How To Use NotebookLM For Beginners In 2024 (NotebookLM Tutorial)

So in this online tutorial, I'm going to show you all how to use NotebookLM as a beginner so that you can get the most out of this amazing AI tool. So you just want to head on over to the link in the description and you can see that you'll come across to this link where it says, try NotebookLM. You do need to have a Google account to access this software. So just make sure you're either logged in or you've created one and are logged in in order to access NotebookLM. And no, you don't need a Google Pro subscription to access. So now you can see here, it gives us this home screen. We can see that we have this area that says New Notebook. And of course, we've got the example notebooks. Just click New notebook because this is going to be a lot easier. And now we get into the area where we can actually start to use this. So essentially, NotebookLM is an AI system that allows you to have multiple sources of data and then combines those pieces of data into interactive pieces of media that are easy to digest. Going to demonstrate this with a few examples and I'll show you why this is just working wonders. So you can see right here that if we want to upload data and we want to upload any sources, we can see that we firstly have Google Drive. So for those of you that have a Google account, this is going to be really easy for you to upload your details. For those of you that have websites, you can see that there is a link to click here. And of course, any YouTube video that you do want, you can also put that link in there. And if you have any pasted text in you, you can also add that. Now, you're currently limited to 50 sources, although I don't think most of you are going to use 50 sources immediately. So I'm going to actually use this to show you guys a recent example that I was working on and something that I wanted to truly understand. So what I'm now going to do is I'm going to get every single link I can in order to give it the data it needs. So I was working on a video where I was trying to understand a new AI architecture called V jepa. So what I'm gonna do now is get the links from that. So you can see right here, I've typed in V Jepa. And what I can do really easily is I just click this button right here, I can click the share icon, and then I can click copy this link. And then now I go down to the YouTube area and then I click paste YouTube URL and I click insert. And now you can see it immediately gives me a quick summary of exactly what's going on. The very best thing about this is that I can ask this questions and this is interactive, but I'm not done yet. I want more sources because I want to have every piece of available information so that I can truly understand what this architecture does. So if I want to add more sources, all I need to do is click plus and then I'll just go to YouTube again, I'll paste a different link and I'll just insert that source. And now you can see this entire thing is based around two sources. So I also want to actually use the blog post. So I'm going to click add. I'm going to click website. And you can see here, this is the article that actually talks about the AI architecture. So what I'm going to do is I'm just going to paste that website URL in and I'm just going to click insert here. So now that I've got three decent pieces of information here, there's many different things that I could do if I want to quickly realize exactly what's going on. You can see that I just clicked Generate Briefing Doc, and basically what that does is it just briefs me on everything that's going on. So you can see right here that it just gives me the main themes, the key ideas and the facts, the problems, the solutions. And this is something that aggregates all of the data from all of the videos. This is something that is really good because it allows you to not just have one piece of information, it allows you to get smaller pieces of niche information so that you find on the Internet that allows you a more comprehensive view. This is really good for those of you that want to make content quickly and make content that is personalized to you, that is actually great. One of the big problems with AI is that when you ask ChatGPT certain things, it doesn't really know what it's talking about in niche examples. So you can see right here that this is a really comprehensive overview that we do get. And interestingly, you can see on the brief document, it gives us these quotes which are actually in the video, which is really, really nice because if we're doing research or stuff like that, we can use these in our content or realize exact what was said. So now if you click this Notebook guide, of course, it gives you some really easy questions that you can ask. You can see you've got suggested questions and basically it's going to give you suggested questions that it thinks you would like to enjoy. So you can click any of these, and of course I can click this right here and it's going to just immediately ask that question. And it's based off all three of these sources. The best thing about this is that when you're interacting with all of these sources, what it allows you to do is it allows you to look at the direct source from where it is. So for example, if we look at any of these things that we think are interesting and we highlight over them, we can see exactly where that information has come from. For example, for point number seven, we can see exactly where this has come from. If we click this point, you can see that this is based on this right here, which is of course the blog post. So you can see it highlights this entire section where it grabs the data from. This is remarkably useful because you want to always know what kind of data the AI is pulling on in order to make its arguments. And of course you can scroll around and see other areas if they interest you too. So now what's also interesting right here is that if you decide that, okay, I have too many sources on the left hand side, what I need to do is I need to just only work on one source or one video. I can uncheck these two and then I can simply ask it a question. For example, I could say, what's the main problem with LLMs? And then you'll see that it is only going to answer this question based on this specific video. You can see here that it actually talks about all of the stuff that is specifically in this video. So for example, If I click 2 here, you can see exactly where it's taking that data from if you want to scroll down certain bits. For example, If I click 5, you can see it brings me to the exact point where that text is. Now, of course, with videos, they aren't as good as PDFs because there is a bit of non structure to the documents. But it's still a really good way for you to immediately get information from a video. Of course, one of the main features of this AI software is the fact that it can generate really interesting podcasts. One of the things that you can do is generate a podcast based on the information that you've collected. So let's say I wanted to generate a podcast based on all of this data on this new stunning AI architecture. What I could do is I'd do this audio overview. So what I want to do here is I could either immediately click Generate and it's going to talk about all of the content in its own way, or I could click Customize. So when I can click Customize, I can basically say, try to focus on why this architecture is a really good way of, of tackling AGI and why this is perfect in layman's terms for beginners. So now I've clicked Generate, and you can see this is going to take a few minutes to generate. No need to stick around. So the audio is somewhere between five and six minutes long, and it's going to be a podcast between two hosts discussing the information that you've added. Now, of course, you can create certain things here, like, for example, you can create an faq. You can also create, for example, a study guide, and you can also create a table of contents. These are going to be things that help you to understand the exact data of what's going on. So you can see here, I've got all of these notes saved, and this is where I have all of this information that's going to be really useful. Now, another thing that you can do, and I think this is really useful, is I would go to my notes, click them, and then I would just rename them. They all say new note at the beginning. For example, this one is a timeline. So I'm just going to put timeline here and then I'll just put this back and you can see it brings it to the front because you've renamed it. So just click it. And then of course, I'm just going to do FAQs. I really wish it did it itself. But this is just where you can rename them yourself. And I'm just going to put Briefing Doc here. And you can see now all of these are immediately available in a way that's easy to understand, of course, as well. What you want to make sure you do is you actually rename your entire project so that you don't get confused on the main page. So I'm going to title this V. Jepper. And you can see right here that now my product is easily identical. And now you can see that after I've renamed this, my project is easily identifiable. If I click the homepage, you can see that. I can see that V. Jepper is right here. Three sources I click in, and now everything's right there where I like it. Now, another thing that you can do is after you've created these notes, you can go even further. So let's say, for example, we look at the frequently asked questions, we can Say, let's do a right click here and of course we can talk with this specific output. So, for example, I could say, hey, this V. Jepper thing is actually really cool, but what about some related ideas? Let's click this and then, of course, we can say, suggest related ideas. So here you can see it gives us some areas where we can go even deeper on the info information we've already explored. This is good for filling in the gaps where you might not be so sure on exactly what you're learning about. It says, here are some related facts and ideas from your sources that can enhance certain sections. So, for example, you can see here, on the original note, how does V. JEPA relate to advanced machine intelligence? It was a bit small, but in the chat that we just had, you can see that it managed to give us a broader discussion on how this actually works. And most of this is based on our data, which is really good, because even if some of the notes that we get here aren't really good, we can take them and then we can expand on them. For example, we can see that there's this study guide here, but we could say with this study guide, let's click this and let's say add 10 easy beginner questions. So this way what you can do is even after you get your note done, you can iterate on those notes and then get even better content. So now with this basic supernote, I can save this to note, and now I've got another piece of information there that I can then once again interact with. So now as this tutorial has gone on, all I need to do now is just click this button right here, this notebook guide, and then I can easily access this podcast. Surprisingly, this one has gone on to 15 minutes long. But I'm going to show you guys what it sounds like.

