before, which is that it fails if we ask it to draw diagrams. For example, we ask it to draw a plan of a building, things like that, technical drawings that require specific measures, right angles, proportions, etc. They usually fail a lot, that is, they usually give much better organic shapes or invented objects, rather than a technical drawing with specific measures, since, as with the texts, it fails a bit in that sense. With this we would have finished the prompts section, and we are going to enter a new section of chatGPT, which would be to interact with files. And that is a section, in my opinion, very interesting, since it opens up a whole range of possibilities, much more than just writing text. And it is that in chatGPT, apart from writing text, we can also upload files. And now we are going to see what types of files can be uploaded, and how we can interact with those files that we are going to upload. The first type of files that we can upload are ... we are going to classify them as documents. They would be PDFs, PowerPoints, Docx, which are basically words, and archivos.txt, basically text documents. That would include you in this category. Then, once we add them, we will be able to ask questions, such as what we said before. A question and, again, quote the specific phrase where you got your answer. A bit like the same structure that we used before in the context prompt section, basically to make sure that we are not getting it wrong. So, to do this test, we are going to do exactly the same that we have done before with Lido, but this time, instead of passing the document by copying and pasting, which is what I was telling you before, that there was a more optimal way to do it, what we are going to do is download it here. We are going to click on download, we save it, it will be saved here, we go to chatGPT, and directly, dragging it from here, from where we just downloaded it, or if not, we can also drag it from a folder, that is, if we have it in a download folder, we grab it here and drag it here. We could interact with it, with this same prompt that we put here, of question, and then quote the specific phrase where you got your answer. So, we are going to grab this prompt itself and we are going to ask it the same thing, what we have asked before, which is the governance token of Lido? And it should give us the same answer as before, or a very similar answer. And as you can see here, it has already given us the answer, it has told us the same as before, it has told us that the governance token is LDO, and the phrase that it gives us for the answer is a different one, that if we copy it here and paste it, we should find it in the PDF, but we are not finding it. This can be due to two things, it can be that chatGPT has invented it, this happens sometimes, or it can also be that it has not copied it well from the chatGPT PDF and that it really extracted the information from here. So, to make sure, what we are going to do is copy a fragment, we are going to simply look for this fragment, and indeed, if we search here and copy this, we see that the phrase seems to be the same. To have a vote in Lido, one must hold its governance token LDO. To have a vote in Lido, in the Lido DAO, one must hold its governance token LDO. What happened here is that it has not copied the commas and that is why we are not finding it in the PDF, but it has indeed extracted the information from the PDF correctly, so in this case, again, chatGPT has done it quite well. And this, as I said, we can do it with PDFs, we can do it with PowerPoints, with Word documents, or we can do it with .txt documents. In the same way that we have done this with documents, we can also do it with tables, we can do it with CSV tables or with XLSX tables, which are the Excel tables. So let's see an example of this. For this example, we have opened a new chat and what I am going to do is that I am going to pass it a table, that is, a document as we count, in this case, CSV. And what this document contains, which I am going to open here so that you can see it, is a lot of data regarding Bitcoin prices, that is, these are the dates of the different days and these are the prices of Bitcoin at that date, that is, on January 4, 2023, Bitcoin was worth this, the maximum of that day was this, the minimum was this, and close this, and adjacent close, I don't know what this is, really, and this is the volume. Basically, they are a lot of data about the price of Bitcoin on the different days, as if they were the data for the general graphic logo. So, with these data, for example, I could ask for the following, which I had prepared here. Using the data from this table, tell me how you would have obtained more performance, making DCA with the same amount every day, or buying the first day and holding continuously. Basically, we are asking you how we would get more money, if we had done dollar cost averaging, which basically consists of making regular purchases of a fixed amount of money, or if on day 1 we had bought everything at once, without making subsequent purchases, and we would have waited until the last day, of which we have data in this table. So, we are going to hit enter, and here, as you can see, the icon of the code interpreter is appearing again. This is basically because it is a calculation that ChatGPTN cannot do with its Deep Learning engine, with its internal AI, and what it is going to do is read the table and make the calculations thanks to the Python integrated code interpreter. That is, it is going to make a code to do all those calculations, and it is going to execute it in order to obtain a reliable result, which is what it is doing here. If we open this here, we see that the code is actually biting. What happens is that, by default, it appears closed. By default, it does not appear here, but here we see that it is actually biting a Python code in order to make that calculation. Ok, it has already finished generating. Basically, it has made this calculation starting from the basis that you are going to invest $100 a day, and it has made the share with the same amount of money, but I have been noticing, and this is why it is important to always review the results that ChatGPTN generates, which says that by doing DCA we have obtained $75,000 and by buying on day 1 we have obtained $241. This, I think, is because it has been wrong, and what it has calculated has been buying $100 a day versus buying $100 on day 1 and not buying again. And I think that is why it has been wrong in its calculations, since they should be much more even results. So I'm going to tell you, and let's see what calculation it gives us now. Basically, what I have told you is that it is impossible that there is such a difference between DCA and buying and maintaining. I think you are wrong, and different amounts of money have been invested in total for each strategy. So we are going to say this and see who responds. And indeed, here it is telling us what we had, that it had calculated to invest $100 only on the first day versus investing $100 every day. So, obviously, by investing $100 every day, you are going to earn a lot more money. That is why now it is doing the correct calculation, which would be, what would be the difference in profitability by investing $100 every day or by investing the equivalent on day 1 and doing the corresponding calculation. And here it has finished generating the answer, and it tells us the following. The total investment for both strategies was $36,700. Ok, so it has used the same amount for both strategies. This is because it has used the data for a year and surely there were a couple of more days. One year has 365 days, surely in Excel there were 367, and that is why 100 times 367 is 36,700. That's why I chose this amount. So far, everything fits me. Then, Dollar Cost Averaging tells us that the second performance to this amount of $75,000 versus buying and keeping the second performance of $88,699. That is, with this analysis that it has done, we see that last year we would have made more profit if we had bought on the first day and we had kept instead of doing DCA, instead of making regular purchases. This makes sense because this last year Bitcoin has gone up a lot and normally the DCA is a more conservative strategy that reduces profits but also reduces losses. So, perfect, we have already seen how we can analyze some data from an Excel, which would be much more complicated to analyze if we wanted to do it manually, and ChatGPT has done it for us through its code interpreter, as we are going to see here. Another type of file with which we can interact in ChatGPT are images. That is, not only can we generate images, but we can pass reference images and ask or ask things about those images. For example, we can interact with JPG images or PNG images. It is possible that in the future there will be more types of files with which we can interact. Specifically, two quite interesting uses that we can give are those of identifying objects or understanding concepts. Let's open ChatGPT here, let's open a new chat, and let's imagine that I have the image of this car here, and I don't know what car it is and I want to know what car it is. We can copy the image. In this case, we don't even need to download it, we can simply copy it, paste it, and say, tell me the model and brand of this car. In this case, and also so that you can see that ChatGPT is not perfect, it has been wrong, since it has told us that it is a Mazda RX-7, when this is really a Toyota Celica. That's why we always have to contrast the information that ChatGPT gives us, and in this case it would be as simple as copying this information, looking for it, and we see that it is not the same car. This car is really a Toyota Celica MK2. So, we have to take that into account. It is true that both are Japanese cars, but it is not perfect. Still, it is something that usually works, and for that we are going to put another example, and we are going to say with this mobile, tell me the model and brand of this mobile. And in this case, it seems that I have done it right, if I remember correctly, since the model and brand of this mobile, if we look for it here in Google, in images, is this one here, Nokia 3310. That is, in this case, here in fact we have the same photo that I had used, here it has achieved the result that we expected, which was to identify what this object was. So, for example, if we want to identify a product, if we want to identify what a product is for, things like that, we can pass the image and identify what this object corresponds to, even much more complex things, sometimes it is able to identify them. Specifically, we are going to do the test with this here, it is a photo of a mountain, and if we ask it what mountain is this, it is telling us that it is the Montaña de Montserrat, located in Catalonia, Spain, and it is famous for its unique rock formations, etc., etc., which we would not know very well how ... well, we can look for it, yes, yes, we look for Montaña de Montserrat, and indeed we are talking about this image. And here, as you can see, it has managed to correctly locate this mountain from an image, that is, it does have some fairly developed capabilities, but as I said, you always have to check, since, as you can see in this case, it had been wrong. Another quite interesting thing that we can do with images is not so much to ask it to identify different objects, mountains, photographs, whatever, but we can ask it, for example, to identify what the functionality of an object or an image is. For example, I have here this image of a chronograph, you can pass it through here, and I can say, tell me what each of the needles of this clock are for. And as you can see here, it has already found the answer, it tells us that the long central needle is used to measure the milliseconds of the chronograph, it is correct, it tells us that the two needles in the center are for the hour and the minutes of the time, it is correct, and here it is reading, and here there was a small error, and it tells us that the needle that goes up the left sphere is the one that ... ah no, sorry, yes, I did it right, the one that measures the seconds, the upper one, the problem has been here, which tells us that the upper needle, when it is really on the right, is the one that measures the milliseconds, and the needle of the lower sphere is the one that measures the hours, that is, despite this small error, which has marked it as superior when it is really on the right, everything else has done it well, and it has correctly identified, let's say, how this object works and what its characteristics will be, and how each of its sections works, so in that sense, another quite interesting thing that we can do with this, we can also do it, for example, if we pass a plane and ask it to identify elements in a plane, things like that, very interesting, and the truth, a fairly powerful tool. In the next section of this course, we are going to deal with the issue of Google Chrome extensions, which now you will see that it has to do with ChatGT. Basically, there is a type of program, so to speak, called Chrome Plugins, that is, so to speak, Google Chrome extensions, which what they do is that, unlike other types of programs, they are installed from your browser and add extra features to interact with the different pages, it is not always like that, sometimes they are, let's say, stand-alone programs that do not interact with the content of the browser, but usually it is like that, since that is like its main function and the grace of installing programs inside your browser. So, within this category of programs, there are some in particular that what they do is connect with ChatGT to offer extra features or to simplify your life, and they usually open in a small window next to your browser. These plugins or these extensions are installed from this website here, which is the Chrome Web Store, and to find them it is very simple, for example, there is a very famous one of these that work with ChatGT, which is called Cider, which in fact is the one that has the most downloads, we can look for it by this bar here, and as you can see here, if we open this photo, what it does is that it opens this window here with the ChatGT features next to it, and some extra features, of these there are many more, we also have, for example, Monica, who does something similar. We also have SearchGPT, which in the same way also opens a window on the side and allows you to interact with the pages, but my favorite and the one that I can recommend, since it has been the one that I have tried and the one that I have been using for a long time, in fact it is the one that I use in my day to day, is Harpa AI. Something important to emphasize about this plugin is that I have found situations in which it creates conflict with ChatGT, that is, as a result of installing this plugin, perhaps ChatGT in itself stops working and errors appear, to solve this error it is very simple, you can install it with a couple of clicks as you will see now, but hey, let's install it first so you can see how the installation process is, we are simply going to click on this button here to add to Chrome, we give it to add extension, and with this it would automatically be installed. So from here we can create an account, which in my case I think is already logged automatically since I have already installed it previously, but if you can't, from here you can do all the management to log your account. Then something important to keep in mind is that to unlock all the functionalities you will generally have to pay, I specifically have the cheapest plan, which if I remember correctly is $ 10 a month, let's see, let's see it there, if we change it, yes, 14 euros a month, not 10, but 14, but personally I think it is worth it since it offers many extra functionalities, some of these functionalities are, for example, the next one that I am going to show you now. We can interact using the chatGPT functionalities in YouTube videos, in this case, for example, it is a video in which this guy teaches 19 watches and what we can do is go to this bar here on the side, which is the JARPA, or if not, we can also open it from here, which in fact if it does not appear up here with the other plugins, what you can also do is give it this plug-in and in this way it appears here, and by opening this and clicking here on the chat with video button, because it has detected that we have a video here, we can ask for things about the video. Here by default it tells us that it is in chatGPT 3.5 but we can change it to 4 and I could ask for example, tell me the 19 watches that appear in this video along with their main feature. Even if I don't want it to take too long, I can tell it to do it in less than 5 words for each of them. And right now what this tool is going to do is that it is going to interact directly with the video, in fact what it does, a little more on a technical level but this is not very important that you know, what it really does is read the transcription, since almost all YouTube videos have transcription and the tool is not really seeing the video as such, but what it does is read this text here, that is, it is not seeing the video as such, but still in practice the result is the same. So if we open this again, it has already done exactly what we asked for, it has looked at the video, in quotes, since the transcription has been read, and it is already saying, well, the first is this universal size and compact, the second is this and so on, and if we go to each of them, for example, we start here, we see that indeed the first is the G-Shock, the second if we advance the video a little more is the Seiko Compass, which is the one that puts us here, etc. So it has done the literally work that we have asked for, to go one by one, here there are other different features but I understand that because it must be surely a mention, well, no, this seems to have skipped, as you can see, it is not perfect, it is something to take into account, but then the next one has already started to do it well, the next one to the Grand Seiko and here I know that it has done it well, then, as always, with the AI you have to go carefully, since in this case one has skipped, but as a general rule, it has done it quite well and the truth is that it is a very useful tool. Then, if in this same window, where we have given Chat with Video to make them ask about the video, we click on All Commands, we will be able to see a list of all the commands that the tool offers, which, the truth, are a barbarity, it has to generate keywords, to generate article descriptions, SEO focus, we have many commands and in this case we only talk about the commands related to the interaction with video, since if we go, for example, to this