Tiago's Reaction Video on Notebook LM

Hello everyone and welcome back to another reaction video today. Why is it a reaction video about notebook LM? Something that I really wanted to test out. Well check this out notebook LM is currently only available in the US We are doomed in Europe man. I want to test this for ages already Well, I could use a VPN obviously, but you guys reached out and you said Tiago just published a video here Tiago Potter the brain behind building a second brain and I thought let's dive into this together. Okay, let's watch this the first time. I didn't watch it before. And let's see what Notebook LLM has to offer. And it seems to be a very comprehensive video. So let's not waste any more time. And let's dive in. Most of what we consider note taking and information management these days will disappear. I'm talking organizing your notes gone, formatting your notes no more searching your notes app and trying. We see here Evernote Tiago is still using Evernote on X he's Talking a lot about Evernote as well, and they keep improving the application, but there are some controvers discussions going on in the X there. This being said, it always depends on your use case and Evernote is still a very established tool. It's just a negative news around Evernote and the migration to a different company and things like that. that makes it feel not being trustworthy. I think, let's see what 2024 will bring. I'm following the new developers behind Evernote as well. They're finally bringing state-of-the-art features into Evernote eventually, like having a slash and you can mention other nodes and things like that, cross-linking, all the things that we would have expected already years ago. Let's see how this goes. But obviously, Tiago is now comparing Evernote to the Notebook LM. And let's see in this video if Notebook LM can actually replace his Evernote setup, which is very well established, obviously. Let's dive in. Anything Google created in this arena would already be interesting and worth investigating because it's Google. They have some of our most valuable data, some of the most powerful AI models, such as the recently released Gemini, which Notebook LM uses, and quite a track record for, you know, completely reshaping our relationship to information. But there's another wrinkle. One of the leaders and creators of the project is Steven Johnson, multiple time bestselling author and the person who introduced me to tools for Thought, the general category that includes second brains. Steven joined a team at Google Labs to create a completely different kind of AI that we have not seen before. And let me tell you, it's the closest one yet to the vision of a true second brain I've been pursuing for years. Let's begin with a little getting started guide. Start by visiting notebooklm.google.com and either creating an account or you can sign in with one of your existing Gmail accounts. Sign up is limited to users in the US who are 18 years or older. Yeah, there we go. I figured that out as well. Well, I'm over 18 years old. That's a box tick there, but it didn't help. Once you're inside, this is the interface you'll see. Everything in Notebook LM is organized in Notebooks. Okay, that's already a very minimalistic view. It's a typical Google style there. Not much to... To say it's structured into notebooks, well I guess it will search across different notebooks. Let's see how this will work. I have to better understand the back end. But what I already like to see here is, add your first source, a Google Doc, PDF or copy text. So if PDFs are searchable, which they likely are, as they already are, if you add them to the G drive, you can search for PDF content. I think we're gonna see some awesome stuff there. which is like a single contained space for all the information related to a specific project, such as a piece of writing. You can create a new notebook by clicking this plus button right here and give it a title such as personal ruggedization. This was an online course that I took recently that I'd like to organize and distill some of my takeaways. This is the main interface that you'll be using to interact with Notebook LM. And there's three main things to pay attention to. The first one is the column on the left, which is the source. This is really the main feature that Notebook LM has, which is you can click that plus button and choose from different existing documents from which the AI will draw its knowledge. And the three kinds of sources you can use are Google Drive documents stored in your own account. You can upload PDFs and you can copy and paste text directly from other places, such as a note taking app. OK, with the new Gemini AI there and being able to, was it one million words? words that it can hold. There's a lot of stuff we can add there. I compare it to chat GPT there where we can build the custom GPTs and we build for example a custom I-Core GPT that holds the I-Core methodology, the core scripts and things like that. So whenever we ask something in there we want to write an article it has the reference to this. And that's exactly how we would leverage this notebook LM. And this sounds very powerful. I also like the idea of having the containers. Therefore, I can be unsure. Other unrelated content isn't leeching into the results that I'm asking there. Well, let's see where this goes to. But obviously, it seems like it's not taking into consideration videos or images yet. Well, let's keep going. Example, I'll use my Google Drive. You can select up to 20 different... documents in drive and each that's strange 20 docs What does this mean? I can have endless long docs and then I have 20 of them. Or I can have short ones. What's the actual limitation of words then in this case? One of those documents can have up to 200,000 words, which... Okay, sorry about that. Obviously 200,000 words. So we have 20 times 200,000 words, which would end up in 4 million words, if I'm correct. So it can hold 4 million words, which is a lot more than Gemini is considered to be able to process. So interesting. Which means you can interact with up to 4 million words of text. That's a far larger context window than any other AI tool out there. I'll go ahead and select my notes on that personal organization class that I took recently and say insert. And I want to show you what happens. So first of all, that source appears here in the left. If I click the source, a couple things happen. First of all, the AI creates kind of a, almost like a summary or an abstract of this source. And these are some of the key topics. And then if I want to actually see the full contents of the source, I can just scroll down here. I wanna add a few more sources because the real power of Notebook LM is not working with one source. It's working and drawing from multiple sources all at the same time and simultaneously. So what... Just very quickly, I really like this, how this gets organized the moment you add a source there. This seems really powerful to me. What would be interesting as well, if there's an API that I can use, SAP or anything like that, and our automatically generated meeting nodes would be then added as a doc and then into a specific... notebook there because then I could draw a conclusion out of specific team meetings that we had or something like that. Really interesting stuff there to automate and increase serendipity of information that was collected on the go. What I'm going to do is click add source and add a few more drive documents. And what I'm going to do is navigate to a folder where I have saved those. All the notes on every book and article I've read over the past few years. Later on in the video, I'll show you how I was able to generate all of these documents. And what I want to ask myself is, what have I already read in the past and even taken notes on that might contribute to and be a useful resource in what I'm working on today? So I'm just going to go through this list briefly and select any other documents that kind of strike me as potentially relevant. Here's one about what the world might look like in 2312. This is a good book here about the power of the power of the power of the power of ...of constraints, climate change, which is the topic of the course that I took, even about what it might look like to terraform other planets, such as in the science fiction book Blue Mars, some big picture perspective about our place in the universe, designing for behavior change. So the nice thing here is that you could search on top in the drive and then it would look into the content of the different docs and shows you even these docs, even if all of them would be untitled. So obviously Tiago knows exactly what's in there. But the key to me really is for business usage would be these meeting notes or working documents and things like that. Also, it would be really interesting if you add a doc to this notebook lm that is a working document, it gets updated. If actually the doc, the content that is then inside my notebook lm also gets updated and therefore the results. Sounds confusing, but this would avoid. having duplicated information that might get outdated over time. And it would be just logical that it is actually just a link to the G-Doc instead of a copy of the content. Change because we may need to have a lot of behavior change when it comes to preparing for climate change. I'm really just kind of noticing which books would jump out at me as in some way related to the topic at hand. So it looks like I have 13. I'm going to go ahead and click insert. All of the sources that I chose are now being added to the sidebar. So let's talk about the two other spaces that you're going to be working with. And now that we have all of our sources loaded up. The main one is this little box down here, which is really like other chat based AI tools such as ChatGPT that we're used to. You're basically going to have a conversation with the AI, but drawing on the sources over there in the left. So I could say, yeah, as I mentioned in the beginning, we are using ChatGPT. I can quickly show you how this looks like. So here we are working on iCore productivity expert, GPT. And when we go here, you know, this iCore expert, now I can ask anything like what is the... Capturing beast and therefore capturing beast within the context of I-Core and productivity frameworks and so on. And now it pulls the data from what we have in the backend. And I show you quickly how this looks when I go to edit this. You see here, I added all the PDFs, which contains the scripts from our courses. And therefore this. AI understands what I'm asking for, tries to find the answer in the scripts. And this is also long form content. It is way beyond what the limitation is of ChatGPT, but I have to say it works very well. And then here is the prompt that you give this and therefore you can train this model and it should work and therefore it should work this way. And that's how we have kind of notebook LM. Now I can share this between Paco and me. and we can use this now to say for example create a linkedin post about capturing beast let's see and then there we go the taming the capturing beast within the echo framework and it's not a hundred percent reliable obviously this gives you a good starting point to write something out but we have still to bring in our expert knowledge to makes sense, makes no sense in the I-Core framework. This is just related to I-Core and not the I-Core framework and things like that. So therefore, it's not so reliable. In the beginning, it worked very well because it started to search in these PDF files. But now we figure out over time, it doesn't even care about looking into these PDF files, but writes out some stuff that it seems to draw from the general prompt. So that's really confusing. That's where I really think... I really hope that this notebook LM will perform a lot better when it comes to this. Something like create an outline of the main ideas related to personal preparation for climate change found in these sources. Hit enter. You can see it takes a minute. And there we go. It's a hierarchical, highly structured outline with main points and then supporting points or evidence or expanding on those points underneath each one. And they're kind of in order from early things that you'll need to think about to much later things. This is just the first step. There's a few different directions that we could go from here. The first thing is I could ask it One of these suggested