Imagine like a robot, right? But it doesn't just, you know, clean up a juice spill. It actually gets why it's a bad idea to just tip over a sippy cup in the first place. That's kind of the common sense AI we're diving into today with vgp.

Yeah, that's a great way to put it.

So vgp.

VGP stands for Video Joint Embedding Predictive Architecture.

Catchy.

But it's all about, you know, teaching AI to learn more like we do. So that's by observing the world around us, making predictions about what's going to happen next, and then, you know, learning from maybe getting it right, maybe getting it wrong.

Okay, I'm into it. So we've Got like a whole stack of research papers here about vgp. And one name that keeps popping up is Yann Lecun. Who is this guy? What's the deal?

So Yann Lecun, he's a pretty big deal in the AI world. Like he won the Turing Award, that's like, you know, the Nobel Prize for computer science.

Okay, so top of the top, top of the top.

And he's actually the chief AI scientist over at Meta right now, and he's been developing this idea of joint embedding Predictive architectures, or GPOs for short, basically as like a path towards this more human, like AI.

So you can see right there that this generate something that is really comprehensive. And I think the biggest use case of this is that you're going to get something that you can listen to while you do your dishes, while you do your laundry, while on the way to work. It's just something that makes learning really interactive and really a lot more fun. So if this is a video that really helped you, let me know if you want me to make a longer video on separate use cases for this, because this technology is absolutely incredible. With that being said, I hope this tutorial helped you out. And if you have any questions, don't forget to leave them in the comment section below.