Transcription
Hello. Today's episode is without Taras. Here I am. What are you talking about? Good to see. Okay, let's go. Most people use their brain 14% or seven. Or seven, and ChatGPT uses even less, five. Let's help these people do something about it. No, then just watch this episode until the end, and you'll learn how to use ChatGPT to the fullest, because we'll give you 44 awesome ways to take over the world. No, not Pinky and the Brain. Most people use ChatGPT at most 5%. Then they wonder why neural networks don't impress them when they just use them as a search engine. At best, an image generator, yes. But during our daily communication with ChatGPT and a bunch of other neural networks, we've collected 44 ways that really boost efficiency and will help you in work, everyday life, studies, and in everything, absolutely everything. Let's go. Let's go. Okay, the first thing you can do is recognize text from an image. A simple story. We take an image with Russian text and ask: "Recognize the text in the image and write it down." We see how the text is generated. In fact, the text should be exactly the same as on the picture. And so it is. It didn't even neglect the short dash. Well, it lost a couple of letters X, by the way. But overall, it's good. Besides this, what else can you ask? You can ask not only to recognize, but also to translate it. Recognize the text in the image and translate it into Russian. Feature number two. I remember the times when I used to photograph signs in incomprehensible foreign languages, and the DeepL app, I think it was one of the first translators that worked on neural networks, translated much better than Google. And so, they also recognized from photos. Now, the same thing can be done with ChatGPT. Use the mobile app, take pictures of signs, take pictures of words in a foreign language if you don't know it or don't understand it, and want to quickly find a translation, it will help, yes. And if you live in Russia and don't know how to install the ChatGPT mobile app on your phone, then subscribe to the channel. We will soon release a video on how to do it. Feature number three. Let's continue with photos. You can ask it to describe what's in the photo. It's interesting in what case this point will be useful to you if you have a task to describe something in a photo, and you don't want to describe it yourself, as an option. Give me a real-life situation where this might be useful. For example, if you're shooting a video of 44 interesting ways to use ChatGPT, I think it might be useful if you use some visual images at the moment when you're writing something and you just need to describe what's happening. Well, it can be useful for those who create content, for example, for those who write something for websites and so on. Why not? Or, if you are a photographer, you need to describe a photo in your Instagram, make a caption. Uh-huh. Well, you can ask it to do it in a specific style, add hashtags, something else. In principle, why not. Next, from the more cool practical things, well, it's also related to photos. Here we'll add a diagram. Make a diagram from the image in Mermaid format, which is a format that can be imported. You can, for example, if you drew a diagram on a board, take a picture of it and ask it to digitize it. This is also a practical thing. If you draw something on a board at work and then need to convert it to Mermaid or somewhere else, then such a thing will definitely help you speed up and simplify your life significantly. This, of course, requires more time, because ChatGPT needs to think. And so, it draws it with text. Then it can be imported anywhere. If you have handwritten text, ChatGPT can also handle it, let's hope. Write the text that is written in the image. Handwritten text in Russian. I found a hardcore picture. Let's see. Word for word. Look, can you read what's written here? Report GPT can. I think it wrote it incorrectly. Why? Oh, how many wonderful discoveries await us, how many wonderful discoveries await us, the spirit of enlightenment prepares. Aha. There's an error here, by the way. Enlightenment is written with a "y", and it corrected it along the way, and experience, the son of difficult mistakes, and genius. Ask it if it corrected the error. Did you correct an error? It needs to think. Let it think. Let's move on. It checks what it did, in general. Yes, I slightly corrected the line. Instead of "enlightenment," I wrote "of enlightenment." See, what a good boy. Out of habit. Out of habit, I corrected one line. I'm used to correcting after my comrades. Let's move on. If you need to generate alt tags, meaningful file names for images, and so on, ChatGPT can also help you with this. Well, this is if you are involved in SEO optimization and similar things. Accordingly, upload a photo, ask it to generate alt tags, meaningful file names for these images. Well, I uploaded one image. You can upload several, by the way. You can upload ten. And it decided for all images. Well, an island on a lake at sunset, of course, interests us. Island on a lake, sunset VP. Silhouette of an island with trees, blah-blah-blah. Here's