Transcription
Hello, viewers. Today I want to discuss the pros and cons of BKLM for those who repackage content and analyze implicit connections between a large number of sources. That is, for corporate analysts, as well as for those who regularly create some kind of text content, or content for self-discovery, for self-study, perhaps some presentations, lectures. So. And first of all, I would like to discuss the cons, uh, and the possible limitations of the service at the moment, at the end of 2025. Let's move on to how we can upload information into notbклам. That is, in the very structure of BкLM, we see the possibility to upload either files, the formats of which you see here. In the future, I promise to also provide the ability to upload these Excel tables. Or we can simply paste text, or upload, uh, then paste a link to a YouTube video or website. And then, as Notebooklm promises, the service itself will go to the website or YouTube video and get some information from there. But what actually happens when we provide, for example, a link to a YouTube video? It goes to the video transcription section and takes the video text. For example, it doesn't take comments to the video. And, uh, if you want, for example, what I did, right? So, here we have a note, mm, here I uploaded, I was just curious how users react to the videos made by a well-known author, who reviews the most popular neural network services. I was interested in how users react to her videos, because, for example, the Web3 Unity channel doesn't make me want to subscribe, but nevertheless, she has a lot of views, and I was interested in what I did. I manually copied 15 links to her most popular videos and, uh, created 10 Marku фай files, which you can see here, right? And basically, in such a messy form, I just copied the posts from the browser. The first 15 posts under each video. I named each file the same as the video it's linked to. And only then, only then does the note, which receives links to videos as context, uh, video reviews of neural network services, only then does this note gain access to the video comments. We must, uh, provide the comments in a separate file, otherwise the service may hallucinate and not provide relevant data on that. So, for example, I asked a simple question: analyze how users reacted to the content released on the Web3 Unity channel. And here, in principle, there are some interesting points that I didn't think of, didn't notice. And based on these points, I can, for example, try to build hypotheses on how to lure away some of the audience that this media reviewer has. What is even more interesting, we can, for example, provide access for the neural network, well, for notetobklm to content that is in Telegram. How can we do this? Let's go to Telegram, for example. Uh, here we are, in the neural networks chat, and in the a-version of the Telegram client, there is an option to export the chat history. Export the chat history in HTML format, right, I'm already starting to stutter. I exported HTML, after which I converted the resulting 13 HTML files to PDF using online converters that you can find yourself. Literally type into the history, well, literally type into the search, how to convert HTML to PDF, and you get such a PDF in the end, with, uh, posts from the Telegram channel. I downloaded 12,000 12,000 messages from TG. And what did I do? For example, I asked it to continue. I've already chatted a bit here, right? Let's remove it now, let's zoom in. First, I asked it to create a series of posts based on what was discussed in this Telegram channel, and in that chat, they mainly discussed the application of Vipeкод tools and where to get the best subscriptions. So. Uh, here's the prompt: create a series of educational posts, and I got some raw, unprocessed, perhaps rough content, but nevertheless, each answer is based on the information available in the Telegram chat. In principle, you can also download, not only Telegram chats, but also export Telegram channels in HTML format, then convert to PDF and upload to notbook. And this allows, first of all, what does it allow? It allows creating similar content, it allows for a deeper immersion into the history of a Telegram channel, into the history of a specific expert field. And after all, we cannot simply provide a link to Telegram. Yes, we essentially cannot provide a link to Telegram withinma. I tried to upload a link, and I got an error, and this tells us that we can interact and analyze content through the Notabklam service, which, in essence, acts as a universal raksystem for finding information within a large number of documents. But we need to export information from Telegram in HTML format and then convert it to PDF. Also, within the service, there is an option to write queries yourself so that the service pulls sources from the internet. But my experience shows that the links that Notebook LM pulls for my query are better and of higher quality. First, ask a specialized service for searching information on the internet, for example, Perplexity, to collect information for me, right? Uh, here we go to Perplexти, for example, take some question, and we can, uh, download the links from here for the query we are looking for, and then transfer these links based on our query to notbook. And this will be more effective than asking notbook to compile information based on our query. Also, I recently watched a video, I also recently discovered an interesting video on this author's channel, dedicated to, uh, notebukl, which came out 12 days ago. The author seems to position himself as an expert in self-learning, in information assimilation techniques, and he gave NotebookLM a final score of three out of ten, which greatly surprised me. And his main point, if his video is 43 minutes long about how good or bad the Notebook LM service is. And his main point is that, as I said, Notbook LM provides a superficial understanding of the material. It pulls out some pieces of information, composes them. At the same time, the functionality of the mobile application, which creates decent audio podcasts, is not as good if you upload a large number of sources, and in particular, he criticizes notkl, that for those who want to truly delve deeply into the material, they need to use, well, get acquainted with, well, spend a lot of time familiarizing themselves with all the sources themselves, rather than studying something through Bukm. And here there is quite, and indeed, a lot of criticism, well, in my opinion, relevant criticism. And after, probably, my two months of active use of notbuklm, that is, I created five to six audio summaries every day. I also have a favorite prompt, it's quite simple, that is, I don't upload many sources, I try to study one source as deeply as possible. For example, I upload a book and ask it to summarize it, well, even on a topic that I already know more or less. I upload a book and ask it to summarize each chapter of the book for someone who is already deeply immersed in the topic. That is, literally a prompt of one or two sentences. And the point is, summarize each chapter of the book for someone who is already in the know. And when I uploaded, of