questions that you see down here, you can see I can actually scroll right and left. And what's interesting is they're based on this particular text. So they're not just generic ones that appear the same way every time. So things like what are the three main things to consider? So let's say, well, this reminds me a lot of mem. That's what they are approaching. Let's check out mem. As you might have seen on another video on our channel that we are not so happy with mem right now. But that's exactly the approach, right? When I ask anything. then it will give me the answer it shows me the sources where it found it from and then it gives me the recommendations what i want to ask next so developer incorporating magic slides and that's just stupid because it shows it's not looking into pdf file at all so probably it comes up with some random thing see i'm sorry tom i'm unable to do this and that's that's really the challenge for mem i think since day one as they were using The GPT and the GPT limitations really limits the whole note-taking app to see their whole knowledge base because the questions you ask, it will look through the whole knowledge base in order to come find the information you're looking for. So therefore, it really relies on the titles and things like that. I'm 100% sure what Tiago is showing here now, you see it's a similar approach, but the model, the AI model... has a complete different knowledge. At the same time, restricting it to several notebooks, it still limits the knowledge that this model can access. So therefore, you are a lot more focused on specific topics that you are asking for and makes it easier for the AMI model to understand what you are asking for and what you are looking for. So, interesting approach. You see, it's not like that notebook alarm is something revolutionary when it comes to this type of tool, as we had something similar in ChatGPT, we had something similar for years in MEM, but obviously it looks like they really refined the thing, and now let's see the results that Tiago got there. I just want something really snappy, really succinct. I could click that, it will ask that question, and there you go. It's kind of distilled it down to the three main things that I need to know. What are some ways that individuals can prepare for and respond to both the personal and societal changes that are likely to occur as a result of climate change? So what I would be really interested in, it says here six citations. If you click there, if it then shows you the sources and actually the different places inside the doc where this was shown by. where I found it in, probably Tiago will just show this later on Let's try that one. As it's working behind the scenes, I'm really having a conversation. I can ask it to expand or to distill. I can ask it to give me more information on one point, but not others. I can ask it to give me different formats, like write this as a poem, write it as a narrative, write it as a question and answer or FAQ format. You can see in this case, it's sort of adapted the content to really be focused on personal preparation, which is the angle that I'm most interested in. There's one more, the third of the main three interfaces. ...that you need to know about. If you get a response, let's say this one, that you especially like, you don't want to lose it. This is a free-flowing conversation, so you might think, well, if I keep going and keep conversing, I'm going to lose track of some of these great responses. All you have to do is click this little pin icon, and it's saved to this big space up here taking up the top of the screen. You can think... That's really nice, I think. That's something I would like in ChatGPT as well. In this long list of conversation, that I could have some blocks that I save elsewhere. And that's exactly there. So therefore you're building up, you're building blocks in order to leverage it later. And therefore you're extracting information from this huge amount of 4 million words down. You're distilling it to take Tiago's words there. And therefore you can leverage this in a much more focused way to build up the content that you want to get out of this. It's a really interesting approach. Think of this as a kind of pin board where you are pinning these kind of little cards with interesting information that the AI has given you just so that you keep them top of mind and easy to see and easy to reference. One more cool detail here. Let's say in the midst of your work you realize, oh there's another source I'd like to add. You can always click right here where it says add note and either copy and paste from somewhere, for example your note-taking app or something that you found online or something from an email, or you can just write directly into the note. For example, Let's say I remember that the creator of that course that I took, his name is Alex Steven, recently wrote a piece on Substack that is much more recent and much more timely. And I want to also incorporate that writing and that thinking into what I'm doing. What I can do is head over to Evernote. Here is the excerpts from his article. I'm going to do Command A for Select All. I'm going to copy all that text and put it right into a new note. And now what it will do is allow me to check this box right here and incorporate. ...corporate what I've just added right into the conversation. That was a very basic tour of three of the main areas that you are going to need to pay attention to to use Notebook LM effectively. So I didn't get it really. So when he ticks the box, it takes it into the conversation. So if I untick it, it's not considered in the conversation. How does this work? Also, I really like that it takes the links there, but it would be interesting, which is probably not the case, if Gemini is actually looking up. the content on these links. Therefore, it would have up-to-date content available as well. You know, that's something that I said in the beginning about the having up-to-date documentation there. So you could keep improving the knowledge in this notebook. Well, let's keep watching. Let's see what comes up. What Google has done is basically redesigned a note-taking app from scratch with modern AI in mind. You still need to take notes, but you can just kind of dump them into one giant pile and use the AI to sort through and make sense of them rather than... Well, I don't know if Google really invented this, as I just showed you with MEM. There are also a lot of other note-taking apps that leverage AI, just thinking about ClickUp Brain, where you leverage the whole business knowledge base and things like that. ...that so I would rather say Google is pretty late to the game when it comes to this. Having to meticulously... But that's my own opinion, obviously. Organize things in files and folders. Notebook LM is different from other AI tools such as ChatGPT because you can train it on a specific set of documents you know and trust. That could be excerpts. I'm not sure about that. I just showed you the GPT that I developed there. And all the other GPTs even can use external data. You can use ChatGPT API. And therefore hook into your external website and it will leverage the data from there. So you can really train your own. GPT models are based on your knowledge base. Um, it seems more complex if you ever talk about it, then having a Google and then you throw in some documents and it gets it. So, uh, I would say the Google approach is something between mem and chat GPT that I just both showed you. But again, it's not the only thing that allows you to build this type of notebooks or knowledge management and leveraging the knowledge. Point the AI at a specific set of documents, effectively giving it instant expertise in whatever knowledge they contain. So you always know it's only drawing on authoritative, private, trusted information. Next I want to take a look at some of my favorite use cases that I've been using Notebook LM for. So I think really thinking about read later apps that this would be a lot more useful if I for example save this video to a read later app. and then I'm building up a knowledge base this way and just by giving it different tags instead of moving it into a closed folder I could then use these tags in order to tell the AI in the chat what it should focus on and therefore starts to only take these tags into consideration again to me this would be a lot more flexible that I can build up a very dynamic knowledge base without The need to think about, oh, in what notebook do I save this now in order so I can leverage it later on? Let's see what the future is holding there because ReadWise ReadLater app, they have already AI. If they just add a chat there, that might be a solution there as well. Sometimes I'm reading a difficult, complex piece of writing, such as an academic paper, and I just need some help understanding it. What you can do is upload a PDF of what you're reading. This is a paper about the role of forgetting and learning. I'm going to go ahead and select it and say, what is the role of forgetting in learning? So you can see here it's basically summarized what is quite a long technical paper into something that I can understand. So I noticed one of the points here which is that knowledge can sometimes have a negative value, kind of peaks my interest. The amazing thing about this conversational interface is I can now ask, what are some examples of ways knowledge can have negative value. That's kind of confusing. So what I rather would have liked here, if he would be able now to highlight this part that is really interesting for him, and therefore it focuses more on this type of information. Obviously, I can keep asking questions now about the document. Again, something that ChatGPT can do for over a year now. probably even beyond because he didn't upload any diagrams or anything like that or business data. And so there's no, it doesn't seem that there is a code interpreter in there or something like that. It's really for, it seems it's really more for knowledge workers, which is Tiago. Obviously, he is a knowledge worker and approach and target group are knowledge workers and students, whereas we are busy professionals and we always look for ways. How are we able to integrate this into the... daily business. So it's really useful in teamwork and daily business work. As I mentioned in the beginning, meeting notes would be a very nice example. So I can quickly leverage what this is all about. This being said, another example would be Rewind. Rewind is a tool that you can install right now on Mac and it's recording 24-7. Whenever you have your Mac on, it's recording your audio, all your meetings, and then you can ask. This AI, what meetings did I have last week and give me a summary of the key takeaways of these meetings. And it will literally go backwards. You can even scroll backwards and shows up the Zoom meetings and all this gives you a summary and you can leverage this. So again, I'm not so much impressed so far about the capabilities of the Gemini, the Notebook LM approach here. Really the difference is really the vast amount this AI can. process. So let's see what happens when the other AIs will catch up to me. So you can see here, this is fantastic. Knowledge that's incorrect, redundant knowledge, harmful knowledge. So imagine if I'm working on a project where I'm trying to convince, let's say, an I administrator. To not save every file that's ever been created across the entire organization. That's something that I could ask it to do for me. I could say, please turn this information into a letter written to an I admin arguing that we don't need to preserve every document ever created by the organization. And there we go. Probably need some editing. I'll definitely want to go through it and add my own touch. But what's so neat about this is what you see right here, which is the citations. Now we are talking. I'm curious what this is now. Let's say I send this letter to the I admin and they want to know, well, where did this come from? What is the source? What is the evidence? What's the proof? By clicking here, I can see specifically where every piece of information or fact or argument in this. Sorry, but this implementation doesn't impress me at all. If I would expect if I click on ten citations that it will give me a drop-down or anything, giving me actually the citations and then the source, and then I click on this and open up the source. So now let's see what happens when he clicks