page that I have opened here, and we open the same option, that is, the option to see all the commands, we see that they are, for example, all the commands that are related to the interaction with video, since if we go, for example, to this page that I have opened here, and we open the same option, that is, the option to see all the commands, we see that they are, for example, all the commands that are related to the interaction with video, since if we go, for example, to this page that I have opened here we see that they are practically infinite, data extraction, content analysis, clickbait text analysis, we have many types of commands with which we can interact with different web pages using artificial intelligence, so this opens up a world of possibilities, the truth. And well, this is a slightly more advanced tool, also directly related to ChatGPT that I wanted to show you, but we are going to return to the focus of the course, since we have many things to see before entering into something much more complex such as this. Anyway, small parenthesis, I want to warn you that this tool has not paid us, nor is it a correlation, nor anything, I just wanted to show it to you so that you can see to what extent the capabilities of ChatGPT can be extended using plugins of this style, I have said it, this is not the only one, there are many more, I have simply shown you this one, I have shown you Harpa, because it is the one I use every day. Anyway, we are going to return to the focus of the course, we are going to return to ChatGPT, since this would be a much more advanced level and it would give us for another whole video, so let's continue with the next block. In this other block we are going to talk about how to customize your ChatGPT, that is, you can set parameters to your ChatGPT, response parameters, so that it responds based on those parameters in a consistent way and without having to enter them through chat. Let's see how that is done, we go back to ChatGPT, we go to Customize ChatGPT, which would be Customize ChatGPT, here it gives us some small instructions on how it works, we click on OK, and basically here we can introduce instructions on how we want to customize our ChatGPT. So, we are going to go back to the presentation for a moment and basically we are going to say two things. In this first block, what we are going to say are things about you, that is, you can say, for example, where you are from, what you work, what you have studied, what are your hobbies, what are your domains of knowledge, that is, maybe I have studied engineering, but I am very good at art history. Well, if I know a lot about art, I can comment on the ChatGPT and take it into account in terms of the answers it gives me. You can even tell it what your goals are and in this way you can guide it a bit. If your goal, for example, is to learn, then maybe it will give you much more developed answers so that you can enter new knowledge. If your goal is more pragmatic and you want to go directly to make money, then maybe it will give you shorter answers with bullet points. All these will be considerations that ChatGPT will have in terms of generating your answers. Then, in this other block below, what it will ask you is how you want it to answer you. It will ask you things about how you want it to answer you. Basically, you have to say if you want it to answer you formally or informally, if you want the format of the answer to be longer or shorter, if, for example, what we said before about bullet points, you can ask it to answer you all by bullet points, how you want it to address you, you can even tell it what your name is so that it addresses you with your name, and you can also ask it if you want it to have a neutral opinion or if you want it to give its own opinion. In the end, this is a program that is based on millions of data, so it will be based on millions of opinions, but basically you can guide it a little in that sense and many more things. Everything you can think of, how you want it to answer you, if you want it to answer you by insulting you, I don't know if it will do it, but maybe you can ask it too. Having said that, we are going to fill in these sections. For this, what I would recommend is that you write it in English and this is something that generally applies to all the interactions that you are going to do with ChatGPT. In the end, you have to think that it has been trained with millions of data, among those millions of data there is a lot of text and a lot of information in Spanish, but there will be a lot more in English. So, taking that into account, what you are going to achieve if you do all your interactions and, for example, this specific configuration in English, is that the quality of the answers, based on the understanding that ChatGPT will have of what you are saying, will be much better. So, in this case, I have prepared a couple of texts, one for each section, I have prepared them in Spanish and I am going to show you another completely free tool, although we can do it, as I said before, through ChatGPT itself, but I am going to show you another free tool, we use this basically because it is faster than ChatGPT, which allows us to make contextual translations. So, we are going to take this out, here we paste EXO, which is the text that I have prepared, and it basically makes us the translation of this first block, which we have to talk to ChatGPT about. In this case, I have said, I am from Spain, I work for the largest YouTube channel in the niche of entrepreneurship, finance and ways to make money, our goal is to make the best videos possible, and that means both interesting and informative videos, as well as making content that is useful for our viewers, etc., etc., etc. So, we are going to copy this, the translation in English, and we are going to paste it here, and we are going to remove this here, which is something that the translator has generated. And the same with the second part of the text, which I have also prepared in advance. We are going to paste it here, and here I have said, I want your answers to be informal but technical, I want them to be rather brief, adapt your explanation to a format of vignettes, I translated this wrong before, oh well, but I translated it well, it is that initially I had written it in English, to a format of short bullet points, etc., etc. So, we copy this, we paste it here, here I see that I have not put the other message, and with this it would be, we are going to save it, and having this saved, if now I ask ChatGPT where I am from, since it was one of the information that we had specified, let's see, here it is not, here, CustomizeChatGPT, if I ask where I am from, which is one of the information that I had specified here, in principle, where do you think I am from? Very well, and here it tells us that I am from Spain, and if I ask it where I work, it is already taking it into account, and if you look at it, it is also making short answers, and these are very direct questions, but what ChatGPT does now that it has constants and information is to take it into account for future answers, so maybe now if I ask YouTube, in fact, if I ask what YouTube is, in fact, we are going to do the test, which is YouTube, I understand that in principle they will take it into account in their answer, because knowing that I work on YouTube, it will give me a much more technical or much deeper answer, or maybe simply, well, look, in fact, it is giving me the specific explanation, but in this case it has already used the bullet points format that we had asked here in the customization of ChatGPT, which we had asked it to make its answers in short formats, with the essential information, and using incomplete sentences to make it fast to read, that is, in this case it has respected it, as you can see, I have said, it will not always respect everything, it will also use a little of its criteria, because in this case, for example, it did not make sense to answer in bullet points, but already with this information we will be able to make a much more optimal use of ChatGPT, since again we will be guiding it much more, and we are going to save each of the iterations and each of the new chats that we will have to pass all this information again. The next block that we are going to see is one of the functionalities of the moment, so to speak, of ChatGPT, it is an update that arrived recently and that completely changes the rules of the game, they are the Custom GPTs, they are chats of ChatGPT specialized in a specific task or in a specific domain of information. Now we will see in more detail what all this is about. Basically, they are chats created by the community, here you have a screenshot with some examples, specialized in different areas, all this is achieved by training and passing information to all these chats, that is, the same thing we have done before dragging a chat to the document, it can be done in a pre-constructed way, pass a lot of information so that that chat responds in a certain way. So you have to take into account some aspects of all this, the first of all is that it is private, even if there is a person who has created that ChatGPT, they will not have access to your information, nor will they have access to the history of chats that you are having with that customized ChatGPT. Then, their uses are specific, even though you can ask questions about cooking to a sports ChatGPT, the idea is that you use that ChatGPT for what it is designed for, since it is specially designed for it and will achieve much better results in that sector for which it is designed. Then another very important aspect is that APIs can be integrated, which is an API for those who do not know what it is, summarizing it a lot, it is a service that offers certain web pages so that that web page can be connected to other web pages. So, for example, the YouTube API allows you to connect a code so that you, from your code, can upload the videos directly to YouTube without having to go through the interface and click manually, that would be an example of an API. Or maybe the Nike API is an API that allows you to collect the prices of all the shoes live and then you can set up a web page where all the prices of Nike shoes come out live without having to go one by one copying. I think Nike does not have an API specifically, but to give us an idea, that would be the example. So this allows you to do very cool things that you normally could not do with ChatGPT, that is, imagine you can grab an API from TradingView, for example, or from any service that lists the prices of shares and assets and make a ChatGPT that has real-time consistency of the prices of the different shares. That would be, for example, one of the possibilities. Another very important aspect of these custom GPTs is the ease of creation, and it is that you yourself can create your own custom GPT adapted to your needs. Imagine that you are a researcher and you are developing a study on a very specific topic and that information only you have it and no one else has it, because there is not going to be a custom GPT about what you are looking for, but you can upload your documentation and in this way you get an AI trained in a certain way. It is not exactly training, but it is more that it has the context of that information so that you can ask questions about that specific thing of which there is no AI specialized in it. So, having said this and with the hype that you can create your own custom GPT, let's see how that works. But before seeing how to create our own custom GPT, let's see how you can use the GPTs that the rest of the people have created, the GPTs that the community has created, that is, these that we are seeing here. You can see if we go here to ChatGPT and we go to this tab where it says Explore GPTs. From this tab we can see several categories. We have the categories of Dalí, which would be the GPTs prepared to generate photos. For example, if we click here to see more, we see that there are many more. We have those of writing, we have productivity, analysis and search, programming, we have a lot of variety. And not only that, because very few appear here, but we can look for many more GPTs. That is, imagine that you live in Brazil and you want to know things about your country. You can look for a GPT from Brazil. If we write here Brazil, we will get a lawyer from Brazil, a legal consultant from Brazil, an IA communicator from Brazil. And well, different things they put here that I don't understand because I don't speak Portuguese. But I said it. And if we want to use any of these, it's as simple as we want to use this one here so that it writes for us. We open it, we click on Start Chat. And we open the Custom GPT with its different functionalities. We can ask it to write an analysis of a thousand words, three thousand, etc. All this we will see in more depth in the block of how to earn money, since there we are going to review some of the GPTs that have seemed most useful to me, especially in the field of this channel, in the field of how to generate income or how to enhance the income that you are already generating. So now let's continue with the part of how to create your own Custom GPT. The first thing you will have to give your Custom GPT is the name of the description and the logo. All this is nonsense that you can put whatever you want, it is not very relevant. Then something that starts to be quite important are the instructions, that is, how do you want that GPT to act, that GPT chat. Another thing that, like the name, the description and the logo, in principle are not very relevant, this depends on the use you want to give it, are the start instructions. You can give it predetermined phrases so that people can click. As we saw here at the beginning that tells us, plan a trip, make me a brainstorm of names, write an email, all these are the phrases that you can customize to appear in your Custom GPT. Then you can pass documents, this is the most important part, that is, you can introduce knowledge to your Custom GPT, this can be the same documents that you would upload here, images, PDFs, etc., can be uploaded to your Custom GPT. And finally, you can add new features with what we talked about connecting it with APIs. Having said this and explained, let's see how this would come out. What we will do is go to our profile, we will click on My GPTs, which will be the section where we can create a new GPT, we will click on Create a GPT, and here we can do two things, we can configure it through chat, which I personally do not prefer because it will be much less precise, or we can configure it from the configuration tab. The first thing then will be to put a logo, for that we are going to click here, we can even use DALI directly to generate the logo, we are going to try with that. And of course, I think it is using this information, it is based on this to generate the photo, so now it generates a random photo or it does not generate anything since it has no information. We are going to leave this here, but I understand that if I, for example, put the title of our GPT, which would be this, and now if I click on Use DALI, I understand that it will do something related to this. Then we are going to name it CryptoAnalysisGPT, since it will be a crypto analysis bot in Spanish. Then, as a description, we are going to put Analyze the price of a crypto currency, since it will be its main function. And then in instructions, again, we are going to put them in English, as we have done with the configuration of our GPT, look, it has generated the same photo, so we are going to give it back, maybe use this as a prompt, but hey, this is the logo anyway, it is not that important. But hey, let's continue, at the level of instructions, as we have done when we have customized our GPT, when we have customized it for us, the interesting thing here is to put them in English, since it will have a better response. So, we are going to paste these instructions here. Ok, well, it is putting things that have nothing to do with it. We are going to leave it like this. And basically we have said, always answer in Spanish, use your knowledge to answer technical analysis and price prediction answers. Do not tell me that investments are dangerous, this is the good thing, because many times when you ask for investment advice, the GPT does a lot of emphasis on, these investments can be dangerous, take this into account, it is very true and you have to take it into account, but to save us all that text of which we are already aware, since as we have seen throughout this video, sometimes the GPT fails and we cannot trust 100% of it, we have not asked it to save it and not give us this warning. I am telling it that I already know and that it only does analysis and gives me a predicted result. I also ask it to use different indicators in our analysis, discarding those that do not show strong evidence, that it gives me a single prediction, both if the price goes up and if the price goes down, and with a percentage of confidence. That is, it tells me, for example, the price will go up with 70% confidence, etc. And finally, I ask it here to base its analysis only on price variations, that is, that it does not look at factors such as volume, etc. This is to make it easier, but we could really put a lot more data. So, in the face of these interactions, how are we going to do it? The first thing we have to upload is documentation, because the GPT has technical knowledge of analysis, but the more information it has, the better. So what I have done has been to download several PDFs from several books of technical analysis on the Internet, which I have here, let's see where they are. Here they are. I have downloaded several famous books of technical analysis, these here, and basically we are going to drag them here. So now it is going to upload all these PDFs. If we open them, we see here that they are very dense books on technical analysis, which explain different concepts, etc. All this is the knowledge that the GPT uses, this one in particular is quite short, it is 67 pages with many drawings, but then we have some much more dense, like this one here, of 500 pages of text. And all this will be the information that the GPT will have in mind when doing its analysis. We would have them all uploaded here, and another thing that we have to do, very important, is to activate the code interpreter. This is what we have said before, that a window will open and Python code starts to run. In this particular case, it will not always be like this, but in this particular case that we are going to ask it to analyze certain data and that the numbers have a very important relevance, it is important to activate the code interpreter, otherwise it will not be able to do the calculations correctly. And then what