course, a book on a topic that I am not familiar with at all, zero, I understood nothing. That is, I created a podcast, listened to the first 40 seconds and was put off by the fact that I didn't understand a single word. And this highlights several problems. First, do not upload too many sources into one note. I am currently working on a summary of Ilya Sutskever's biography. Uh, yes, I'll show you with this wonderful cover. I hope I will eventually release a video with this cover. I hope so. And, essentially, what I encountered. I tried to upload the entire biography of Ilya Sutskever into one note, and it turned out to be a mess, of course. Although I want to highlight four segments from Ilya's biography that I want to talk about. What did I do? I divided these periods that I want to talk about into several notes. Only then did I achieve a more competent result. Therefore, I do not fully agree with the author that the notblm service is a bad service for self-education. By the way, the author correctly notes that, well, flashcards for self-learning are not so beneficial to create. That is, to briefly summarize his main point, that, mm, indeed, question-answer flashcards are only good when you have the opportunity to return to the cards that cause you difficulty. And at the moment, at the end of 2025, notebookm does not offer the ability to sort cards. It, first of all, does not offer to export the resulting deck to any application. Secondly, it does not offer any mechanism for sorting cards. That is, for example, a question causes you difficulty, and you want to return to it the next day, not in 2 days. And in this Bukl, it will not help you in any way. Therefore, for self-checking, it is better to create, of course, quizzes. Also, yes, we do not upload too many sources into a note. We first study the sources, think about which ones are more or less suitable, which ones are not. And only then, when we have manually selected the sources ourselves, yes, if necessary, we have downloaded the chat from Telegram, if necessary, we have attached the comments as a Markd file ourselves. Also, in the future, they promise to add the ability to work with tables. In principle, you can already build some tables now, yes, you can already build, returning to the analysis of the Telegram channel, you can already build some tables, ask it to create a table of trends, but for now it creates a table in Markdown format. If we export it by copying and pasting, it will be Markdown, not XML. And in the future, Google has directly stated that it will add the ability to work with tables. Also, what else do I want to say? Regarding Google nano, well, notoglm has added the ability to create infographics. Infographics are created one unit of infographic at a time. Or we can create a deck, a deck of slides. And with a deck of slides, it can turn out interestingly. Look at the subtlety. it creates one slide per source that you uploaded. I didn't quite understand it at first, but then I did. And in general, this can simplify work with presentations. That is, I created a presentation on working with, well, a presentation on the history of the chat and, yes, discussing a presentation on the history of the chat, discussing Vipec tools. And, well, in general, it's quite, quite nice. Uh, it turns out that each file I uploaded, that is, here are 13 files, 13, okay, the beginning will connect. Okay, yes, I uploaded 12 sources, and it created 13 slides. Or it can create, that is, you can ask it to create some unit of infographic, and it will create one image for a specific request. That is, you can upload several tools, well, several sources and, for example, ask: "Create an infographic for me on how Cod is compared? Well, how is one specific coding tool contrasted with another specific vipingtool?" That is, you can, mm, create small tasks for a specific, well, specific infographic. You can create several infographic units, assigning each new infographic its own small task. So. And I also can't help but say that, first of all, they improved the audio version, they added more customization, yes, they added styles in which it will be, well, in what style the audio summary of the sources you uploaded will be created. Again, we don't upload too many sources, because then it can turn into a mess, because the LM itself will decide which sources are important, which are not. And, uh, especially for content creators, the create report function is important. Essentially, it's the ability to create an essay, uh, or a series of tweets, or some textual, well, a textual unit of content based on the sources you uploaded. Both pre-proposed formats with some pre-configured prompts are available. That is, the formats that it offers you to choose from with some classic headings are simply formats for which a prompt has already been written, and the prompt may not match your goals. That is, you can choose the most correct create format and, for example, ask: "Create a series of tweets for me based on the sources I created. Create marketing campaign ideas for me with text headlines for my company. And all that sort of thing." Therefore, we do not rely in most cases on the formats offered by BKLM, but create our own prompt pursuing our goals. And this is truly, well, a very good, very good tool, both for a new perspective on the sources you already have. It's just really nice to see how notuklam processes information and interacts with it. As well as just, you know, I don't know, you need to push some internet resource, you don't particularly want to push it, and you're like: "Bang me, please, based on the sources I uploaded, bang me a series of LinkedIn posts." A good prompt for such an audience with such a limited character count. Go ahead, dear. I won't even read it. I'll just publish it under my name and forget about it. I'll have some activity and everything will be fine. So, to summarize, at the moment there are, uh, eight ways to create some kind of content. Flashcards are not so interesting, as they do not allow validating which questions cause you difficulty, which do not. So from the phone, we use audio summaries to, as I already said in the previous video, better learn, for example, all the chapters of some book. And in the computer version, we create, like, you know, a content factory, we create infographics, that is, pictures for posts, we create presentations. Just an easy way to create a presentation on the topics you need. Uh, we create quizzes. Well, and, yes, the most important thing, infographics and creating Create your own, uh, well, and, like, creating textual content based on the prompt you need. So. All this is available both in the free version. And in the free version, you can, if you have a VPN, I've already said this, but there is such a service, yes. Apapap. Where is it? Yes, in general, there is such a service in Telegram, yes. Therefore, I hope this video was useful for you. If it was useful for you, I would be glad to know what services you use and what service you use that no one has reviewed yet, and you yourself are just itching to tell others about this service. I also hope that you will cherish your time and attention. So, see you soon.