on one of the numbers. ...conversation came from. So for example, let's click on number six here. And not only does it show you what is the original text that it's drawing from, but over here on the left, it will open up a window. and highlight in purple that piece of text in context. This is so important to have. Yeah, I really like this, that it shows up the exact place where it found this information. But again, I have to go through all the numbers in order to understand what the different types are. So having a quick summary or a title showing me what is behind this number would make it much easier to snipe down on the source that is really relevant for me to share. The value citations need to be shown in context. What was being argued before that appear? What was being argued after that? What is the holistic kind of situation in which this knowledge arose? Let's do another one, which is meeting notes. How often have you been a... There we go. Now I'm really looking forward how he will leverage this one. Part of a meeting, let's say a long meeting, and you get the transcripts of everything that was said, but they're so long and convoluted. You don't really want to spend the time to read through all those. Even just the notes that you take yourself. As part of the meeting that you're in, this is a great use case for Notebook LM. Here is a Google Doc, which is the full transcription of a phone call that I had with a client of mine who's named Duck. And you can see it's very detailed, very comprehensive, has who was speaking, when they spoke, what they said. So there's definitely value here. But when I start to scroll, you can see this is 26 full pages of very detailed text. So that's something we are not doing at all, as there are again a lot of tools out there already for a long time that do not only the transcript for you, but do summaries, highlighting. You can even click on the highlights out of this meeting and it will jump into the place inside the recording of the meeting, showing you the source inside the video. So we personally, we are using MeetGeek. As usual, we are not sponsored by these companies at all. There are other alternatives like Fireflies. There's another one, Spynack, AIO. I think even Potter does now the summarizing thing as well. There are many, many tools out there that do the summarizing for you in the first place. And obviously, you could always use ChatGPT to ask the same things to give a summary out or highlights or key takeaways or action points out of your conversation that you had with somebody if you have the full transcript available. I really don't want to spend the time to find the one or two facts or opinions or whatever it is I'm looking for from so much text. So what I'm going to do is just copy the title of this Google Doc, go back to Notebook LM. I've created a new notebook here called Meeting Notes. Click Add Source. Let's go back to Google Drive. I'm going to paste it in the search just to make it easy to find. There it is. You can see one selected. I'm going to hit Insert. And what I can do now is essentially have a conversation with that. A document which if we look here has over 12,000 words. It's much, much bigger than can fit in ChatGPT for example. So what I'm gonna ask is what? The key is really how good the AI can find information in long form text. So Gemini really excels at this point. There was this example that you can have all the different books of Harry Potter uploaded to Gemini. And then if you ask for one specific sentence, Throughout all these millions of words, it will identify the sentence. And that's what chat GPT or GPT is really struggling with, that it comes up, as I showed before, with some information that is not true. So it's dreaming. So that's really the superpower here. And obviously, if you then use transcripts like this, it will make a lot more sense out of this, I can imagine. But therefore, there would be the need. to test this side by side. So I kind of wait to get my hands on Notebook LM as well to make one of these comparison videos then. Are the main points of my conversation with the duck. So you can see here, Doug expressed interest in creating a training program. What kind of training program? Well, he believes a group workshop would be most effective. What does he want to incorporate into that workshop? Elements of my course. I mean, this really just saved me 30 minutes to an hour at least. So I could even go a step further. Please create a proposal. for a group workshop training program for Doug's organization. And there we go. And here it is really important. That's what I showed you with the custom GPT. So if Tiago would have now also the additional information about his program, how much it costs, you know, the boundaries of... all this, then it would come up with a much more specific proposal based on the general knowledge. So what I try to say here is having these buckets, I would need to add this information in each of the buckets whenever I want to recall it. So it would be great to have one global bucket for this notebook LM that AI can access anytime and whenever I ask anything about my business. So the business relevant information would be just in this one thing. And then if I ask anything about my business, like here for Tiago's example, proposal for his workshop, then it will pull the information from this relevant information. I think that that's something that's missing. See, it's written in the language of a proposal. There's an introduction, objectives, target audience, program structure, features of the program, benefits. I mean, this is really fantastic. But I want to take this even one step further. I want to actually add that right in. Why not just dump in the entire text of my book and allow the AI that I'm working with to draw on those details? Well, that's exactly what I tried to do with Chad Chippity. And I'm really curious with his book. I think it's more than 270 pages. So a lot of words will be in there. Let's see how good the answers will be then. I'm going to go ahead and click add source right there. I imagine one thing Doug will want to see is the full curriculum. What is the full content that you plan on training them on? So why don't I anticipate that and just say, now please, I don't know why I always say please when I talk to AIs, create a full detail. Well, when they take over the world, we want all be the guys who are nice to AI. I also say please to AI all the time. Well it seems Tiago and I, we are just nice people, no matter if it is humans or robots. Curriculum drawing on my book, Building a Second Brain, that covers the topics above for a full day group workshop. And there we go. You can see there's a target audience, goals, it has the morning session and what we're gonna cover specifically, introduction, the four pillars, fantastic. Sounds very familiar because I wrote that myself. Here's the afternoon session, here's the closing session. What's so cool about this is the reason I I can feel completely comfortable uploading the full text of my book to this platform. Is the sources that you use in Notebook LM don't go anywhere else. They're not used to train any model. They don't get shared with anyone. The second that I leave this notebook, all of that will be wiped clean from the AI's short-term memory. So those were probably half a dozen use cases that have to do with your day-to-day work, reading, understanding. Yeah, but I'm not so sure about this with the short-term memory of AI. I would like that AI would Goods keep learning the more I interact with it about a specific topic. So whenever it makes a mistake, I can correct the mistake and say, no, no, no, you misunderstood this. It will reinterpret the things. And then next time I go into the notebook, I know it's already trained, trained just with me and not global training. So it's specifically trained not only having access to my knowledge, but having also access to what I'm expecting. from the AI to answer and how it should answer and so on. Another thing that he didn't mention here, I'm not sure about the tone, if you can change this, but this will be pretty obvious, that be more professional, be more casual, and things like that. Translating from one format to another. But next, I want to dive into a very specific use case that is very close to my heart, and I know close to the heart of the people who created this, which is writing, especially writing that is creative, is creative. or that is complex or multifaceted, that is trying to simultaneously pull from many different kinds of really diverse sources while also synthesizing it into something that is your own creation, that has your own style, your own voice, your own perspective. There we really see what Tiago's passion is. He really gets excited talking about writing there. Something I wouldn't get excited at all, but our co-founder, Duck on Harry, he loves writing as well. So therefore this is his part now there. This I want to show you is using notebook LM, almost like an editor. So this is a piece of writing that I've done. It's in draft mode. And what I'm going to do is copy the title. I've created a new notebook here called mastermind insights. I'm going to go ahead and add a source, go to drive, paste that in the search field. There it is. You can see one selected insert. You can see if I click on the source. It will show me what I really would find interesting if I could just select a folder on my G drive and then I can keep adding things to this folder. That would be really useful so I don't have to search things all the time, but I can move the information something like I mentioned before with the Relator app where I just give the tags and I just throw the information there and I don't care about the moment I information, but it's ready whenever I need it and to retrieve it. That I think would remove a bit of a friction if I want to build up a knowledge base for a specific project or writing project, for example, here as well. What they call a source guide, which includes a summary and key topics. Let's exit out of that. And what I'll ask is, please suggest ways this piece could be improved. So what's interesting is it's giving me lots of ideas, but what I wanted to actually do is say, please rewrite this piece in full following the instructions above. It's funny, it wrote it in a bullet point format. So I'll ask, please convert this from bullet point format into prose. Now, This is already probably better. It's certainly much more succinct, which I'm always trying to go for. There's one more thing I can do here, which is add a source, go back to Drive, and let's see if I have any interesting sources on writing. So this is one that jumps out at me immediately. I can tell from the title, it's a my notes on a book called Write Useful Books, which is really about non-fiction prescriptive writing. So I'm going to insert that and say, please rewrite. My article on takeaways from my first mastermind, drawing on advice found in right, useful books. It's interesting. I'm curious looking at the citations, which tell me some of the specific ways that it did that. For example, not using titles or headings that are clever or punny. because that is interesting or entertaining but not useful is a great piece of advice for this kind of writing. The next use case for writing I want to show you is really earlier in the ideation process. Let's say you're not even sure what you want to write in the first place. You just kind of have some messy notes and you're looking for a collaborator that can help you give them some meaning and some structure. I'll go ahead and create a new notebook, which is a new article. Let's say I don't even know what it's going to be about. I'm gonna go ahead and add a source, go to Drive, Let's go back to Readwise, which contains all the notes and all the books that I've read. And what I'm going to do is pick one, let's say, that I've read recently, such as A History of the Index. And what I'm going to ask it is, please suggest related ideas or avenues of research related to this book. So it's giving me ideas from indexing prehistory to alternative indexing systems. This book is about indexing the impact of digital indexing. Each one of these is potentially a rabbit hole that I could run after. The one that jumps out at me is historical evolution. And just seeing that reminds me that I have another book that I've taken notes on in the past, which I believe was called Glut. There it is. So let's add that