a cool thing. It even did more than I wanted. Excellent. Let's insert a thick PDF. Some number of pages. Highlight the sections. Make a table of contents with brief theses for each chapter. Let's go. Nine tips for mastering the game of nets. Master the speed of the sword, the direction of the will, and reading the court. Excellent. Can you make a detailed decision tree for these nine tips? It still won't say that the five-point system answers. Yes. Now what can we do? Uh, tip number eight. We can ask it to extract pages, glue them together into a new PDF. Remove the first and third pages from the file and send the PDF without them. Here it runs Python code to prepare the file. Accordingly, it knows how to use external tools and will provide a PDF. Let's see, download. Bam, it seems to have handled it, probably. Let's move on. So, the next cool case. You can convert a PDF table into an Excel file. We upload a PDF, a to-do list, and ask: "Extract the table and save everything in CSV." Let's go. It runs Python tools again. It's easier and better for it to work this way, more reliable. And so, here's our CSV file. It opened in Numbers. Everything is there, the to-do list, dates. And now I'll show you more. Well, it was prettier, of course, but all the text is extracted. The date is extracted and the task is there. Everything is cool, it handles it. Let's make a copy based on our CSV files and paste some text into it. Like this, for example, 1009. Extra task. Now you can ask it to upload two files. We upload two files and ask the following: "Compare the two versions and highlight what has been added, deleted, and give a summary of the meaning." Let's see. It uploaded both versions. What will the result be? Mmm, it's running code. Place your bets. I don't think AI will kill humanity. So, changes. Here it separately writes apologies, says it compared two versions, one new task appeared. Extra task deleted. No semantic change. An additional item has been added to the to-do list that was not there before. This is very convenient if you are sent, for example, contract updates, with comments, or something else. You can just upload two versions and write what the difference is. Let's move on. So, let's take our previous book and do the following. Break the text into eight Telegram posts. For each, create three headlines and a CTA (Call to Action). Let's go. Content creation has become easier than ever. Here are the short posts. True, it only created headlines. It skipped writing the actual post texts. Yes, yes, write the posts themselves. Format: headline, post, networks. It started writing posts that you can publish in your Telegram channel today. Everything is absolutely wonderful. Read the book, did everything. You don't need any AI. Got it? It works, and that's fine. It works, and that's fine. So, let's move on. What else cool can be done? You can generate a QR code. Generate a QR code with text for me. And what should we write? For what with text? With an image of a monkey showing a middle finger. That's a QR code. Yes, I know. Why would they need an image? You point your phone at it, and there's an image of a monkey. Can it do that? No, it doesn't work like that. You can put a link to an image with a monkey. Well, then give me it. Okay. Let's generate a QR code with the text: "Hello, world." Well, okay, okay, let's do it with the text "Hello, world." It runs Python to generate everything. You don't have to write Python yourself. ChatGPT does everything for you. And we download it. We open it. Bam. It works. Point your iPhone at it, and now it will steal all your bitcoins. Hello, world. Hello, world. It works. Cool. Everything, let's move on. What else besides this can be done? You can also do exports. If you have a lot of text, collect the text into a Word document and give it the text. Let's see what document comes out. These are the same texts that it generated for the posts earlier. It assembles, says, let's download. Let's go. Downloaded. Let's see what's inside. Bam. Beautiful. It even marked up the headings and so on. Isn't it a good boy? You can also export not only to Docx, but also to PDF. Now export everything to PDF. It can handle it. It can handle it. You know everything about it. I just believe in it. No. And on the third try, yes, it says it exported all posts to PDF. Let's see what happened. Along the way, it didn't manage everything. It couldn't put everything in as beautifully as in the Word file, and it lost those emojis, but everything is still there, everything is in the PDF, as requested. Next, from the interesting stuff. You can work with time, schedules, planning in a cool way. In fact, it's good at time zones. For example, I schedule calls with people from different cities: Porto, Dubai, Moscow, and Beijing. Find suitable time slots so that no one has a late night or too early morning. Very realistic, I must say. Yes, damn. Well, I would take Alaska. Alaska is a good place for a meeting. It says that the optimal option is to have calls from 8:00 to 12:00 Porto time. That's 11:00 to 15:00 Dubai time, 10:00 to 14:00 Moscow time, and 15:00 to 19:00 Beijing