we can deactivate, for example, is the DALI image generator, because in this case we will not need to generate images. I have done all this and now we would have everything ready. Here we can enter the questions that we mentioned before, for example, if I put hello here, it does not appear because it is loading, I think. But basically here the questions that we have taught before will appear when there is a GPT chat. And in this case, in principle, we do not need it, so we are not going to put it. And with all this done, from additional settings we could add everything that is connecting with APIs. In this case, it would be to make a much more technical video and we will not go into detail, but we could do all that we have told you before, connecting it with APIs, which we remember are those tools that allow interactions with different pages through code. What we have said before, upload videos through the YouTube API, you could connect and interact from here, etc. So, having said that, in our case we are not going to put it and we are going to give it simply to see where the create button is here. So, we can make it available either for anyone who has the link to this GPT, since with a GPT you can generate a link, or only for me, or it can be published in the GPT Store, which is that dashboard that we have seen before where the different GPTs appeared. In this case I am going to put anyone with the link, but with this example it is a little different. And we already have our CryptoAnalysisGPT created here. As you can see, the name comes out here and it also appears here in the sidebar and here in the middle. And this is the customized GPT that we have just created with all that technical asset analysis documentation that we have passed. So, if we want to start using it, what we could do is, for example, here I have a document that I am going to show you now, which contains all the historical prices of the last year of Ethereum. Well, I could get this document, I drag it, let's see, let's close it here, I drag it, let's see, it doesn't close, ok, now yes, I drag it in my GPT and I can ask it, based on these data, let's see, based on these data, tell me your projection for Ethereum. And we can predict it in the next three days, in the next week and in the next month. All this utilizing your technical analysis knowledge. Ok, here it is telling us what will be the indicators in which it will be fixed. It will use mobile means, the Relative Strength Index, RSI, this is quite famous, the NCAD, Moving Average Convergence Divergence, supports and resistances and candle patterns. So, based on all this information, it will do its analysis. Here you see that it is already writing code to do that analysis. Well, specifically here it is opening the file, more than writing code, which the Python code interpreter uses to open that file. Here it seems that what it has done has been to look at the price of the last five days. So, it is taking this into account when doing its analysis. And it would have finished its projection, so we see that it has done exactly what we asked for, a projection for the next three days, another for the week and another for the month. And at the end of it all, it has also done what this GPT chat is really programmed for, which if you remember, we had asked it to give us a percentage of confidence regarding its result and it has done exactly that. It has told us that it projects a short-term alzista trend and with a confidence of 60%. Still, remember, this is something you can do to experiment and that you can do to leverage your analysis, but it is not something reliable. That is, as we have seen before, with much simpler things, it has failed and in fact, in this case, it seems that it has only taken as a reference five days at the value level when doing its analysis. I don't know if it has used more values, but that's what it seems. So, taking all these data into account, it is an interesting experiment to create our GPT chat for this video example, but it would not be the only thing, let's say, on which I would base my analysis if it depended on my money. Still, I have said it, if you now want to create your custom GPT, you know how to do it. Let's go to the next block. Another very interesting thing that we can do with GPT chat is to interact with it by voice. That is, we all know the famous Siri or Alexa, which are those devices or those artificial intelligences with which you can interact by speaking, but they have, so to speak, predetermined answers. It is true that based on what you ask it, you can ask it in different ways and it will almost always answer you well, but in general they are predetermined answers with very limited robotic voices, etc. In this case, what we are going to do is that we will be able to speak to GPT chat and that the same answer that it would normally give us if we had written it, it will also give it back to us, also by voice. So this, where I think it makes more sense, is to do it through the mobile, since it is a situation in which maybe you are on the street and you want to ask a quick question and you don't have a laptop to take out and start writing in GPT chat, and that's why we're going to do it from the mobile, and for that we will need to download the app. The GPT chat app, the official app, let's say, because there are many that are not official, can be downloaded for both iOS, that is, for Apple devices such as an iPhone or an iPad, as well as for Android, for tablets and mobiles. So depending on the device you have, you are going to download it on any of these devices. In my case, I have an Android, so I'm going to show you from Android. Ok, so once we are inside the GPT chat app, the first thing we can do is go up to the menu, up to the left, and if we click here on the three dots to configure, and we go to voice, we can change the voice tone that we want. We are going to turn up the volume so that we can hear, but basically ... As you can see here, we have different voice options, I'm going to give it this one, this is not very relevant. And then from here, from the main page, let's say where we would normally write, instead of writing, we simply click on this little icon here on one of the headphones and we can talk to GPT chat. We wait a little ... Hello GPT chat, what do you think about Bitcoin? Do you think it is a good investment?

Speaker 3: It is important to do an exhaustive investigation and consider your risk tolerance and your investment goals before deciding to invest in Bitcoin or any other asset. Do you have experience ...

Speaker 2: Ok, let's cut it now, I've been talking for a long time, but basically you have seen that it works 90% well. Sometimes the audio is cut and although it has an English accent, and I understand that this will come in future updates, because what it is doing is reading it in Spanish with an English accent, we can say that it works quite well and it can be quite useful. Imagine that you are on the street, you do not want to type, you do not want to stop to read an answer, you put on your headphones, you ask something to GPT chat and you continue with your routine, with your displacement, with whatever you are doing. Another quite interesting tool that we can do from the app, which would be exactly the same as we have on the web, but on the mobile. Another very important and very interesting aspect of GPT chat, and in fact it is what all those applications that we taught before, such as Harpa, use, is the GPT chat API. Basically, as before we put the example that YouTube has an API and through its API, through a code, you can upload videos and you can connect that API, in the case that we put before with the custom GPTs, because GPT chat at the same time has an API that can be connected with third-party services. That is, from an application, either programmed by another or programmed by yourself, you can connect your account, so to speak, in the form of a GPT chat. I don't know if you understand an account, since it works with some things called API keys, but basically it would be that. So this would be for people a little more advanced, since usually most people, unless you are programmers or have programming skills, you are not going to touch this, but I think it is important to review it, even if it is a little above, in a theoretical way, so that you understand the capacity that this aspect of GPT chat offers. So, as I said, we are not going to show it in detail, since for this we would have to do another programming course within this course, but I am going to give you an example so that you can see how far you can go. A few years ago I prepared this project, in fact I think I showed it in another video, which is basically an automated YouTube channel, that is, a YouTube channel that uploads videos alone. Much more automated than the automated YouTube channels that you see on TikTok or in different people who promote this type of business model, since this channel what it did was upload videos 100% automatically, that is, we gave it a button and it automatically looked for the news of the day, took out the script from those news, generated the thumbnail and uploaded the video 24 hours a day, working non-stop. As you can see, some of the videos worked more or less, this one reached 154 visits, but usually they got few visits, that's why I have it on standby and if I have time at some point I will resume it. But basically I wanted to show you this example because I used the chatgpt API and I used it largely within the code, it is true that I touched on other aspects, but I have made a small scheme of what aspects the chatgpt API influenced, so that you can see the importance of using this type of tool as long as you have the knowledge for it. Basically what the code did was extract the news from the internet, then convert those news to a script, for this I used the chatgpt API, that is, all this that you have in green were processes for which the chatgpt API was used, then that script extracted the representative words, that is, if the phrase said today bitcoin has increased by 50%, what it did was extract the word bitcoin and use that word bitcoin to later download videos related to bitcoin. As you can see here, if we open this here, it was putting stock videos as it was talking and changing the different stock videos throughout the video. Then it also generated the voice from that script that we had previously generated with chatgpt, joined those stock videos with the script, generated headers like this, generated headers that were also generated using chatgpt from the script previously generated with chatgpt, generated a title, description and labels, that is, all the videos had a title, a description and hashtags and labels also generated with chatgpt. Then I put everything together, I chose the copy, that is, the text, the photo and the crypto for the thumbnail, because as you can see here I had some pre-finished photos and what I asked chatgpt based on the emotion of the video was to choose a photo or another or choose a cryptocurrency or another based on the theme of the video and the same for the text. And finally, I put all that together and uploaded the video. So for all these steps that you see in green, which are as you see half, I used chatgpt, then I used some other intelligence to generate the voice, the youtube API that we commented on before to upload the video and all this was code without any API, that is, basic bytecode. But as you can see, you can get very interesting projects with chatgpt and quite lucrative if we apply the programming knowledge and use this API. And we are approaching the end of this course, but we actually enter one of the most interesting parts, which is the section of how to make money with chatgpt. So in this section we are going to review both ways in which you can increase your income sources, your earnings, as well as sources or ways, rather, to generate new income sources. We will review both for the conventional sector, that is, if you are employed, as well as some freelance opportunities or business opportunities that exist with this new chatgpt tool. You will see that in most cases we are not talking about giving a button and starting to make money, but what we are going to see are ways in which you can leverage your skills and existing ones, that is, in which you can enhance your skills and existing ones using different custom gpts, using different commands of the chatgpt itself. That said, let's start with professional employment and growth. And it is that within chatgpt there are different tools that can facilitate many of the processes within our work, either to optimize our workflow within the company or to get new opportunities. In this case, for example, something quite useful is that we can do in that sense is to answer emails, that is, with commands like this, answer this email in a tone and we would introduce the tone here, for example, friendly or formal or serious, any of these tones, angry. I want your answer to contain the following information and we tell you the information that we want the answer to contain. In this case, it will be something very basic, but something that we can do and leverage a little in that skill that chatgpt has to interpret a lot of information, even if it is messy, is to copy the entire email list to which we want to respond and pass that as a context to chatgpt. Then a fake email is prepared, let's say a false email chain to give an example, but basically I am simulating that we are talking about a thumbnail. So, what thumbnail do you want to make for the chatgpt video? I can't think of anything. It depends on what photos we have. I think it would be cool to update them because lately we have been repeating photos. Could you take photos if you called what? They are these. Do it together. There is no do it together, but it is for an example. I would like 1. that you review the photos and choose which one is better. 2. that you tell me what you think of the free full course copy, that you tell me what you think of the title free full chatgpt course and that you give me the ok for the index that I have prepared for Telegram. Greetings, Adrian. This is an example email, it is a fake email, but it will help us to make the example of this. What we would do would be to copy all this, we are going to go here, we are going to open a new chat of chatgpt and we would pay the prompt that we commented. Then we go down all the way, we skip a few lines so that chatgpt detects well that we are in another section and we can even put all this in quotation marks if we want to make sure that it understands that this is the context. And here, answer this email in a friendly tone. I want your answer to contain the following information and we are going to answer and we are going to say I like the third photo. Then I do not like the copy, I would rather say free full course, that is, instead of free full course, free full course. Yes, the title. I like the title. Then this will be a fairly vague response to the email, to put it in a way, quite poorly ordered, but with all this information that we have given to chatgpt, what it will do is grab this information, grab the context and give a contextual response based on all this. As you can see, you are already writing an email in a much more friendly tone as we have asked. And all this is something that we can apply in our day-to-day, in our work. Another very interesting thing that we can do for your professional career is to negotiate your salary thanks to chatgpt. In chatgpt there are custom GPs trained specifically, just as we have trained one before with documents on technical analysis of investments, trained on how to negotiate salaries or trained in negotiation. These are some of the interesting prompts that we can use. Investigate if it is usual to negotiate your initial salary in the sector D and your sector N and your country. You can only enter the data you get necessary to reach a conclusive answer. So, basically, based on this, maybe there are some salaries that are agreed by an agreement and cannot be negotiated. It is important that you ask this to chatgpt before even asking any possible negotiation or any figure that creates the agreement. Then, develop a strategy to find out what is the maximum salary that I can request for this job in this company. You can request the data you consider necessary to reach a conclusive answer. We put this last phrase because this way we encourage, let's say, we give an incentive to chatgpt to ask us questions. And as long as it asks us questions, the answer will be much more effective and much more conclusive. That is, we are telling chatgpt, hey, I'm here for what you need. If there is a specific data that can help you reach a better answer, tell me and I will look for it. And in this way, we work hand in hand with chatgpt in order to find out what would be the maximum salary that we are going to request. And then it would be directly that we can ask him to respond to a job offer. For this, I have prepared an example. This would be with this custom chatgpt called salary negotiator, which I already have right here. And basically what I have prepared here is a very short document that represents a company called Burgos Tech that is offering us a job offer. So basically we are going to copy this, we are going to paste it, we are going to put the quotes and we are going to put this same prompt that we have here. Responding to this job offer, asking for a salary increase, etc. Then we are going to copy this prompt and replace the fields that need to be replaced. Responding to this job offer, asking for a salary increase to 15,000 euros. We are going to ask for 25,000 euros, for example. For this, he uses the following arguments. So here we put the arguments that we want chatgpt to use. And all of these, he will choose which are the best in terms of negotiation and how to propose them. We can say, I live far from the location and that means an extra expense. It is a very stressful job that can affect my health. For example, all these are examples. Here you would put the real arguments. Or, for example, it requires a very risky move on my part because I have to leave my current position and employment. And with this, we would hit enter. And from here, chatgpt will do its magic. Here it tells us what we said before, it is going to connect to an API. We are going to allow it and start working. And here, as you can see, he has already written the entire email. He has addressed the person as Laura Gómez, who specializes in human resources. He has addressed the company well. We have put Burgostech and here we see that the offer, let's see if I can find it here, comes from the company Burgostech. That is, he has introduced all the data correctly. 