in, which is more about the past and about how our relationship to information has changed over time. So now I have two sources I can ask, please find connections or relationships between the ideas in these two books. Let's see if it found anything interesting. So this is the one that strikes me. I would like to write a piece about the democratization of knowledge, how access to knowledge and ways of structuring knowledge has become more accessible over time. So I'm going to say, please suggest a few ideas for articles I could write about the democratization of knowledge. through history. And these all kind of strike me as somewhat generic. What I think I would like to do is use a specific story, use a specific example. And what comes to mind, I don't know why, is a book I read about paleontology. Let's see if I can find it. Yeah, it was called Fossil Men. Let's add that in here. And now I'm going to say, suggest the title of an article about the democratization of knowledge using an example. From paleontology or fossils. Okay, that's an interesting title, but now I'm gonna say expand on that title with an outline of key points this piece could contain. Interesting. Fossils in the history of life, fossils in human origins. So it's really going hard on this idea of fossils, but kind of weaving together or weaving through it different ideas about democratizing knowledge. This would obviously require more development. element I'd want to go in and look at the sources that it found, maybe ask some of these follow-up questions. Let's say I'm satisfied with this initial outline for a piece. I could even go so far as to say, please write the opening lines of this article, making them as attention-grabbing as possible. Like a hidden world awaits, a chronicle of life frozen in time. This is fantastic. Just notice what we've done. We went from not Not even knowing what I wanted to write about, really just a single source that I had been reading recently, all the way to the opening lines of what sounds like a pretty interesting piece if you ask me, through a completely conversational back and forth collaborative exercise with an artificial intelligence is totally amazing. And that's a very important point that Tiago points out here, that people really need to understand how to approach AI. So for us as well, it's really just... to get started so you're no longer there's no longer an excuse that you're sitting on a blank canvas on a blank piece of paper and you get writer's block or something like that there are hundreds of ways to get your thought process going and then you go down a rabbit hole in a conversational style with ai and i think this is really the superpower of ai this is how we use it with chat GPT on a daily basis. And the results don't need to be perfect, obviously. Otherwise, we would be swamped with only AI responses. But it gives you an idea. So when I approach something like this, I usually always ask for, give me five titles or give me even 20 titles, okay? And then out of these 20 titles, there's one that really caught my attention. And then I ask, give me five more variations of this specific title. And therefore... I can keep refining and then I exchange one or two words myself and therefore I still have the feeling that I developed this result together with AI but it's still original it still feels human the moment I use this content then and I think this is really key how we can use the AI in this case. Now you may have noticed that what allowed this, what made this possible, was that I had all my notes and highlights from every book and article I've read over the past few years already in Google Drive. But you can do that too, and I wanna show you how. It's possible with a really powerful tool called Readwise. Check out our in-depth video on how to use Readwise for more information. But what you can do is go to Readwise, which is a... paid service and once you create an account and log in you can click dashboard here if you go to the import section you want to make sure that under the connected section right here you have your Kindle ebooks or your relator app or social media or whatever is the source that you want to be able to save in Google Drive Which you can then use in notebook LM. So for me, the two main ones are Kindle right here and Readwise reader, which I use to read online content. Let's head back to the dashboard. And then the second thing to look at is under export. You want to be sure that Google docs is listed here under the connected section. Normally it will appear down here. It's really amazing that allows you the export into the. G drive, this not only allows the notebook alarm to look at this, but also again, I mentioned ClickUp Brain, but even inside G drive, you can search for the information and it will find the information you're looking for. As it looks into the content of these notebooks, I think this is really the way it should go. So that's exactly what I said that I can now throw the stuff into my read later app, which was shown there as well. The Readwise Reader read later app. and it connects to your readwise and therefore it saves then your highlights from even youtube videos and other things into g drive this is really amazing you have to do is click connect sign into your google docs account the same account that you use to create the notebook lm account and then google docs will appear right here if you click into the google docs settings the key feature that you want to pay attention to is this third one right here if this is not checked Then Readwise will send all the excerpts from each source to its own individual Google Doc. So the excerpts that I saved from one book will go to one Google Doc. That makes it really easy to go through like you saw and select only the sources that you want to use in Notebook LM. But if this is turned on, what it will do is it will send all the highlights to one, or maybe if you have many highlights, just a few Google Docs. so that you can fit it within that 20 notebook limit. What that allows you to do is essentially have a conversation with your entire history of highlights from whatever sources you imported into Readwise. This is really powerful, guys. That's really powerful. Once you click here where it says start export to Google Docs, you'll want to leave it running for probably a few hours depending on how many highlights you have. But once it's finished, you'll see a Readwise folder appear in Drive. Alright. So it's not syncing. I have to double check, but it would make sense that it keeps syncing because I don't want to get back into readwise every time and then start syncing. It's another friction point there. Probably it will automatically sync there. I know that was a lot. I want to turn now to the other side of the story. There's always another side. Which is some of the limitations and pitfalls. This is an experiment's brand new platform, which means it's a little rough around the edges. There are bugs. They're having to invent a completely new interface, which means there is a bit of a learning curve. You have to figure out what the different buttons do and what the different parts of the window mean. Currently, you have to provide the sources to Notebook LM in the form of basically Google Docs or PDFs. It doesn't have the ability to directly connect to a note-taking app. Such as Evernote or Notion or even Google Keep. You'd have to use something like Readwise as an intermediary to get your notes into Notebook LM. You're limited to only 20 sources at most. So if you have more sources than that that you want to use, you're going to have to do some manual copying and pasting. That could become interesting for scientists, but they have loads and loads of papers they are working through. 20 is not... enough there obviously. ...to get them under 20. Just like a lot of other AI tools, Notebook LM has trouble doing math, reading PDFs that have messy formatting, and finding a lot of precise details if the sources that you're using are really large. I've also talked to the... That's interesting, something that I mentioned earlier in this video, that this should be the superpower of Gemini, but maybe it's not part of the Notebook LM thing then yet. Notebook LM team. And they said fixes and improvements to a lot of those issues are coming soon. This is way more than a note-taking tool, way more than a writing tool, and it's way more than an organizational tool. Here's how I describe what Notebook LM is trying to do. Until now, the stages of the creative process, like reading, researching, note-taking, and writing, were completely separate. We had to use completely different tools for each step, which means we were constantly having to switch between them and worry about things like formatting and compatibility and security and privacy no one software tool could do it all notebook lm i'm not sure what about notion q a where you can ask your database i just mentioned clicker brain asking your database what about mem where you can ask your whole knowledge base at the same time you can read it you read the things for researching i wonder well you do research on the sources that you add to notebook lm however Researching, in my opinion, is it should also be able to access the web and using and search the web and find relevant and latest information, maybe even access actual paper and articles and studies and things like that. So but the things that he is talking about is very revolutionary. It's already happening for more than a year now. And we will just get used to this style of reading, researching, writing and note taking in one app. If it is notebook.lm, I don't know. The superpower of notebook.lm I see here, it's for free as it is part of the Google workspace. I think it would be really interesting if there are specific sources that you can add and log and then give your team access to it so everybody could ask questions about the specific knowledge base. These are the things that we do in ClickUp right now, where we just We have all our SOPs, the standard operating procedures, the work instructions, but also the relevant projects and all the things that we worked on. And there you have also a chatbot that you can ask anything about your business knowledge base inside ClickUp and you will get relevant information and the sources as you have seen here with the citations and things like that. So as I said already, to me, Google is late to the game here when it comes to this innovative product here. But it will become the new standard. That's the point. And other note taking apps like Evernote certainly have to catch up. When they introduced their AI search, it was still behind what other tools were already capable of. And now let's see how quickly Google can adapt to this new pace there. Gets all those steps and folds them into a single integrated space so you can remain in that wonderful flow state and keep moving forward. They call it working at the speed of thought. It's an AI collaborator trained on the data only you possess with your unique view of the world. I suggest signing up for a free account at notebooklm.google.com. Let me know what you think about Notebook LM in the comments. Anything Google... I have trouble saying the word Google. How do you say that word? Google. Google. Google. Google. So it seems obviously that he talked about the Notion AI. If you want me to react to the other video here as well, especially in comparison now to the Notebook LM, as he pointed out all the advantages, I'm really curious about this one, but I won't watch it before I do an actual reaction video. So go to the comments below it if you'd like to see me reacting to the other video as well. I think it's more like goo like you say the GOO first Google good it's really interesting I mean I'm not a native speaker at all I'm overthinking all the pronunciations as well but it's good to see that native speakers even think about how to pronounce Google Google Google As it is just a made up word, right? Google. I would maybe refer to the Google, but which would be then goggles. I'm not getting into this conversation either. Otherwise, I cannot speak a word anymore as I'm overthinking things. How do you say that word, Alex? Google. Just like a normal person, basically. Like a normal person. Everybody, I really enjoyed this video. Thank you very much for all the insights. Tiago was really interesting. I think it was very comprehensive. And well, I cannot wait to catch you guys up in the next one.