time. Well, do you agree at all? Yes. Generally yes, it seems to add up. What else besides this can you ask? You can ask: "Create an ICS event for Google Calendar." For this event, it's an important meeting. It will set the time itself based on the time zone it sees in the browser, as far as I understand. So this also works coolly. It runs Python, as usual, checks that the time is UTC based, adds it to the file, created an event. August 12, 2025, it says: "Download, import." Everything works. I won't show my calendar. Trust that it works. I've done it a million times. You can work with tasks in a cool way. By the way, I talked about this in detail at my webinar. You can buy it in recording via the link in the description. You can plan tasks, break them down into sprints, put them into timeboxes, and so on. Well, for this, you need tasks, right? What tasks do you have? And what would you like? I don't know what you want at all. For a sprint, 10 hours. Let's break down the task. It was already in your, in your master class. Maybe we can skip it. What tasks do you have? Come on. What are you doing? Write a book. Write a book? No, that's too much. You won't manage it in 10 hours. Well, I'll try. Let's write a book. Write the first chapter of my own book on finance. What other task is there? Start running in the mornings. What else do you want? And what else do you want? I want to earn a million dollars. A million dollars. What else do you need to do in these 10 hours? Cook soup. Cook soup. Well, a standard housewife's routine. So, the Ice formula. Impact, Confidence, Ease. It even collects a score, look. Write the first chapter of a book on finance. A significant step in a long-term project. Realistic to do in a sprint. Will require concentration. Start running in the mornings, improve health and discipline. Easy to start. A simple first step. Earn a million. Huge impact. Too ambitious for one sprint. It says you won't succeed. Why? Too ambitious for one sprint and extremely difficult quickly. Prioritization by Ice. First, start running in the mornings, then cook soup, then write one chapter of the book, then you'll earn a million dollars, that's like a normal day. No, actually, it only broke down the first three tasks. It considered the last one unachievable. Well, logically, more or less fair. Write how long it took you to earn your first million dollars, in 10 hours or less. Besides this, by the way, you can ask it to collect checklists and habit trackers. Collect a monthly habit tracker for me in Markdown with a progress bar. I want to adopt the habits of reading a book, running, exercising, and what else do you want? Playing the piano. And playing the piano. Where is your piano? On the way. On the way. In general, here is a habit tracker in Markdown. For each area, for each day, there's a mark to put completion. All this can be downloaded and put into Google Sheets or Notion, and you'll have your own tracker. So, now let's talk a bit about video. In fact, you can upload a video that I downloaded, ah, and I thought it was a video, and ask it to extract the audio and save it in MP3 format. Done. Let's check if it will work. The MP3 file is there. Apple Music launched. I still don't believe that casting was done and it works. What are you trying to do? I exported the MP3 from the video, so it can also do that. So you uploaded the video and extracted the MP3 from it? Yes. Do you want a life hack from Evgeny for this video? Yes. Ask ChatGPT to write you a script in Shortcuts. This is an app on Apple. It's interesting, how many of our listeners have Apple computers? How I use it, for example. I use Shortcuts, which allows automating various things, something like N8N, which is built directly into Apple, well, automation, basic automation. Okay, what to do? So, you can write what script? And there you can also schematically set up flows. And here, for example, one of the options is to run scripts. Of course, Shell, extract from the video, which I'm opening here. Yes, the audio track MP3. When you execute this, it will offer you to upload this video or select it. Write a ready-made Shortcut file. I don't really believe it will work, but okay. Finder couldn't handle it. Well, to be fair, I think it's because of the lack of understanding. I'll ask again. Write a script for Apple Shortcuts for extracting MP3 tracks from video. So, nevertheless, Shortcut wrote a native shortcut through the share command app. MacOS shortcuts fm fast flexible batch requires installed ffmpeg. So it offers a script. I don't know if you did it this way, if you generated these shortcuts. It doesn't look good. It looks like some kind of hassle. I can run it on my end and show you how it works. Yes, well, of course, you need a codec to cut the audio track from videos and then recognize the text in it. No, well, you see, it says there's a native shortcut that can be done without a codec. I think without a codec, it's a subroutine that can't be launched. But yes, and it recommends it, but ffmpeg will be there, but at least this can be imported into the system. Well, in short, it works. Okay. Let's test 3D modeling. Create a 