25,000 euros, etc. And he has used the arguments that we have put, but much more developed and in a much more professional way. Another very interesting custom gpt for your professional career, to put it in a way, would be the work advisor. That is, this custom gpt from here. Basically, this custom gpt allows you to ask commands like this. What are my rights with respect to extra hours in Spain? Or what is my minimum salary in Spain? Or how do I manage my unpaid leave? Or under what conditions can I apply for a paternity leave in Spain? Basically, any of these questions could be asked to chatgpt and it would even answer specific things. For example, this month I have worked 20 extra hours, but they have paid me only 17. How do I use this decision? And now I have finished generating. As you can see here, in this case we are talking about a custom gpt, which is a work advisor focused on Spain. And consequently, as it is trained with the information corresponding to the laws of this country, a recommendation has been made based on these laws. First of all, communication with the company. This would apply in any country, but here specific things come in from the data with which it has been trained. Then this process would be carried out in a mediation, arbitration and conciliation service. Finally, I would recommend you to review the applicable normative texts, such as the labor statute, the deregulation of social jurisdiction. All this is to say, legal aspects directly extracted from the information with which this custom gpt has been trained. Which we could not do if we had directly written the message in a new chat of the normal gpt. Having explained these tools for a person who works in a conventional job or who wants to work in a conventional job and wants to negotiate his salary as we have seen before, we are going to see now in this other aspect everything that has to do with freelance and ways in which we can either make more money or enhance our skills or generate new sources of income as freelance. The first tool that we are going to see for freelancers is the Fiverr ad creator. Fiverr, which I understand that many of you will know it, but we are going to do a brief review for those who do not know it, is this platform here, which is basically a marketplace for freelancers, that is, people who can offer their services to other people. It is going to come here, it is going to post ads and in this way it is going to get customers. For example, if I look here for design, we are going to get a thousand ads from different people offering different types of design for packaging, for shirts, for logos, a thousand things. So what happens? Really, Fiverr is very good, but you are competing with a lot of people, so you will always be interested in getting all the advantage you can on those other advertisers, that is, to stand out from the others both in terms of presentation, that is, once your ad enters, you give some image as a level of SEO, as a level of marketing and that is to say that your ads are positioned better than others by keywords, etc. Now to give you an example, here we have two ads, both are about the same thing. This one here is a guy who does Python programs and as you can see, it has a very short description, it has few labels, in the order description it only has three lines of text and as you can see, I think it has no order because nothing comes out. And normally here stars should come out and orders and orders should come out because it is a pretty bad ad. On the other hand, if we go to a professional ad, such as this one, we see here that it has 249 orders, 5 stars and that it has a much better description. Well, something like that is what we are going to do with the Fiverr ad creation tool, with the Fiverr ad creation custom gpt. So this custom gpt is specifically called Easy Geek Creator, you can also open it from this presentation itself, which we remind you that you will have it in the description and once it is open, this window will open here. So in this case we do not have to write it directly, but we will click here. Ok, I would have already generated the answer and basically it has asked us four things, it has asked us about what our ad is about, that is, what we are going to sell, our target audience, who we are going to sell it to, the only selling points, it would be the literal translation, but the short translation would be something like our differentiating elements, that is, what we are going to do better than the competition and keywords if we have them. So in this particular case, for example, we are going to put, in case we are selling a poster design service, we are going to put 1. I will design a poster, we are going to make it more specific, in fact, we are going to make them posters for graduation parties, for your graduation party. 2. Target audience would be students, teachers, parents of students, what else could it be? People who work in schools, universities, anyone. 3. Unique selling points, differentiating points, for example, my offer will be the cheapest of Fiverr, because we are starting and we want people to buy us, we are going to put this as a selling point. 4. I put a lot of effort in my designs, I also use tools from it to get the best possible result, and what else can we put? We can put that we make very fast deliveries in less than 48 hours. Then we would put keywords, we are going to put poster, party, design, photoshop, and we are going to put that it adds its own keywords. And with this we are going to hit enter. And this is something that I think throughout the course we had not seen, but basically there will be situations in which ChatGPT will tell us to choose between one of two answers. This is basically a feature that ChatGPT has that they do to train their algorithm, in this way they have feedback from users and they can know which has been the best response generated and consequently, based on that, improve their own tool. But it does not have more, we are going to choose any of the two, so we are going to put, we hope to generate both and choose one of the two. We can choose this one, I think, I do not like this one to put keywords. And with this we would have, as you can see, we have gone from a very small text, which is this here, and very simple and without much hook, to all this here, which is much more striking for an ad. Description of the Geek. Transforms your graduation party into an unforgettable event with unique and personalized poster designs. As a designer, I specialize in capturing the essence of your celebration, creating striking visuals that leave a lasting impression on your guests. Here all the information, such, such, such, I even added a section of FAQs, of frequent questions, in case you have any questions, leave them calm, so that you can do your job well. What do you need to start? I just need the details of your event, such, such, such. Can you request changes? Of course. In what format do you receive the design? JPG, PNG, all this has even taken it out of its own harvest, which is very good. How do you advance quick deliveries? My process is optimized for efficiency, from conception to final delivery, such, such, such. As I said, a practically impeccable ad that will compete with the best Fiverr ads in that regard, and much better than those examples that we have seen at the beginning, which were simply three lines of text. In the end, as I said, ads are one of the most important parts to position yourself, that is, to get more sales, since with a bad ad you will not get sales, as we have seen in the example before. And with this we would have, to put it in a way, 50% of the problem solved. The other 50% would be the photo, let's say, of the ad character, but that, well, at the moment the idea is very limited in that sense, and we just focus on the text. Another fairly useful tool is Freelance, which I have put like this as an extra since I understand that it will not apply to everyone, but I think it is interesting, is this custom GPT from here called Nomad List. Basically this is a custom GPT trained with different information from different possible locations to live if you are a digital nomad. So you can give it different information here, that is, if you work from home and you don't need anything more than your computer, and you can live from anywhere in the world, you can give it different information so that it recommends you different places where you can go to live if what you want is to live abroad and you want to live maybe with a more adjusted budget because in your country maybe the expenses are higher. So we could say, for example, this, I live in Portugal and I earn very little, $ 1,200 a month, I like to be focused on my work, I also like to meet people to have a drink, you don't need to go out to party, I want a place where I can live comfortably without leaving Europe. We send this to him and let's see what he tells us. Now what he is going to do, I understand, will be to analyze his database, correct. This sometimes happens, apparently there was an error, but then it seems here that it is already being solved, let's see. Yes, it is running again. And it would have finished generating and the sites that have been recommended to us with our current budget are Krivoy Rog and Ivano Frankivsk, both in Ukraine, which as you know now is a very desirable destination for you to live. And he tells us that they could be interesting options within Europe with the budget that we are commenting on. So I said, here I understand that maybe he is using a relatively outdated database, since indeed at the budget level it does not fit, but right now going to live in Ukraine is complicated. So I said, always take it with tweezers, but within the files that this has, you can get interesting information. And this tool concretely, I think with a couple more prompts and giving a couple of indications, for example, if I answer him now, it would not matter to try to earn more and live in a place that is not Ukraine. What do you recommend me? Let's put 1,600 so I have more margin. And here it would have finished generating and as you can see, it has already given us a much more accurate answer. It is true that many of the answers that it has given us are in Russia, which of course would not be bad, but it has also given us a couple of quite interesting answers. We can go to Adana, Turkey or Skopje in Northern Macedonia, which if I remember correctly is the capital. So we would even be well connected. The next tool that we are going to see within this category of freelance is Hacienda GPT. It is another custom GPT that has been trained with legal documents in order to help us manage our tax obligations. That is, to manage the whole issue of taxes, etc. This is very useful, especially in Spain, since if you are an employee it is already quite simple to manage the whole tax issue. But if you are self-employed, if you are a company, it is extremely complex. And if we talk about investments, well, more of the same. That is why it is good to have a tool like this, even though it does not replace a manager or a tax advisor. Since we remember that in the end, if you have any problem and you have problems with justice, you can't just tell him, no, he told me, take out the GPT, this is very important, underline it. But within a framework, let's say, of orientation, it can help you a lot. So you can ask him things like, for example, should I give myself a high rate of self-employment if I earn less than 10,000 euros a month? This is a myth that is said a lot. You can ask him things like this. Or how much should I declare my income if I am self-employed in Spain? Or how do I declare my earnings if I sold Tesla shares and I have gone from 1,000 to 1,345 euros? That is, if I have a profit of 345 euros. Or I have even won the lottery. How do I do to pay the taxes that correspond to my prize? What percentage should I pay? Things like that, we can ask him. So, to give you an example, let's see some of these. Specifically, I don't know, this one, the first one. And let's see what he tells us, Hacienda GPT. It is important to emphasize, again, this is a GPT chat focused on the prosecution in Spain and that has been trained with documents in Spain. The good thing is that, being documents created by the community, if we go here to Custom GPTs, we can look for the same. For example, imagine that you are Peruvian, you can look for Peru. And surely, Peru Cuisine, Travel Peru, Lawyer Peru. We already have one here of lawyers. Or Mentor Startup Peru. We already have several focused, let's say, on what would be the legality in Peru. So, in that sense, you don't have to worry. And maybe, specifically, this one is not going to work for you, but there will be some for your country. And regarding the question we asked him, he told us, with the precision of details, it is important to access the website of the tax agency, etc. And it says, according to the information found, there is no minimum income import defined by law to determine when to give up with the car. The key is in the habitability criterion. If your activity is not sporadic to generate income regularly, you should give up, regardless of the amount. Basically, he has already given us a complete answer. There is jurisprudence that suggests that if you do not earn the minimum interprofessional salary, it would not be necessary to give up. But this excludes obligations if the activity is considered usual. He even puts us here a link to the source, I understand, where he got all this information from. That's right, one more tool that can be quite useful to you, if you are freelance, at least in Spain, and if you are freelance in another country, then looking for the equivalent of custom GPTs for your country. Another sector or another field, let's say, in which we can use these custom GPTs and these tools that ChatGPT offers, is in content creation. In fact, we have found quite a few tools that can help you start a YouTube channel without having to do many of the designs or many of the processes that would require creating a channel or maintaining a channel. Specifically, we have found this one here, which acts as an assistant on social networks, that is, a kind of coach on social networks. It can help you, for example, answer questions about how a certain social network works, how a certain algorithm works. Let's see it here. It's called Social Butterfly AI and we can ask it things like, for example, what does the YouTube algorithm value more? Visits or likes? It has already answered us and basically, to a large extent, the answer it has given us is that it is not a question, well, that the answer is not as simple as the question, but apart from the visits, it also takes into account the retention of the audience. In fact, it is quite cool because it marks us in black the keywords. As for the likes, it tells us that they are important, but that it is not the only thing that matters. But other interactions such as comments, sharing or viewing time, which would be the same as the retention of the audience, are also important. It has given us a fairly complete and fairly accurate answer when it comes to answering this question here. That is, we have come up with a question that was not even well thought out, and yet it has managed to give us a very well-planned answer, in a way. The next custom GPT that we are going to see is the GPT logo, which is basically a custom GPT that allows you to create your logo. That is, it prevents you from having to talk to a freelancer or having to get a stolen logo from the Internet. One of the most interesting tools that this custom GPT has is that you can upload a logo sketch and with that you can create your logo with this custom GPT. So let's see, for example, let's open Paint, let's make it very crappy, and let's draw our logo. Let's imagine that I want my logo to be a moon. A moon. Let's put it in yellow. Look, let's put a background. Since we are here, let's do this. Let's put a blue background with a yellow moon. Okay, and let's also paint this in yellow. Let's see. It's been a while since I used this, since I used Paint. I can't remember the word. And let's put a green triangle underneath. Red underneath. This will be our logo. And we're also going to paint it red. So we're going to copy this. And we're going to pass it to the GPT logo. And we're going to say, create a logo based on this sketch. And we're going to say, it's for my YouTube channel. About astronomy. Astronomy. What's it called? And we're going to put, what's it called? Triangular astronomy. I can go with the logo. Here it has already asked us what style we want our logo. So we're going to put, I don't know, of all these, let's put geometric, for example. And it also asks us if we want a specific color for the background. Well, we're going to put blue, since we've put blue here. I want the background to be blue. And we would have the logo created here. Even so, we have to emphasize something. The truth is that it has been very, very cool. It is not exactly at the same disposition, but it has made the red triangle. It has made the moon in the same position. It has made the background blue. It has done everything perfect. The only thing, what we commented before, that we have not seen. We are going to review to see if I find it. But we have seen it in this slide here. The issue that we mentioned before, of the texts that are missing, about the letters, replacing letters, etc. Well, here we have a clear example. Instead of astronomy, it has put astronomy. And instead of triangular, it has put, well, it has put triangular with dosis. So, as you can see, the letters are perfect, but it has failed in the letters. In this case, for example, we could either edit it with Photoshop. Or we can say, make another similar logo, but without text. And here we will have it solved. In fact, in the end, it has not put a triangle, but it has done something pretty cool. That is, it has put the YouTube logo here, in the logo itself. And as you can see, this would be as simple as copying it. And with this, we could download it and put it in our channel. Well, we could not download it directly. I mean, put it directly, because I'm seeing that it is in the web format P. But hey, this is already extra. We have free PNG converters on the internet. This one specifically is called image.online-confer.com. And this would be as simple as dragging this file that we have in web format P. And we click Start. And we click here to download. It would be as simple now as uploading it to our channel. But we already have our perfect logo here. Continuing with the custom GPTs, or content creation tools. We have this one here, which is called Awesome Banner. Which we have also put here in the link. And if not, you can also look for it in the custom GPTs. But basically, it is a tool that creates a YouTube