I want to test this for ages already Well, I could use a VPN obviously, but you guys reached out and you said Tiago just published a video here Tiago Potter the brain behind building a second brain and I thought let's dive into this together. Okay, let's watch this the first time. I didn't watch it before. And let's see what Notebook LLM has to offer. And it seems to be a very comprehensive video.

So let's not waste any more time. And let's dive in. Most of what we consider note taking and information management these days will disappear. I'm talking organizing your notes gone, formatting your notes no more searching your notes app and trying.

We see here Evernote Tiago is still using Evernote on X he's Talking a lot about Evernote as well, and they keep improving the application, but there are some controvers discussions going on in the X there. This being said, it always depends on your use case and Evernote is still a very established tool. It's just a negative news around Evernote and the migration to a different company and things like that.

that makes it feel not being trustworthy. I think, let's see what 2024 will bring. I'm following the new developers behind Evernote as well. They're finally bringing state-of-the-art features into Evernote eventually, like having a slash and you can mention other nodes and things like that, cross-linking, all the things that we would have expected already years ago. Let's see how this goes.

But obviously, Tiago is now comparing Evernote to the Notebook LM. And let's see in this video if Notebook LM can actually replace his Evernote setup, which is very well established, obviously. Let's dive in. Anything Google created in this arena would already be interesting and worth investigating because it's Google.

They have some of our most valuable data, some of the most powerful AI models, such as the recently released Gemini, which Notebook LM uses, and quite a track record for, you know, completely reshaping our relationship to information. But there's another wrinkle. One of the leaders and creators of the project is Steven Johnson, multiple time bestselling author and the person who introduced me to tools for Thought, the general category that includes second brains.

Steven joined a team at Google Labs to create a completely different kind of AI that we have not seen before. And let me tell you, it's the closest one yet to the vision of a true second brain I've been pursuing for years. Let's begin with a little getting started guide.

Start by visiting notebooklm.google.com and either creating an account or you can sign in with one of your existing Gmail accounts. Sign up is limited to users in the US who are 18 years or older. Yeah, there we go.

I figured that out as well. Well, I'm over 18 years old. That's a box tick there, but it didn't help. Once you're inside, this is the interface you'll see. Everything in Notebook LM is organized in Notebooks.

Okay, that's already a very minimalistic view. It's a typical Google style there. Not much to...

To say it's structured into notebooks, well I guess it will search across different notebooks. Let's see how this will work. I have to better understand the back end. But what I already like to see here is, add your first source, a Google Doc, PDF or copy text. So if PDFs are searchable, which they likely are, as they already are, if you add them to the G drive, you can search for PDF content.

I think we're gonna see some awesome stuff there. which is like a single contained space for all the information related to a specific project, such as a piece of writing. You can create a new notebook by clicking this plus button right here and give it a title such as personal ruggedization. This was an online course that I took recently that I'd like to organize and distill some of my takeaways. This is the main interface that you'll be using to interact with Notebook LM.