3D model of a cube for me in OBJ format. It can also create a cube. It runs Python. In general, it works. Everything seems okay. By the way, how to check, I don't know. I'm not a 3D modeler. But if you are a 3D modeler, tell me how good everything is here. I think everything is wonderful. You can create something specific from any text in a specific format, like for Excel, something for Markdown, or for LaTeX itself. If you write something mathematical, write me a LaTeX formula for or for the Schrödinger equation. I think it can handle it, yes. I actually launched the rendering. I don't think it's needed. Aha. You can display it. Let's see. Can it be done like this? I don't know. It converts equations into images. Well, by the way, it drew something here, but it's naturally interpreting it there. WolframAlpha. It renders the image like this, and LaTeX does it. If you need something like this, it also works. By the way, it even says it will save it to PDF, although I didn't even ask for it. Let's see. Well, look, beauty. How can we do without wave functions? It made a PDF, although I didn't even ask for it. It's highlighted, it works. Oh, from the cool stuff. Make a graph of sales growth by month for 2025, from 10 sales in January to 150 in December. Extrapolate it to 2026 in HTML format. Make it interactive. Naturally, you can upload your own data from Excel or Google Sheets here, and it will also handle it. Accordingly, you can make an interactive graph, or a non-interactive one. Correct. Correct. Two points in one. Let's look at the interactive HTML graph. Downloaded sales. Let's go. Op. Well, beauty. Evgeny, look. And extrapolation. Wonderful. Absolutely. I would even say masterful. Masterful. Nothing less. What else can we do? We can do structured research with quotes from the internet. Turn on research. We can't get far without it. Collect five sources on the topic of cryptocurrency and provide one to two quotes per source and links. Result: a short summary. Well, yes, I'll look for something interesting to read. Aha. New York Post, Guardian, Coin Telegraph, Dec. Here's such a simple thing. Working with maps and Geo. Collect GeoJSON. Three points with a name, description, and marker color. I, damn it, don't know where to import it. Honestly, do you know? I think so. So, in general, it generated GeoJSON with three points. What do we have here? Prasudo, Commerce, Porto, Faro, Old Town. Excellent. All in native Portugal. It set the coordinates. In general, this can be imported and placed on a map. Here's such a thing. Can you generate an HTML map with markers from the list above? If you hate Google and Yandex and need to navigate in difficult conditions, show a taxi driver where to go when GPS and GLONASS are not working. Possibly, ChatGPT can help you out if you have access to ChatGPT. In short, there are moments. Download HTML Map Points. Let's go. Both. Wow, it really placed three points. There's a map. Not Google, Old Town. Prasudo Commerce. It works. Isn't that cool? What else can be done? Let's go back to design. You can create SVG icons by description and with optimization. For example, draw a minimalist microphone icon in SVG format. Size 24x24 with an outline, no fills. Let's go. Rendering has started. Well, it will probably generate it through a script again, I think. Place your bets, gentlemen. Op, the icon image. SVG, look, it did everything. Absolutely a beauty. Do you like it? Yes, I do. Magnificent. What do you think, as part of your large multi-agent system, should you have one agent that only generates icons? Of course, it's very easy to convert from format to format. Now convert it to PNG format. It will handle it easily. Transparent background. Let's see. Here comes the PNG. Open it. Downloaded. And the same icon in PNG. Well, of course, I didn't specify the resolution. It made it this tiny. But still, it handled it. What else can we show? Naturally, you can generate simple scripts. You can ask it to make a script that renames all files in a folder according to the pattern YYYY. Month, day, name, and extension. Write it in Bash. Accordingly, the file does everything. Well, let's open what we had here. File open folder if you use open file crypt m bash, but it's clear that you'll need a little knowledge of the terminal, but is it critical? I don't think so. So, uh, it's offering too much. Wow, you're already, it seems, rewriting the Bash system of the world. Let's check. Op, it renamed all the files according to the pattern. Isn't it a beauty, look. It did everything as it should, as requested. Write in the comments which of the chosen methods you really liked and will use at work, in your personal life, and in everyday life, wherever. Yes, write us also how you might use GPT in your own ways, which we haven't mentioned and which you want to share with the community and GPT enthusiasts. There are, of course, an infinite number of these methods. We could list them for weeks without interruption, but we'll leave that to you, so to speak, for self-study of GPT. Subscribe to the channel, give likes, bells, all that stuff. And until next time. Everyone, bye. [music]