banner for your channel. Just like what we are going to do with the logo, we can make a banner. So what we can even ask for is, we can pass the logo that we just created. I think, in fact, we have it here. And it will not be exactly the same, but it will be strongly inspired by this design. So we can ask for, Inspire yourself from this logo to create a banner for my YouTube channel focused on astronomy. And in advance we are going to ask him not to put text, because we have already seen that in that ChatGPT does not excel. So we are going to say, do not include text in the banner. Ok, I would have already generated the image here. As it is true, I have done it quite well, in my opinion. The YouTube logo has done it very well, which is not usual. Generally, as with texts, the logos are usually done quite badly in ChatGPT. But here it has come out quite well. And, as I said, although it is not exactly the same style, it does have many similarities. Well, it keeps the YouTube logo, the shades of the colors are similar. In both it has made this style of triangles in mosaic style. It has added these little stars as well as crosses. So, pretty cool. If we do not like it, we can either ask him to change some specific effect. For example, we can ask him to make it with flatter colors. Or, so that you have the other option, we could simply give it here to regenerate. But now you can. For example, we are going to ask him to do it with flatter colors. So that the style is more similar to that of the logo. And we already have it here. Here, in my opinion, it has happened. And he has done it with too flat colors. But hey, this would be a matter of playing until we get a result that we like. This, the truth, is not bad at all and could be used directly. Especially if we are talking about getting an MVP, a minimum viable product. That is, trying to get our channel to work quickly. Which is usually the best. If you have an idea, maybe try to test it quickly to see if it works on the market. And not spend too much time on details like this. That in the end, you could spend a lot of time and effort to get a good banner. That if the channel does not work, nothing has served you all that time to get a banner. So I think the ideal would be to get a quick logo. Get a quick banner with this. Have a presentable and beautiful channel. And from there, I would start uploading videos. How are we going to upload those videos? Well, we can either write the script. Or using the next tool that we have here. Which is Script Pro. Basically, it is a custom gpt. Which will allow you to generate scripts for your YouTube channel, for your videos. That is, it will help you generate scripts. For example, you can ask it to create it from scratch. You can ask it to create a section. You can ask it to improve a script you already have. Or you can ask it to make you an advisor. You can ask it to fix errors such as lack of spelling, lack of expression, lack of documentation. That is, there is some information in the script that is wrong. You can ask for all that. But in this particular case, I think that maybe it is not worth using mode 1. And it is more worthwhile to do this part of the script creation process using Harpa. Using the tool that we have seen before. In this case, what we can do, since we are having an astronomy channel, is to go to Wikipedia. Look for, for example, Saturn. The planet. And ask Harpa to generate us a short about Saturn. So I have already prepared this prompt in advance. Make a script for a 60-second short. Before that, sorry, you have to click here on Chat with Page. We can change from 3.5 to HTTP 4. And we say this. Make a script for a 60-second short. Talking about 10 curiosities about Saturn. It must be dynamic and entertaining. We click on Enter. It seems that it gave error. We are going to try with 3.5 and sometimes it happens that if it is saturated it does not work. And here, as you can see, not only the part of the script has not been put, but a proposal has not been made at the level of how the video should be. Quick image of Saturn from space. Emotional music in the background. Get ready to discover the secrets of Saturn in just 60 seconds. Animation of the dazzling rings of Saturn. Saturn is famous for its impressive rings, which are mainly composed of iron and rock particles. And it is saying all the indications of how the video should be. The truth is that this proposal had not been approved and it seems to me a barbarity. And then, at the end of the talk, as you are looking here, I also added a call to action. End screen with the channel logo and subscription. Do not miss more content about the universe and its wonders. Subscribe for more amazing videos. Well, we could copy this same video. And we could ask him here to improve it in some way. In this case, in principle, it is quite good. So we are not going to touch it. But we could combine these two things. Harpa and Skip Pro. In order to get the best possible skip. As you can see, with these four tools, we would already have many of the aspects that imply creating a fairly solid YouTube channel. We could, on the one hand, look for guidance on what steps to take and if we are going in the right direction with the network assistant. We could create our logo, create our banner, create the script of the videos. And we would simply have to reach enough views in order to monetize it. Either through ads or sponsors. If we are creating a source of income out of nowhere, quite leveraged in artificial intelligence and having to put a lot less work than if we were doing it from scratch, manually, as has always been done. Even so, what I have said, what we are repeating throughout the video. ChatGPT, on the one hand, is not perfect. It makes mistakes. And on the other hand, on YouTube, human connection is highly valued. That is, if we now make a channel with this, with an artificial robotic voice and everything, let's say, prefabricated with artificial intelligence, sometimes it is difficult for the channel to work. And if it works, it is very easy for it to be copied. So it is important, what we have been saying throughout the video, to make a mix between your own work and the work in which the AI helps you. That is, to leverage your abilities in artificial intelligence. The next section in this section of how to make money with ChatGPT will be programming. And do not be alarmed, do not skip this section directly, since although you do not know how to program, you can find jobs for other people without knowing how to program. And in fact, we are going to give an example, but first we are going to do a brief review of what the strategy is. Basically, what we are going to do is find jobs or positions with potential, which means finding positions in Fiverr that have purchases, that is, that people are using and that they are having clients. And we are going to try to compete against those positions, either by offering a cheaper price, offering better conditions, for example, more revisions with the Easy Creator that we have seen in the freelance part. So, let's see an example of this type of ads, which would be this one here. We have a person who has a lot of orders, has 145 orders, and basically offers you three different packs. Here, for example, in the basic pack it simply helps you solve some bugs, in the most advanced pack it helps you solve bugs, but in more languages, and only in the premium is where it makes you the full code. That is, we could lower this price to compete with it, so we have a margin to compete, we have a margin to do it better, and if we have a margin to do it better with a person who is making 145 orders, we have a margin, maybe, to make 500 orders, since we can do it better than him by offering a more competitive price, getting more orders, and making optimal codes, thanks to AdShockJPG. How are we going to do this with the codes? We are going to use this tool here, which is called Python, which is a custom GPT that we have here, and this, basically, what they have done has been to pass documentation about this programming language, Python, and that will allow us to create code from scratch without having knowledge, simply giving indications of what we want. For this, we are not going to stay here, we are not going to do it simply in theory, but we are going to make a practical example. In the case of this ad that we saw, the codes that this person performs are quite simple, that is, if we put a video here of one of the orders he has made, information comes out like this in a box, he has the option to start the quiz, and he has a quiz with different questions, they are simply buttons and answers, something extremely simple, and as you can see here, with an interface that leaves a lot to be desired, that is, it has a lot of room for improvement. We are even going to make it a little more complex, and in our example, we are going to program a calculator without knowing how to program, that is, we are going to use this tool that we mentioned before, to create a functional calculator in programming without knowing how to program it. For this, we are simply going to put this prompt, we are going to say I want you to program a calculator in Python, and we are going to give it some extra indications, we can say I want the calculator to be used to add, subtract, divide, and multiply. Ok, we have the same problem as yesterday, so I'm going to change to another account of the GPT chat and we continue. Ok, now that we are in another account, we can continue without having the limit of GPT-4. By default, it will be strange for you to skip the warning of the limit, which is that in this course, obviously, we are using it very intensely, and that is why we reach the limit very often. If you make a normal use, by default, you will not have this problem, you will have to use GPT-3.5 or even pay two accounts, as I have done to record this video. So, we are going to paste here the prompt that we are preparing and we are going to wait. Ok, I have already finished generating the code, as you can see here, it has given us all the explanation of the design process, since from what I have been seeing here, it has made a lot of emphasis that it is very well programmed and very well structured. So, we are going to go down here, and down here, we get this one here, and paste it in a code editor that allows us to run it later. If you do not know what a code editor is, it is simply a program that allows you to paste code and run it, edit it, etc. And, well, if someone is interested in knowing how it is installed and how to use it, and wants to have some extra knowledge, rather than just talking with GPT-3, we have a very interesting course on our channel, a free Python course, how to learn Python and make money as a beginner, which was taught by Randy, who is quite complete and teaches us how to install the same code editor that I will use, which is called Visual Studio. So, in this case, we are going to open Visual Studio and simply paste it here, in the code editor that we commented, and we click on this arrow to run it. And here, as you can see, we have not opened any window or anything, and it is that the code that this custom GPT has generated for us is a code without an interface, it is a code that works by terminal. So, if I want to add, subtract, divide, multiply and exit, this is quite common in programming when different things are prototyped, that you do not generate the interface, but you simply do it by commands. So, for example, if I want to add, I put add, and it tells me, enter the first number, I put 25, enter the second number, 10, and it tells me, the result is 35. Ok, this is fine, but for example, for the order, this Fiber that we talk about is very limited. If we enter it to a person, for example, for the custom GPT, I like the code, but I would like it to be seen in a window with buttons, that is user-friendly, which means that it is, let's say, comfortable for the user. Ok, the generation is over. As you can see here, it makes us a review of all the steps it has followed. In principle, in the hypothetical case in which we do not care, and we simply want to generate the code and paste it, all this would not matter. We simply go here, we copy the code, we go back to Visual Studio, where we have pasted another code, we paste it, we remove this, this I have put it so that it does not run inside the chatGPT, since if not, chatGPT would try to run it and it could not, because we are asking it to generate a window. And with this, and this out, let's see, this out here, and then, from here, we simply hit play and automatically, the calculator that we have programmed appears on that side. As you can see here, 1 plus 1 equals 2, plus 6 equals 8, multiplied by 8 equals 64, divided by 3 equals 21.336, and we have a completely functional calculator. Even if we want to go further, because it is, let's say that in terms of design, the truth is that it is quite good, it is much better than the quiz that we have seen before, but we could ask it for more things. We could, for example, go back here to our custom Python GPT, here it is, and we can tell it, now make the interface more futuristic and attractive. And we can ask it, for example, to use a black background, to use orange and blue, well, orange, white and gray, and this dark gray, better. For the buttons, use the font, let's put white to make contrast, and in black. And use animations in the text field of the result. We can even ask it to add other aesthetic changes that you think will improve the attractiveness of the calculator. And now again, we wait for all this to be generated, and we have just generated the response to this prompt and we are simply going to see how it has done it. We are going to copy, we are going to open Visual Studio again, we are going to paste it, again, we have to remove this, remove all these symbols so that it works correctly, and we are going to see it again here. And now, as you can see, the calculator looks much more attractive and it works exactly the same as the previous one. 36, we are going to put it divided by 8, and everything there is perfect, and it has given us the title Futurista Calculator. Here I think it has passed because it is fine, but I don't know how to call it Futurista. But hey, as I said, it has improved a lot, we could even ask it to make us the round buttons, maybe put the numbers in black, things that we want to ask it specifically, we can ask it without any problem. And seeing this, it would be to do what we have commented, simply publish different gigs, and in this way we are trying to get orders, and we are going to solve them, and you see that in a moment with ChatsGPT we have made this code that this calculator programs, which is also a perfectly structured code, that has no error, so if for example it is for someone's work, it could be done without problem, or for example, I'm going to create a quiz code like the other girl that we have seen in the Fiverr ad, so that you can put it for other people like freelance, thanks to ChatsGPT, having minimal knowledge or practically non-existent programming. It is also important to mention before moving on to the next block, that what we do to program using ChatsGPT can be done much more efficiently using a payment tool called Copilot, which is basically a GitHub tool, this here, that integrates directly into the code reader, that is, if you look, this is the same interface and you integrate here directly the same program. This has several advantages, the first obvious thing is that you have it all on the same side, the second, that it has more context of all the code you are programming, that is, if your code has several sections, it has them all present, without having to go through pieces, without having to go through first a window to the code and then the other, it has everything much more present in a place where data is stored, where people can publish their programming projects. Specifically, and relating it to videos that we have seen on the channel, recently we made the video of Influencer Conia, that code that we use for the video of Influencer Conia is published on GitHub, that is, that code could be used by Copilot for its information and help you solve problems of that caliber. And as you can see, here we have all the information about the machine we use for mining Bitcoin, in the last video, the cheapest machine for mining Bitcoin, all these codes were also on GitHub, and many more have been used to train this AI, so the results it will give you will be much more efficient than the one that the GPT has given you. But hey, as I said, this one here is not free, it has a cost of $10 a month and it is a little more advanced. Still, I found it interesting and much easier to understand, since you can ask for explanations about the code, explanations about different solutions, etc. Having explained all this, we are going to move on to the next block, which is art and design, since, as we can do in freelance positions as a programmer, we can also do them at the design level, as we have seen before when we did the example geek. So, in this specific field, we also have many custom GPTs that can help us, for example, we have this one here called Video GPT by Vid, which helps us to create videos. And now you will see, because it has a bit of a trick, it is not that it creates videos just like it creates GPT images. We are going to go here and we are going to go here, for example, to Let's create a video, which basically we are asking it to create a video for us. Here it is asking us some questions, if you do not speak English, we can translate them here if you want to get an encouragement and how long do you want the video to last. So, we are going to tell it, look, the video is about the importance of climate change, for example, it is aimed at older people over 60 years old. I want it to be in an informal and friendly tone and we are going to put that I want it to last, we are going to put 15 seconds, that it is very short so that it does not take too long. And with this information we wait and here the script has already been generated, it tells us that climate change affects us all, but it is not too late to act, small actions such as recycling, basically a script has been made for us. Let's say we want an avatar, let's say yes, let's say we want Devon, for example. Here it is asking us to give access to vid.io, which is the tool with which this custom gpt is connected, we are going to give it to it. And here the video has already been generated, then it tells us that to see it we press here, we are going to press, another window opens for us, in fact I am seeing here that the video is not really finished, it says here 10%, so let's wait. And as you can see here, it has already finished loading and we are going to see the result.