And there's three main things to pay attention to. The first one is the column on the left, which is the source. This is really the main feature that Notebook LM has, which is you can click that plus button and choose from different existing documents from which the AI will draw its knowledge. And the three kinds of sources you can use are Google Drive documents stored in your own account. You can upload PDFs and you can copy and paste text directly from other places, such as a note taking app.

OK, with the new Gemini AI there and being able to, was it one million words? words that it can hold. There's a lot of stuff we can add there. I compare it to chat GPT there where we can build the custom GPTs and we build for example a custom I-Core GPT that holds the I-Core methodology, the core scripts and things like that.

So whenever we ask something in there we want to write an article it has the reference to this. And that's exactly how we would leverage this notebook LM. And this sounds very powerful. I also like the idea of having the containers.

Therefore, I can be unsure. Other unrelated content isn't leeching into the results that I'm asking there. Well, let's see where this goes to.

But obviously, it seems like it's not taking into consideration videos or images yet. Well, let's keep going. Example, I'll use my Google Drive.

You can select up to 20 different... documents in drive and each that's strange 20 docs What does this mean? I can have endless long docs and then I have 20 of them. Or I can have short ones.

What's the actual limitation of words then in this case? One of those documents can have up to 200,000 words, which... Okay, sorry about that. Obviously 200,000 words.

So we have 20 times 200,000 words, which would end up in 4 million words, if I'm correct. So it can hold 4 million words, which is a lot more than Gemini is considered to be able to process. So interesting.

Which means you can interact with up to 4 million words of text. That's a far larger context window than any other AI tool out there. I'll go ahead and select my notes on that personal organization class that I took recently and say insert.

And I want to show you what happens. So first of all, that source appears here in the left. If I click the source, a couple things happen. First of all, the AI creates kind of a, almost like a summary or an abstract of this source. And these are some of the key topics.

And then if I want to actually see the full contents of the source, I can just scroll down here. I wanna add a few more sources because the real power of Notebook LM is not working with one source. It's working and drawing from multiple sources all at the same time and simultaneously. So what...

Just very quickly, I really like this, how this gets organized the moment you add a source there. This seems really powerful to me. What would be interesting as well, if there's an API that I can use, SAP or anything like that, and our automatically generated meeting nodes would be then added as a doc and then into a specific...

notebook there because then I could draw a conclusion out of specific team meetings that we had or something like that. Really interesting stuff there to automate and increase serendipity of information that was collected on the go. What I'm going to do is click add source and add a few more drive documents.

And what I'm going to do is navigate to a folder where I have saved those. All the notes on every book and article I've read over the past few years. Later on in the video, I'll show you how I was able to generate all of these documents. And what I want to ask myself is, what have I already read in the past and even taken notes on that might contribute to and be a useful resource in what I'm working on today?

So I'm just going to go through this list briefly and select any other documents that kind of strike me as potentially relevant. Here's one about what the world might look like in 2312. This is a good book here about the power of the power of the power of the power of ...of constraints, climate change, which is the topic of the course that I took, even about what it might look like to terraform other planets, such as in the science fiction book Blue Mars, some big picture perspective about our place in the universe, designing for behavior change. So the nice thing here is that you could search on top in the drive and then it would look into the content of the different docs and shows you even these docs, even if all of them would be untitled.

So obviously Tiago knows exactly what's in there. But the key to me really is for business usage would be these meeting notes or working documents and things like that. Also, it would be really interesting if you add a doc to this notebook lm that is a working document, it gets updated. If actually the doc, the content that is then inside my notebook lm also gets updated and therefore the results.

Sounds confusing, but this would avoid. having duplicated information that might get outdated over time. And it would be just logical that it is actually just a link to the G-Doc instead of a copy of the content.

Change because we may need to have a lot of behavior change when it comes to preparing for climate change. I'm really just kind of noticing which books would jump out at me as in some way related to the topic at hand. So it looks like I have 13. I'm going to go ahead and click insert. All of the sources that I chose are now being added to the sidebar.

So let's talk about the two other spaces that you're going to be working with. And now that we have all of our sources loaded up. The main one is this little box down here, which is really like other chat based AI tools such as ChatGPT that we're used to. You're basically going to have a conversation with the AI, but drawing on the sources over there in the left.

So I could say, yeah, as I mentioned in the beginning, we are using ChatGPT. I can quickly show you how this looks like. So here we are working on iCore productivity expert, GPT.

And when we go here, you know, this iCore expert, now I can ask anything like what is the... Capturing beast and therefore capturing beast within the context of I-Core and productivity frameworks and so on. And now it pulls the data from what we have in the backend.

And I show you quickly how this looks when I go to edit this. You see here, I added all the PDFs, which contains the scripts from our courses. And therefore this.

AI understands what I'm asking for, tries to find the answer in the scripts. And this is also long form content. It is way beyond what the limitation is of ChatGPT, but I have to say it works very well.

And then here is the prompt that you give this and therefore you can train this model and it should work and therefore it should work this way. And that's how we have kind of notebook LM. Now I can share this between Paco and me. and we can use this now to say for example create a linkedin post about capturing beast let's see and then there we go the taming the capturing beast within the echo framework and it's not a hundred percent reliable obviously this gives you a good starting point to write something out but we have still to bring in our expert knowledge to makes sense, makes no sense in the I-Core framework. This is just related to I-Core and not the I-Core framework and things like that.

So therefore, it's not so reliable. In the beginning, it worked very well because it started to search in these PDF files. But now we figure out over time, it doesn't even care about looking into these PDF files, but writes out some stuff that it seems to draw from the general prompt. So that's really confusing. That's where I really think...

I really hope that this notebook LM will perform a lot better when it comes to this. Something like create an outline of the main ideas related to personal preparation for climate change found in these sources. Hit enter.

You can see it takes a minute. And there we go. It's a hierarchical, highly structured outline with main points and then supporting points or evidence or expanding on those points underneath each one. And they're kind of in order from early things that you'll need to think about to much later things.

This is just the first step. There's a few different directions that we could go from here. The first thing is I could ask it One of these suggested questions that you see down here, you can see I can actually scroll right and left.

And what's interesting is they're based on this particular text. So they're not just generic ones that appear the same way every time. So things like what are the three main things to consider? So let's say, well, this reminds me a lot of mem. That's what they are approaching.

Let's check out mem. As you might have seen on another video on our channel that we are not so happy with mem right now. But that's exactly the approach, right?

When I ask anything. then it will give me the answer it shows me the sources where it found it from and then it gives me the recommendations what i want to ask next so developer incorporating magic slides and that's just stupid because it shows it's not looking into pdf file at all so probably it comes up with some random thing see i'm sorry tom i'm unable to do this and that's that's really the challenge for mem i think since day one as they were using The GPT and the GPT limitations really limits the whole note-taking app to see their whole knowledge base because the questions you ask, it will look through the whole knowledge base in order to come find the information you're looking for. So therefore, it really relies on the titles and things like that. I'm 100% sure what Tiago is showing here now, you see it's a similar approach, but the model, the AI model... has a complete different knowledge.

At the same time, restricting it to several notebooks, it still limits the knowledge that this model can access. So therefore, you are a lot more focused on specific topics that you are asking for and makes it easier for the AMI model to understand what you are asking for and what you are looking for. So, interesting approach.

You see, it's not like that notebook alarm is something revolutionary when it comes to this type of tool, as we had something similar in ChatGPT, we had something similar for years in MEM, but obviously it looks like they really refined the thing, and now let's see the results that Tiago got there. I just want something really snappy, really succinct. I could click that, it will ask that question, and there you go. It's kind of distilled it down to the three main things that I need to know. What are some ways that individuals can prepare for and respond to both the personal and societal changes that are likely to occur as a result of climate change?

So what I would be really interested in, it says here six citations. If you click there, if it then shows you the sources and actually the different places inside the doc where this was shown by. where I found it in, probably Tiago will just show this later on Let's try that one.

As it's working behind the scenes, I'm really having a conversation. I can ask it to expand or to distill. I can ask it to give me more information on one point, but not others. I can ask it to give me different formats, like write this as a poem, write it as a narrative, write it as a question and answer or FAQ format. You can see in this case, it's sort of adapted the content to really be focused on personal preparation, which is the angle that I'm most interested in.

There's one more, the third of the main three interfaces. ...that you need to know about. If you get a response, let's say this one, that you especially like, you don't want to lose it. This is a free-flowing conversation, so you might think, well, if I keep going and keep conversing, I'm going to lose track of some of these great responses. All you have to do is click this little pin icon, and it's saved to this big space up here taking up the top of the screen.

You can think... That's really nice, I think. That's something I would like in ChatGPT as well.

In this long list of conversation, that I could have some blocks that I save elsewhere. And that's exactly there. So therefore you're building up, you're building blocks in order to leverage it later.

And therefore you're extracting information from this huge amount of 4 million words down. You're distilling it to take Tiago's words there. And therefore you can leverage this in a much more focused way to build up the content that you want to get out of this.

It's a really interesting approach. Think of this as a kind of pin board where you are pinning these kind of little cards with interesting information that the AI has given you just so that you keep them top of mind and easy to see and easy to reference. One more cool detail here.

Let's say in the midst of your work you realize, oh there's another source I'd like to add. You can always click right here where it says add note and either copy and paste from somewhere, for example your note-taking app or something that you found online or something from an email, or you can just write directly into the note. For example, Let's say I remember that the creator of that course that I took, his name is Alex Steven, recently wrote a piece on Substack that is much more recent and much more timely.

And I want to also incorporate that writing and that thinking into what I'm doing. What I can do is head over to Evernote. Here is the excerpts from his article. I'm going to do Command A for Select All. I'm going to copy all that text and put it right into a new note.

And now what it will do is allow me to check this box right here and incorporate. ...corporate what I've just added right into the conversation. That was a very basic tour of three of the main areas that you are going to need to pay attention to to use Notebook LM effectively.

So I didn't get it really. So when he ticks the box, it takes it into the conversation. So if I untick it, it's not considered in the conversation. How does this work?

Also, I really like that it takes the links there, but it would be interesting, which is probably not the case, if Gemini is actually looking up. the content on these links. Therefore, it would have up-to-date content available as well. You know, that's something that I said in the beginning about the having up-to-date documentation there. So you could keep improving the knowledge in this notebook.

Well, let's keep watching. Let's see what comes up. What Google has done is basically redesigned a note-taking app from scratch with modern AI in mind. You still need to take notes, but you can just kind of dump them into one giant pile and use the AI to sort through and make sense of them rather than... Well, I don't know if Google really invented this, as I just showed you with MEM.

There are also a lot of other note-taking apps that leverage AI, just thinking about ClickUp Brain, where you leverage the whole business knowledge base and things like that. ...that so I would rather say Google is pretty late to the game when it comes to this. Having to meticulously... But that's my own opinion, obviously.

Organize things in files and folders. Notebook LM is different from other AI tools such as ChatGPT because you can train it on a specific set of documents you know and trust. That could be excerpts.

I'm not sure about that. I just showed you the GPT that I developed there. And all the other GPTs even can use external data. You can use ChatGPT API. And therefore hook into your external website and it will leverage the data from there.

So you can really train your own. GPT models are based on your knowledge base. Um, it seems more complex if you ever talk about it, then having a Google and then you throw in some documents and it gets it. So, uh, I would say the Google approach is something between mem and chat GPT that I just both showed you.

But again, it's not the only thing that allows you to build this type of notebooks or knowledge management and leveraging the knowledge. Point the AI at a specific set of documents, effectively giving it instant expertise in whatever knowledge they contain. So you always know it's only drawing on authoritative, private, trusted information.

Next I want to take a look at some of my favorite use cases that I've been using Notebook LM for. So I think really thinking about read later apps that this would be a lot more useful if I for example save this video to a read later app. and then I'm building up a knowledge base this way and just by giving it different tags instead of moving it into a closed folder I could then use these tags in order to tell the AI in the chat what it should focus on and therefore starts to only take these tags into consideration again to me this would be a lot more flexible that I can build up a very dynamic knowledge base without The need to think about, oh, in what notebook do I save this now in order so I can leverage it later on?

Let's see what the future is holding there because ReadWise ReadLater app, they have already AI. If they just add a chat there, that might be a solution there as well. Sometimes I'm reading a difficult, complex piece of writing, such as an academic paper, and I just need some help understanding it.

What you can do is upload a PDF of what you're reading. This is a paper about the role of forgetting and learning. I'm going to go ahead and select it and say, what is the role of forgetting in learning?

So you can see here it's basically summarized what is quite a long technical paper into something that I can understand. So I noticed one of the points here which is that knowledge can sometimes have a negative value, kind of peaks my interest. The amazing thing about this conversational interface is I can now ask, what are some examples of ways knowledge can have negative value. That's kind of confusing.

So what I rather would have liked here, if he would be able now to highlight this part that is really interesting for him, and therefore it focuses more on this type of information. Obviously, I can keep asking questions now about the document. Again, something that ChatGPT can do for over a year now.

probably even beyond because he didn't upload any diagrams or anything like that or business data. And so there's no, it doesn't seem that there is a code interpreter in there or something like that. It's really for, it seems it's really more for knowledge workers, which is Tiago. Obviously, he is a knowledge worker and approach and target group are knowledge workers and students, whereas we are busy professionals and we always look for ways. How are we able to integrate this into the...

daily business. So it's really useful in teamwork and daily business work. As I mentioned in the beginning, meeting notes would be a very nice example. So I can quickly leverage what this is all about. This being said, another example would be Rewind.

Rewind is a tool that you can install right now on Mac and it's recording 24-7. Whenever you have your Mac on, it's recording your audio, all your meetings, and then you can ask. This AI, what meetings did I have last week and give me a summary of the key takeaways of these meetings.

And it will literally go backwards. You can even scroll backwards and shows up the Zoom meetings and all this gives you a summary and you can leverage this. So again, I'm not so much impressed so far about the capabilities of the Gemini, the Notebook LM approach here.

Really the difference is really the vast amount this AI can. process. So let's see what happens when the other AIs will catch up to me.

So you can see here, this is fantastic. Knowledge that's incorrect, redundant knowledge, harmful knowledge. So imagine if I'm working on a project where I'm trying to convince, let's say, an I administrator.

To not save every file that's ever been created across the entire organization. That's something that I could ask it to do for me. I could say, please turn this information into a letter written to an I admin arguing that we don't need to preserve every document ever created by the organization.

And there we go. Probably need some editing. I'll definitely want to go through it and add my own touch.

But what's so neat about this is what you see right here, which is the citations. Now we are talking. I'm curious what this is now.

Let's say I send this letter to the I admin and they want to know, well, where did this come from? What is the source? What is the evidence? What's the proof? By clicking here, I can see specifically where every piece of information or fact or argument in this.

Sorry, but this implementation doesn't impress me at all. If I would expect if I click on ten citations that it will give me a drop-down or anything, giving me actually the citations and then the source, and then I click on this and open up the source. So now let's see what happens when he clicks on one of the numbers.

...conversation came from. So for example, let's click on number six here. And not only does it show you what is the original text that it's drawing from, but over here on the left, it will open up a window.

and highlight in purple that piece of text in context. This is so important to have. Yeah, I really like this, that it shows up the exact place where it found this information. But again, I have to go through all the numbers in order to understand what the different types are.