Speaker 4: Climate change affects us all, but it is not too late to act, small actions such as recycling, basically a script has been made for us, let's say yes, let's say yes,

Speaker 2: basically a script has been made for us, basically a script has been made for us, and from the information we have generated in ChatGPT, that is, we have chosen the avatar and such, what it has done has been to generate the video in another different service, and really if we wanted to download it, we would have to pay, that on the one hand. And that's why I said before that it has a trick, because if we want to use this video creation service, really what we are going to be using is another page, and ChatGPT is just a way to interact with that page through the chat, but on the other hand, and the most important thing, which is as you see here, what has happened is that from the Spanish script it has tried to make a video in English, and that is why it has done it quite badly. So, I said it, in the first instance it is a service that can be fine, but I think it makes no sense to do it through ChatGPT, it was an example, let's say, of the limitations we have today. Even so, we will see later that the subject of videos may soon be solved for what is ChatGPT and artificial intelligence in general. On the other hand, the subject of language, if we make a video in Spanish, you see here that errors are made. Can we do it? Yes, but if we want to do it, it is in English, and if we do it in English, then it is better to do it directly through this service or any other. In fact, in the channel we have a tutorial on how to use QuickTalk, which is a service very similar to this and I think even cheaper. So, I said it, this is just an example so that you can see that many times there are custom GPTs that are not exactly what they promise you or that do not work as you expect. But the others I do think they are quite interesting, so let's go with the next one, which is the image generator. You may ask, why do we need an image generator in ChatGPT if we already have Dalí, we already have the one we saw before integrated? Well, basically what this is going to do is from a fairly simple prompt that you are going to give it, it is going to put certain extra elements, let's say to that prompt, so that the generated image is much better. Imagine that I want a car, futuristic, flying. Basically what this is going to be doing is to put extra information to that prompt so that the result is much better. And here, as you can see, with very little information, that is, with a very short prompt, we will already be getting a very good result at an aesthetic level. Basically what this custom GPT does is the same as we have done before with the Ferrari Testarossa, that is, we have generated a very long prompt so that our resulting image is much more beautiful, it does it automatically without having to worry about taking out that prompt. That is, it is using the same engine as Dalí 3, that is, it is using the same image generation engine, as in the previous program, so that the results are better. But the image generation really uses the same technique, it just does that extra previous work. The next design tool that we have included is this one here, Cartoonize Yourself, which basically turns a photo of you, a selfie, let's say a portrait, into a drawing. I have included it because, despite the fact that it may seem silly and that it is very limited and it is in a very specific aspect, to make a photo, you generate a portrait, let's say, cartoon, like this example here. So I'm going to show you what the results are with a photo I took before, we're just going to copy it and we're going to paste it here. And we're going to say Cartoonize Myself, which would be something like turn me into a cartoon, into an animated drawing. And we already have it here well, that is, I got curly hair, I got a beard, I got brown eyes, I got the shirt, pretty cool results. If we wanted to create more, we could click here on regenerate to create variations of the same. But as I said, I think this is a service that may be interesting to sell in Fiverr and it would be as simple as copying the other custom GVT, the one we saw before so that people can compare and get an idea of what the result is. And from there, sell this as a service in Fiverr. And in the same line of creating characters, in this case we are creating it from a photo, but also what we can do, also for the sale of services, for example in Fiverr, is to create consistent characters. That is, imagine that I want to create a character for a sketch or an animated drawing and I want this to be my service in Fiverr to sell it to other people. Well, we would go to this other custom GVT that you have in the description and what we would do would be simply click here. We wait. It asks us if we want it to be male or female. The character, let's say we want it to be male. We put male. The name of the character, I'm going to put mine, I'm going to put Ramon. It asks us to make a description of the character. Here we are going to do a test, I do not know if this can be done and if it goes wrong, let's go. But I'm going to try to pass him directly to the base limit. Or rather in English, we are going to put it in English so that it catches it well, because here it has asked us in English. Well, we are going to put it in Spanish, we are going to take it to the limit to see if I can do it in this photo for the character. And we are going to ask, for example, it is asking us here to put what variations we want, that is, how we want the photo to vary. Expression. Okay, and it asks us what the first expression is, we want it to put happy. Let's put happy. And here we have created the first photo of the character, then in theory it will not be consistent with respect to this photo since this would be like the description, but from here they should be consistent in the next photos. So let's do the test. It asks us what we want the next expression to be, we are going to put that we want it to be sad. And as you can see here, it is doing well, because this photo and this photo are effectively the same character, the only thing that is happy, and here it is sad. And we could ask for more photos and you see here that, unlike before, that with the descriptive photo it does not generate the same character, here it is generating consistent characters. The shirt is the same, the face is the same, the eyes are the same, the proportions are the same, the only difference that we could see is that of realistic photos, that basically what we can do with this is generate photos, generate images of people that do not exist in a realistic way. So let's open this and see how it works. Basically here we can say different descriptions of how we want the photo, but let's put, for example, a man older drinking a cup of coffee. And this custom gpt, which is this one here, which as we said was still using the I and the only thing it did was extend the prompt, that is, it did a previous job but it worked with the same artificial intelligence, it works with an external artificial intelligence, that is, it works with this called geni-gpt.net and basically what it has done is generate the image that we can see through this link. It says it will take a while, so let's see if it has already been loaded and indeed we have generated a perfect image that honestly I would not have to use an external IA with Dalí, no kidding. Of course it is perfect, the number of fingers that seems silly is perfect, but if you have seen our video creating an influencer with artificial intelligence it is quite difficult for an IA to generate a person with the correct number of fingers and the truth is that all the proportions, the details, everything is ideal in that sense. Anyway, something important that I want to emphasize is that from my point of view you are very limited in terms of image creation if you do it through gpt chat and I think there are much better options. Specifically this one here is called Focus, it is a local format on your computer, that is, you download it and run it on your computer. It can be downloaded from github. I'm going to show you now what the repository looks like. Look, here we have it. In fact, we have shown it before as an example, but basically you go to this page which is the one we have linked here in the document that you will have in the description and basically you go to the download section which is download, you download it, you have three files like this and you can run it. If you want to see a little how it works, we have it here. You have a thousand parameters, you can put an image as a reference, you can paint parts of an image, you can change the style, you have a thousand styles to choose from and each one with an example image of how it would look. You have different resolution options, you have a thousand options, many more than you have in gpt chat. If any of you are interested in seeing a little how it works before downloading it, we have a video, this one here, which is called Influencers Con Ea are getting rich and they do it, in which we basically explain the business model of Influencers Con Ea with this artificial intelligence here, which the truth is that it is very complete and very cool. So if it is something that interests you, generating these images with artificial intelligence I would recommend you go directly to focus, do not use any of these since you will have much more freedom and many more possibilities if you use this intelligence here. In fact, it is the intelligence that we use in most of the animations that you see on the channel. The base photo has been created with artificial intelligence and later it has been animated through After Effects. The next sector in which gpt chat can help us make money is in everything related to investment and we will see several tools related to it. The first of these tools is this one here, which is a custom gpt called Finance. And this custom gpt here has the particularity that it is not a custom gpt focused on investment advising in a first instance, that is, it will not tell you invest in this, invest in this, this will go up, this will go down, but they will be much more abstract investment recommendations more related to financial responsibility. For example, how can I organize my budget to save so much or how should I distribute my different investments. And we are going to do a couple of examples so that you can see what I mean. Imagine, for example, that I earn 1000 euros or 2000 euros and I want to save 10,000 euros for the end of the year. Well, I can ask if I earn 2000 euros a month, how should I invest and in what amount in order to save 10,000 euros after a year. And here we can say, so that you take into account the whole investment issue, contemplate the investment in indexed funds and remunerated accounts. And here it would have given us a strategy and basically tells us that if we want to save 10,000 euros a month for the end of the year, we have to save a minimum of 833 euros a month. And this 833 euros a month, this financial advisor is recommending that we put 200 euros in remunerated accounts and then tells us that we can put the rest in an indexed fund. So basically the reasoning behind this is that it is good that you have an interesting amount in remunerated accounts that have instant liquidity and that allow you to face unexpected expenses. Because if not, what would give you more return is to put it all in an indexed fund but it has more risk and maybe in the face of an unexpected expense it forces you to sell in times where that indexed fund has fallen. So the truth is pretty good. Another thing we could ask is, for example, simpler questions like what is more profitable, the S&P 500 or a remunerated account? And here, as you can see, it has already answered us and basically the final answer and the conclusion, because it has given the explanation in several points, is that the S&P 500 has a longer profitability as long as we invest in the long term. But if we have to invest in the short term, it may be too risky to be investing in the S&P 500. So the truth is, quite good advice and an interesting idea for finance. If, on the contrary, we are looking for a much more focused advice on market projection, that is, to predict what will happen with prices, then we have this one. This custom GVT is basically something very similar to what we have assembled before, to the custom GVT that we have made for ourselves, to which we had passed documents on technical analysis, but much more elaborate. In fact, it is one of the most popular of the custom GVT. So we have it here and we can ask things like, for example, make a projection for the price of Tesla shares. So here, for example, it is trying to access another website, I understand that it will be to get updated data, we will allow it. And as you can see here, it gives us the disclaimer that this is just a technical analysis and should be considered as such. I am not a financial advisor and this prediction should not be taken as a financial advice. This basically tells us why in the end, as we have seen on multiple occasions throughout this video, the custom GVT is not perfect, it can be wrong. And even if it is not wrong, in the end, if this goes wrong and Tesla goes down, the only culprit, the only responsible and the only one who is affected is yourself. So you have to take it into account when making decisions. It is something with which you can help yourself, but it should not be your only source of information, obviously. Another quite interesting thing that we can do with the custom GVT is backtesting. Basically, backtesting consists of choosing a strategy and testing it over time. That is, going back in time to see how that strategy would have turned out. This is usually used a lot in trading bots. That is, you are going to program a strategy and based on that strategy that you have programmed, the bot will automatically make the decisions based on the guidelines that you have programmed. In this case, we have a custom GVT for this, to backtest. This same code can also be used to apply this strategy on web pages of trading bots. But basically it is done with this custom GVT here, which is called TradingView Bindstrip V5. So let's see how backtesting is and how we can do it with GVT. To carry out this backtesting, we are going to execute three steps. The first will be to get a strategy. The second will be to generate the code that will later allow us to visualize that strategy. We are going to do this with this custom GVT. It is a code in a language called Pinescript. And finally, it will be simply to visualize that strategy. For this, we are going to do it on a page called TradingView. Step by step, first of all, we are going to get the strategy. For this, we will go back to the custom GVT from before, to Finance Wizard. And we are going to ask for the same. And what we have asked for is to give an investment strategy for Bitcoin and USD. That is, between Bitcoin and the dollar, based on different technical indicators in order to draw the best conclusion. The strategy must generate buying signals based on those indicators with complex calculations. Let's see what it tells us. We give it to allow. Ok, the strategy has finished generating. Basically, it has given us four different indicators and signals or indications of when it can be a good time to buy or sell based on those indicators. What we are going to do now is copy it. We are going to go to the second custom GVT, the TradingView Pinescript. And we are going to say the following. It generates a code using all these indicators. It generates buying and selling signals for each time one of these criteria is met. And below, we are going to paste the code. Well, the code, rather, the indicators that have happened to us before. Finance Wizard. We are going to press Enter. And now I have to wait. Ok, and with this, we would already have the strategy generated in code format. It would be up to here. And if what we want to do is to do that backtesting that we mentioned. To see what results this strategy would give us. If we had applied it in a certain time period. What we do is that we go here to TradingView. We look for the asset we want. In our case, Bitcoin. But it is not necessary to look for it because I see that it already comes out here. And in this section below, which says PINE Editor. We paste the code we have generated here. It is exactly the same. And we are going to add it to the chart. And as you can see here, a lot of things have been generated. Now we do not see it well because we have a lot of zoom. But if we enlarge the chart. Each of these. They are buying signals and selling signals. And if here it says. Buy, buy, buy, sell, sell, sell, sell. And with this, it is difficult to know if it is profitable or not. But if we go to the strategy simulator. Ok, this is already solved. Apparently nothing appeared here. Because the code is presented as an indicator. Not as a strategy. To solve it, very simple. We have simply told the chatgpt. Nothing appears in the strategy simulator. And the solito has already understood what is happening. He has already changed the code as long as it appears now. So let's copy this new code. We paste it here in the same place. Exactly the same. Update the chart. So how do we interpret these data? Well, really the only data or the one that interests us the most. It is the gain factor. Which is basically the multiplier number. Of what we have won. We are talking about 1.299. We are talking that we have won approximately 30%. Also, if we go here to the settings. We could modify the quantities. That is, here, for example. We get an initial capital of 1 million dollars. And that is why here the gain figure is 11,000 dollars. But this could be modified. And we can play with all those numbers. But as I said. The important thing and what interests us. It would be the gain factor. And with this we would have done our backtesting. The same thing we just did. We can do it with different strategies. They don't even have to be Finance Wizard strategies. But we can search in blogs, on the internet, in videos. Even using Harpa. Different strategies. And bring them here so that it generates the code for us. And then do the backtesting. And another thing that can also be done from here. From TrainView. Right now we have it applied, if I'm not mistaken, to six months. But if we change the time range here, for example, to a year. Well, we get other different data. Since right now we are applying this strategy. About much more time. And in five years, the same. We have other very different data. In fact, in five years I think this same strategy will be lost. So that. Contemplating all the time horizons. Since in general it will not be as easy as looking for a strategy on the internet. Do the backtesting. And put it in a bot to run it. Since if not, everyone would be a millionaire. That said. What we said before. It is very important that you keep in mind that at all times. Regardless of which of these tools you use. You are the first and last responsible for your results. So. If something goes wrong, you will not be able to blame ChatGPT. And you will have lost your money. So do not blindly listen to what ChatGPT tells you. Which in the end is an artificial intelligence tool that can be wrong. And even if it is not wrong. Even if it applies the technical data correctly. There may be fundamental data. What is it called? That is, news. That go against it. That make that investment you made, that you thought was going to go up. Suddenly it goes down. So be very responsible. With all the decisions you make. And even if it is giving you the indication. An artificial intelligence that should be extremely intelligent. Do not go like crazy. Listening to everything that ChatGPT tells you. Another way in which you can make money. Thanks to ChatGPT. Which can be either through setting up your own agency. Or enhance. If you already have one. Or simply using these tools for yourself. That is, to enhance your business. Your YouTube channel. Whatever. The first of these. That we find very interesting. Is this one here. That is called Keyword Research. That basically helps you find keywords. This can be used, for example, for Instagram posts. For YouTube posts. Everything that requires keywords to position. Even, for example, what we have seen before. Fiverr ads. You can use this tool to find keywords. The difference this has with doing it directly with ChatGPT. Is that this tool is connected to the internet. Then, being connected to the internet. It will give you up-to-date keywords at the moment. For example, for a video that we have uploaded recently. Which is this one here. It's called Launchpads. The popular way to make money with cryptocurrencies. It's a video that we uploaded recently. And this is the title. You can say. Make me a list of 10 keywords. With little competition. This is important. Since a keyword can be very popular. But if it has a lot of competition. It's useless. So, this is the answer. For this YouTube video. That I'm going to upload. I would have finished generating the answer. And basically what it has done. Has been to give us a phrase. Or a series of keywords. That can be applied. Depending on the niche. Or depending on the sector. Or how we are going to pose the video. For example, it has told us. If you want your video to be more. As general educational information. We could use this. Both in the description. As in the labels. As in the title. And thus help position the video. Investment strategies. And as you can see here. It has put it to us from different pieces. Trends and future. Future of investments in Launchpad Crypto. And this, as I said. Applicable to any format we want to use. Even if what we want. They are not phrases. We could also ask for it. The next tool. Within what would be SEO and Marketing. It would be this one here. Which is called Viral Visionary. Which is basically a tool. A custom GT. That will give you viral ideas. That is, from a theme that you can pass. It will give you viral ideas. Related to that theme. And this is quite good. Because it is also updated in real time. Imagine that I want a list. Of the most viral cryptos of the moment. Well, what I ask. Here I say. Make me a list. Of the 10 most viral cryptocurrencies. Of the moment. I wrote it wrong. But hey, you know. It will understand it anyway. So now what it is doing. It is connecting with its database. I understand. Indeed. And as you can see here. It has already made us the list. This is an article from November 2024. So. It is not very updated. But hey, I said it. We always have to verify these things. Because GVT chat is not perfect. Another thing we can do. For example. Instead of asking for the 10 most viral cryptos. Because here it has really put us. The typical ones. We could ask him to tell us. The 10 meme coins. And in this way we find different currencies. Of which people do not speak. But those cryptos. He knows them. All over the world. Tell me. The meme coins. And coins. Less known. That are. Being. Viral. Right now. In April 2024. And as you can see here. It is finishing to generate the answer. Because he has done a much better job. Because he has used different sources. We see here that there are different pages. And in that sense. Much better. Because in fact some of them. It sounds to me to have seen them on Twitter. Or on YouTube. As very viral cryptos. Specifically this one. The maga. It became very viral. Because it was for the elections. And it was a coin like Donald Trump. And I said it. Although it has grown a little at the beginning. If we are controlling and guiding a little. We can get interesting results with this tool. The next one we are going to see. It's this one here. It's called Cell Generator. Which is not so focused on SEO. But more purely on marketing. And it is that from this tool. What we can get is to create descriptions. For the products of our store. Either for example a dropshipping store. Or conventional e-commerce. Or if we are selling an infoproduct. Focused on SEO. That is, that it is positioned on our website. And that it is positioned in its product. So let's see how it works. We have it here. And for this example we are going to do the following. I have taken a random Aliexpress mobile. You see here it has a lot of information. And basically we are going to copy everything. Without discriminating whether it is useful information or not. We are going to take all this. We are going to paste it here. And we are going to say. Make me a description. In Spanish. For this. Product. That is, in an easy way. Without breaking our heads. Directly we have copied all the information. From the provider. In the hypothetical case that we were doing. That we were doing dropshipping. And basically we want a description. Based on all this information. Which is messy. And obviously we could not put it like that on our website. And we would have finished generating the description. And as you can see, from all this mess of information. It has created a pretty cool description. The Xiaomi Mi Pad 6 is a latest generation tablet. Which highlights its performance power. High-end characteristics. Equipped with a Snapdragon 870 processor. Basically it has collected all the information. And it has put us in a very interesting product description. If we want another format. For example, if we wanted it to be a shorter format. And with bullet points. We can say. Short paragraphs. With. Bullet points. To get a description. Attractive. And easy to read. And as you can see here. It has done literally that. It has put us by bullet points. It has been combining paragraphs with bullet points. It has divided it by sections. And we can literally do the same. Pass all this information to it. Copied, badly done. And tell it to do it in the same style as this description. And in this way we will have consistency. In the different product descriptions. So nothing. Another quite interesting tool. Especially for what would be e-commerce. If what we are doing is selling through e-mail. That is, e-mail marketing. We have this here. The photo has made me especially funny. Because it's called Impossible Sales Rep. That is, representative of impossible sales. They are represented with a guy trying to sell him. An igloo, an eskimo. There is a refrigerator, an eskimo. Let's see how this tool would be. And basically it would be this here. We are not going to go into too much detail. Since this has no mystery. Simply create optimized e-mails for sale. But as you can see, here are some examples. An e-mail for sales for a technology company. For a health service company. For a clothing store company. Fashion retailer. Or for a restaurant. And basically we can apply this. Imagine that what you are trying to sell is a card design service for restaurants. Well, you can use this tool to send massive e-mails to many restaurants. And e-mails that you know are optimized to get a sale. Or if what you are selling are T-shirt designs for clothing stores. Well, the same. You can use this to generate the optimized e-mail that will get you that sale. Finally, in the same line that we had seen for content creation. That is, we have here in content creation Social Butterfly. Someone with a lot of knowledge in social media that could help you in making decisions for your YouTube channel. For your Instagram account, for whatever. We have this other one that would be the equivalent but for your business. Which is called Visionary Business Coach. That basically you can ask questions about your business. Ask him to make you a marketing strategy. A sales strategy, etc. So we have it here. And basically you can tell him, look, I want to sell ice cream in the North Pole. What do you think of my business idea? And you see here, it says your idea of selling ice cream in the North Pole is undoubtedly challenging and out of the ordinary. And here, even though I wrote something silly obviously, but for you to have an example of how it would work. He does an analysis, of what things you should contemplate in order to carry out this business model. And really go through all the important points. That is, as if you were talking to a professional. He tells you about the limited target audience because in the North Pole there are very few people. That there are problems with logistics and supplies. And here on the other hand, in another section he tells us about potential strategies that we could use. Unique experience, sustainability, strategic associations, etc. Basically, he can guide you if you have a problem, if you have a doubt about your business. How can you solve that problem or doubt from the perspective of a professional creating this custom GVT. For the last two blocks that remain within the section of how to make money with GVT chat, we are going to see two aspects that, although they are not directly related to making money, they are indirectly. The first of these two that we are going to see is productivity. And we are going to see some tools related to that. The first one we have here is the productivity coach. Basically this, what we are going to do is that we are going to see how to deal with that, with productivity. We can ask him, for example, I drink too much coffee. My problem is that if I stop drinking it, I am not so productive. I would like to go down from three cups to one a day without lowering my productivity levels. So, with this prompt, we hit enter. Look, I'm going to listen to him and I'm going to hydrate. We can reduce it to two cups by drinking half a cup the first week, that is, half a cup less, the second cup less. Then week three tells us to start replacing part of the coffee with green or black tea. A pretty interesting strategy to reduce caffeine consumption. Complementary strategies, hydration, active rest, sleep well, healthy diet and relaxation techniques. Basically a fairly complete explanation that a human coach could have given us, with coach knowledge. The next tool that we are going to see in that sense is to make mental maps. The mental maps, I'm going to give you an example, are this here. They are basically diagrams that you can make to organize some ideas. Well, we can do this with ChatGPT. We comment that Dalí is not an expert in this, but what this person has done, the one who has created this custom GPT, is to connect his own mental map creation tool with ChatGPT, then for this we are going to open the corresponding GPT, which would be this one here, Diagrams Show Me. And in this specific example, we are going to resume the issue of reducing caffeine intake, and we are going to copy directly the answer that the previous custom GPT has given us, and we are going to tell it to explain it to us in a diagram. So we are going to put this information about how to reduce caffeine intake with a diagram. Ok. And we put the information, as always, in quotation marks. Let's see, why am I not copying it well? Now. Ok. Here we have to allow it, as always, since it is connecting to an external application. Ok, it seems that there is a problem. I had previously tried this and it had worked, so we are going to ask it to try it again simplifying the information. Simplify. Let's see, wait. Simplify. And reduce the information to generate the diagram successfully. We allow it, as before. And now it seems that it has worked, what happens is that, as you can see, it has not done it completely, it has done it on the weekly plan, let's say, although it is quite good, but what we could ask it now is, wow, I also want you to do it on the other sections, on the complementary strategies part, the part of maintaining productivity. So we are going to ask it the same. Let's copy these two things. Let's say, now make another diagram showing the sections of maintaining productivity and complementary strategies. Here again we see what we commented before. This is so that Chatterbox knows which is the most valid answer of the two. Now, as soon as they finish generating, we see. We allow both. Ah, well, I clicked on it by accident. There it is. Ok, well, you remember that two options had appeared. In one of them it seems that it has failed, but in the other one it has worked, and here it appears to us. We see the diagram in full screen and we have it here. Complementary strategies all the time. Well, all the time, all the time. But well, you see that in the end artificial intelligence has its limitations in that sense, but still it has done a pretty good job of illustrating that information in the form of a diagram, in the form of two diagrams specifically. Finally, another tool that we can use for productivity is to generate presentations using artificial intelligence. That is, you can generate PowerPoints without having to write them by hand. You can pass the content, for example, we could pass the same information about the reduction of coffee consumption and it generates a PowerPoint for us. Even so, it is testing the ones that in theory are better, and the two best ones, because they are very simple, today I think it is very limited and that it is not worth it. Even so, if it is something that may interest you, there are pages that allow you to do it much better, which are outside of chatGPT, let's say, as it is here, which I was seeing before, and the truth is that the presentations they generate are much better than those generated by this tool, which are quite simple presentations and with some errors, especially the language, since it is done in English. The last block within this section of how to make money is the productivity block. It is not something that, following the steps one by one, you are going to earn money, it is not a job, let's say, but it is something that indirectly can help you earn more money, since if you are fine with your body and you are fine with your mind, you are going to be more productive, you are going to work better, you are going to get better results, either in your entrepreneurship or in your work in the conventional world, in a way. The last one is this one here, which is a custom GVT, which is called Gymstreak Workout Creator, which can basically help you to create training sessions. If you are not very familiar with everything that is the world of the gym, this can help you. You give a few indications about yourself, even if you are familiar, it can also give you some advice and in this way it can generate a training plan. In this case, I already have a program here prepared in which I basically tell you this. One by muscle group, I tell you that I train every day, back, biceps, pectoral, triceps, shoulder and leg. I want to lose weight, I do 4 series of 10 repetitions, I want my workouts to last around an hour and a half, make me a training plan. And well, I already finished with day 1, it is now generating the part of day 2, but we are already seeing that it has done quite well, that is, in the first instance we have asked that we want to train 4 days, I see that it is dividing it well, day 1 back, biceps, in that perfect sense, then in terms of how long the workouts last, maybe it is true that I will miss an exercise as long as it reaches the hour and a half, even 2 exercises, in that sense I think it has been very short, since with 4 exercises of 4 series of 10 repetitions, we do not reach the requirement that I had put, that I want my workouts to last an hour and a half. But hey, this would be a matter of simply adding this in the next program, that is, add a couple of exercises in each of these workouts. And then in terms of the exercises, quite well, because apart from having respected what I told him that I like to train with 4 series of 10 repetitions, he has been pointing out several benefits that each of the exercises has, so that you are aware of what you are exercising, which part you are working on, because it is a good exercise, etc. So, on this side, quite well, since it is true that he has not done it perfectly, he would have missed the fact of making the workout last an hour and a half, but we could always ask him now, to add more exercises. So, having said this, we go with the next one, which is also a coach, but this time of nutrition, that is, everything related to what we did with personal development. In this case, it is this one here, which is called Nutrition Pro, and basically what this is going to do is something similar, but focused on food, that is, we can prepare a PROM in which we give a brief description about ourselves, that is, from a chair, I want to lose weight, maintaining the maximum muscle mass, and I want to cook the minimum, and cheap recipes. So, we tell him this, and let's see what Nutrition Pro does for us. Ok, well, he has already finished generating the answer, and he has done it quite well, because basically, based on the information I have given him, that is, the main feature is that I weigh 74 kilos, that I train from time to time, and that I want to lose weight to 70, he has drawn the conclusion that, in the first instance, I have an average of 2,000 to 2,300 calories. Based on those calories, he has divided the different macronutrients, on that perfect side, and he has even made the food plan, with different recipes, and with the energy contributions of each of them, and the amounts that should be, let's say, to adapt to these, to these caloric requirements that he has previously calculated. So, on that perfect side, lunch, chicken salad, 150 grams of chicken breast, etc., etc. So, on that perfect side, unlike the other one that had a small error, in this case, in the case of Nutrition Pro, it has been done perfectly. Having said that, let's go with the next tool, which would be up to here, Top 1% Sleep Performance, which is basically a coach of something quite important also in personal development, which would be the issue of having a good sleep, since it is no longer useful to train well and eat well if you do not sleep or sleep little. In that sense, we have this custom palette here, which is worth redundancy, questions related to sleep. In this case, to do the test, I have prepared this prompt here, in which we basically ask him from what time should I stop using the mobile if the next day I want to wake up at 7, from what time should I stop eating and from what time should I relax and finish my exercise session. We can even tell him that other things should stop doing a few hours before going to sleep. Ok, he is already finishing generating the answer, but basically he tells us to use blue light blocking glasses, which is what keeps you a little awake from 7, quite well as a recommendation. He also says not to eat after 6. This seems quite heavy to me, but I understand that if he says it out there in Johnson that now we will look for who it is, it must be a good recommendation. And then regarding exercise, he says a bit the same, that at least 3 hours before going to bed, caffeine, obviously, it is not enough to have a coffee before going to bed. Relaxation practices, good. And a sleepy environment, good. I mean, having a quiet, dark room, all this quite out of the box, but with respect to the rest, quite interesting. And if we look for this person, Brian Johnson. Ok, in this case, based on his answers about the information provided by Brian Johnson, who is now a YouTuber, but before he was a businessman, he sold surprises if I remember correctly. For example, the longevity, that is, what are the optimal habits to live as long as possible. That is why some of the tips were a bit extreme, like not eating before, I mean, not eating after 6, but basically because this guy optimizes everything to the maximum, he does a thousand analyzes, takes a thousand vitamins supplements and takes a thousand measures in order to live as long as possible. In fact, I think he is about 50 years old and his examples are quite extreme, but it is good that he gives you the advice of someone who is taking it to practice and who seems to have scientific evidence behind it. And the last tool that we are going to see in this block and therefore also within the course since this is the last block, is this one called Books, which would be something like a bookseeker, that is, here we have it, you can tell him what types of books you like, what category, or you can even tell him, well, I read this story, I like to read books, so to try it I have prepared this prompt to see how it works, and basically I have said the following, I recently read Astrophysics for People in a Hurry by Neil deGrasse Tyson, who is a very famous science publisher in the United States, and I liked it a lot, recommend me three similar books, we hit enter, ok, and I have finished here to generate the answer, I had already tried it before, that is, I already knew that it worked quite well in terms of books, or that I already knew, or that I already knew the author, or that I had already read, as is the case of this one, then in that sense, quite well, because I can tell you firsthand that this Custom GPT works excellent and it can be very useful if what you are looking for is a new reading, or even educational books that can help you develop a new ability I am going to review the latter and I see that no one is talking about it and the truth is that it seems very interesting to me and I do not know how it is exploiting us more, and you can make money creating Custom GPTs, that is, creating these chats that we have seen so far, you can generate income, two examples, the truth is that I found them preparing this video, would be that of Finance Wizard and that of Gym Strike Workout Creator, basically what these people have done here we had, we could say instructions and such, because what they have done, what I understand that they have done is within the instructions they have asked their Custom GPT to put links to other applications, or references links, or links to applications and services that it is trying to sell, if you look around here, in Finance Wizard we have asked this and in the end it has told us for more advice on how to operate with options click on this link and you will see that there are different types of services or platforms that these people are using either to sell or to create a community, etc. that is, they are leveraging the fact of creating these tools that people use later and that in fact, this one in particular is used by many people, we can see here that it has more than 200,000 conversations, that is, 200,000 chats have been opened with this tool in order to capture people and carry out a more detailed segment of your workouts as well as calories and macros I invite you to download the app Gymstrike in the following link and as you can see here we get the app to download How can you earn your money? Doing exactly the same as these people but taking your niche imagine that you want a YouTube channel and you want to make it grow you can prepare a Custom GPT of the niche that your YouTube channel deals with and at the end of everything if you want to make dog food you can prepare a Custom GPT that talks about dog food and that is trained with information about dog food in order to redirect people to your store later at the end of everything you put some instructions to your Custom GPT that says at the end of each message write this text and put your link to your store so nothing, honestly a very interesting way and as a conclusion we are going to review the future updates that are coming in the chatGPT ecosystem and within the OpenAI company basically because I think it is important to have them present especially if chatGPT is going to be an important part of your income source or your business in order to update them as they come out or maybe to pivot your business taking into account that in the future this other update will come out and take all that into account the first of all and perhaps the most important is the new update of chatGPT I don't know if you remember that behind chatGPT we are saying that there is an AI model that before was GPT-1, then GPT-2, GPT-3 right now it is GPT-4 and in the future, between the second and the third quarter of this year GPT-5 should come out the extra features that GPT-5 brings us to make it faster as well as all its previous versions it has greater understanding capacity basically it is trained with more data in the case of GPT-4 in the first positive view that we saw that it is trained with 1.76 trillion parameters in the case of GPT-5 it will be 5 trillion that is, it is not an increase as important as the previous ones that we talked about multiplied by 10 multiplied by 10 and multiplied by 10 again this time we multiply the power by 3 but it is still a very substantial improvement then it will also improve the data extraction capacity basically that we have seen before I don't know if you remember if the quote had actually been extracted from the pdf and the commas were not removed which is a detail but it is still a mistake in theory with GPT-5 these things should not happen since it will have an impolite data extraction and that will also improve the custom GPTs because basically when you upload some files to train since in the end you are not training but you are giving context to a custom GT when extracting data from that information that you have uploaded it will have its integration with IoT devices Internet of Things devices for example your Alexa or your Smart TV all these devices that use the internet in a passive way will have integrations with GPT-5 this for example can lead to having artificial intelligence devices such as Siri or Alexa or some of these that directly work with GPT chat and have many integrated functions a bit the same as we have seen with the mobile and finally they also say that it brings improvements with the content creation so that is a bit ambiguous it is to be seen but I understand that it will go in the line of video creation and why do I think it will go in the line of video creation? basically because of the next update that comes within OpenAI which is one of the biggest we have seen to date which is Sora which will come out between the next few months with the avatar that spoke badly he spoke in English and he spoke weird and put stock videos no, here we are talking about generating videos 100% from scratch in fact we have already seen it in the introduction of this video with this same video but basically the videos it generates are a barbarity to date this artificial intelligence is already functional that is, it is already generating videos as we commented this mainly because although it can already generate videos on the one hand it may be that this is what is known as cherry picking cherry picking is an English expression that basically means that you are choosing from all the videos that have been generated the best so if it is the case and you have to generate a thousand videos in order to reach this result it means that it is not ready that a little development is missing in order to optimize the code and that it is not so expensive at the computational level to generate this type of videos that as you can see the truth is that they are quite incredible it looks like a real video there is no difference with reality some of the interesting characteristics that it has, that I have put them here also in a link is this here called video to video which is basically that you you can pass a video and you have to change the settings to make it look like it is in a jungle then it goes from being this video to being this other video which is basically a very similar car but in a jungle or here he asked me to do it in 1920 if you look at it, the curves are the same it is all very similar but it changes the whole scenario I have done it because in a first instance here we talk about what would be text to video as we are seeing here on Sora's website but here it would be something completely different that as far as I know it would be something like this even so, as I said surely the reason why it is being delayed or why they do not want to release it to the market rather, despite having these videos that are being generated perfectly is because it must be quite expensive let's say at the computational level it must require a lot of computing power and therefore the people who wanted to use it should pay a very high price that said, even when it comes out it is expected that the cost will be quite high it is estimated around 12 cents per frame this is based on community speculations there is nothing confirmed and that would mean that if you are paying 12 cents per frame that is, 12 cents for each frame because there is an image that composes each of the seconds of the video it would mean that for 5 minutes you should be paying $2,160 if we talk about the video being at 60 fps that is, at 60 frames at 60 frames per second therefore, it will not be cheap it will be quite expensive but well, if they manage to optimize it enough it can be affordable for most mortals and if they have managed to put it in the same subscription as the rest of OpenAI tools that is, put it inside XGBT as they have done with DALI much better finally, something to contemplate that the future of artificial intelligence is that the trend is to try to reach something known as AGI in English, Artificial General Intelligence which would mean General Artificial Intelligence and it would be something like the Definitive Artificial Intelligence Why is it definitive? Basically because it is better than a human in any task, to understand, learn and reason that is, it not only has the goal to solve your doubts but also to feed back into your knowledge it can have either an autonomous learning or supervised but basically, based on its interactions with humans and based on the knowledge it has the idea is that it goes self-programming and self-improving based on the AI we have today and better comprehension although in this aspect we are quite advanced too in theory, the interaction with this type of AI would be perfect that is, even better than any human it would understand you better than your brother, your sister or any member of your family basically because it would have almost complete information about everything that happens and on the other hand, it would be autonomous and it would be radical for humanity since we were talking about that we go from the computational power quite limited in our brains as I said, it is difficult for us to make a somewhat complex multiplication but a mobile calculator which is not a very powerful device does it in a moment so basically we would be applying this to any type of reasoning but as a result of being such an unknown field there are many people who consider it dangerous not only in the sense that we can all imagine in the first instance in movies like Matrix or Terminator but more in the sense maybe even that there are politicians or companies that can use these powerful tools to somehow control us in this sense, if someone is interested in this more philosophical aspect of artificial intelligence I recommend you to watch the podcast that Lex Fieldman gave to Eliezer Yuvkovsky who is a publicist and computer engineer very focused on everything that is, I have it here three hours of pure content related to all this of the philosophy applied to artificial intelligence very interesting this man specifically is more in favor of limiting all progress related to artificial intelligence due to potential dangers but in the same way that there is this person who is a crack in his field advocating for that and I highly recommend him since he has very interesting conversations in which he defends the completely opposite position although in this case he is a bit biased since we are talking about the CEO of the 3D company ChatGPT With all this explained I can only thank you for having listened to this course I hope it has served you I hope you learn a lot and see you in the next video