So having a quick summary or a title showing me what is behind this number would make it much easier to snipe down on the source that is really relevant for me to share. The value citations need to be shown in context. What was being argued before that appear?

What was being argued after that? What is the holistic kind of situation in which this knowledge arose? Let's do another one, which is meeting notes. How often have you been a... There we go.

Now I'm really looking forward how he will leverage this one. Part of a meeting, let's say a long meeting, and you get the transcripts of everything that was said, but they're so long and convoluted. You don't really want to spend the time to read through all those.

Even just the notes that you take yourself. As part of the meeting that you're in, this is a great use case for Notebook LM. Here is a Google Doc, which is the full transcription of a phone call that I had with a client of mine who's named Duck.

And you can see it's very detailed, very comprehensive, has who was speaking, when they spoke, what they said. So there's definitely value here. But when I start to scroll, you can see this is 26 full pages of very detailed text. So that's something we are not doing at all, as there are again a lot of tools out there already for a long time that do not only the transcript for you, but do summaries, highlighting. You can even click on the highlights out of this meeting and it will jump into the place inside the recording of the meeting, showing you the source inside the video.

So we personally, we are using MeetGeek. As usual, we are not sponsored by these companies at all. There are other alternatives like Fireflies. There's another one, Spynack, AIO. I think even Potter does now the summarizing thing as well.

There are many, many tools out there that do the summarizing for you in the first place. And obviously, you could always use ChatGPT to ask the same things to give a summary out or highlights or key takeaways or action points out of your conversation that you had with somebody if you have the full transcript available. I really don't want to spend the time to find the one or two facts or opinions or whatever it is I'm looking for from so much text.

So what I'm going to do is just copy the title of this Google Doc, go back to Notebook LM. I've created a new notebook here called Meeting Notes. Click Add Source. Let's go back to Google Drive.

I'm going to paste it in the search just to make it easy to find. There it is. You can see one selected.

I'm going to hit Insert. And what I can do now is essentially have a conversation with that. A document which if we look here has over 12,000 words. It's much, much bigger than can fit in ChatGPT for example. So what I'm gonna ask is what?

The key is really how good the AI can find information in long form text. So Gemini really excels at this point. There was this example that you can have all the different books of Harry Potter uploaded to Gemini.

And then if you ask for one specific sentence, Throughout all these millions of words, it will identify the sentence. And that's what chat GPT or GPT is really struggling with, that it comes up, as I showed before, with some information that is not true. So it's dreaming.

So that's really the superpower here. And obviously, if you then use transcripts like this, it will make a lot more sense out of this, I can imagine. But therefore, there would be the need.

to test this side by side. So I kind of wait to get my hands on Notebook LM as well to make one of these comparison videos then. Are the main points of my conversation with the duck.

So you can see here, Doug expressed interest in creating a training program. What kind of training program? Well, he believes a group workshop would be most effective.

What does he want to incorporate into that workshop? Elements of my course. I mean, this really just saved me 30 minutes to an hour at least.

So I could even go a step further. Please create a proposal. for a group workshop training program for Doug's organization.

And there we go. And here it is really important. That's what I showed you with the custom GPT. So if Tiago would have now also the additional information about his program, how much it costs, you know, the boundaries of...

all this, then it would come up with a much more specific proposal based on the general knowledge. So what I try to say here is having these buckets, I would need to add this information in each of the buckets whenever I want to recall it. So it would be great to have one global bucket for this notebook LM that AI can access anytime and whenever I ask anything about my business.

So the business relevant information would be just in this one thing. And then if I ask anything about my business, like here for Tiago's example, proposal for his workshop, then it will pull the information from this relevant information. I think that that's something that's missing.

See, it's written in the language of a proposal. There's an introduction, objectives, target audience, program structure, features of the program, benefits. I mean, this is really fantastic.

But I want to take this even one step further. I want to actually add that right in. Why not just dump in the entire text of my book and allow the AI that I'm working with to draw on those details?

Well, that's exactly what I tried to do with Chad Chippity. And I'm really curious with his book. I think it's more than 270 pages. So a lot of words will be in there. Let's see how good the answers will be then.

I'm going to go ahead and click add source right there. I imagine one thing Doug will want to see is the full curriculum. What is the full content that you plan on training them on?

So why don't I anticipate that and just say, now please, I don't know why I always say please when I talk to AIs, create a full detail. Well, when they take over the world, we want all be the guys who are nice to AI. I also say please to AI all the time.

Well it seems Tiago and I, we are just nice people, no matter if it is humans or robots. Curriculum drawing on my book, Building a Second Brain, that covers the topics above for a full day group workshop. And there we go.

You can see there's a target audience, goals, it has the morning session and what we're gonna cover specifically, introduction, the four pillars, fantastic. Sounds very familiar because I wrote that myself. Here's the afternoon session, here's the closing session.

What's so cool about this is the reason I I can feel completely comfortable uploading the full text of my book to this platform. Is the sources that you use in Notebook LM don't go anywhere else. They're not used to train any model.

They don't get shared with anyone. The second that I leave this notebook, all of that will be wiped clean from the AI's short-term memory. So those were probably half a dozen use cases that have to do with your day-to-day work, reading, understanding.

Yeah, but I'm not so sure about this with the short-term memory of AI. I would like that AI would Goods keep learning the more I interact with it about a specific topic. So whenever it makes a mistake, I can correct the mistake and say, no, no, no, you misunderstood this.

It will reinterpret the things. And then next time I go into the notebook, I know it's already trained, trained just with me and not global training. So it's specifically trained not only having access to my knowledge, but having also access to what I'm expecting.

from the AI to answer and how it should answer and so on. Another thing that he didn't mention here, I'm not sure about the tone, if you can change this, but this will be pretty obvious, that be more professional, be more casual, and things like that. Translating from one format to another.

But next, I want to dive into a very specific use case that is very close to my heart, and I know close to the heart of the people who created this, which is writing, especially writing that is creative, is creative. or that is complex or multifaceted, that is trying to simultaneously pull from many different kinds of really diverse sources while also synthesizing it into something that is your own creation, that has your own style, your own voice, your own perspective. There we really see what Tiago's passion is.

He really gets excited talking about writing there. Something I wouldn't get excited at all, but our co-founder, Duck on Harry, he loves writing as well. So therefore this is his part now there. This I want to show you is using notebook LM, almost like an editor. So this is a piece of writing that I've done.

It's in draft mode. And what I'm going to do is copy the title. I've created a new notebook here called mastermind insights. I'm going to go ahead and add a source, go to drive, paste that in the search field.

There it is. You can see one selected insert. You can see if I click on the source. It will show me what I really would find interesting if I could just select a folder on my G drive and then I can keep adding things to this folder. That would be really useful so I don't have to search things all the time, but I can move the information something like I mentioned before with the Relator app where I just give the tags and I just throw the information there and I don't care about the moment I information, but it's ready whenever I need it and to retrieve it.

That I think would remove a bit of a friction if I want to build up a knowledge base for a specific project or writing project, for example, here as well. What they call a source guide, which includes a summary and key topics. Let's exit out of that. And what I'll ask is, please suggest ways this piece could be improved.

So what's interesting is it's giving me lots of ideas, but what I wanted to actually do is say, please rewrite this piece in full following the instructions above. It's funny, it wrote it in a bullet point format. So I'll ask, please convert this from bullet point format into prose. Now, This is already probably better.

It's certainly much more succinct, which I'm always trying to go for. There's one more thing I can do here, which is add a source, go back to Drive, and let's see if I have any interesting sources on writing. So this is one that jumps out at me immediately. I can tell from the title, it's a my notes on a book called Write Useful Books, which is really about non-fiction prescriptive writing.

So I'm going to insert that and say, please rewrite. My article on takeaways from my first mastermind, drawing on advice found in right, useful books. It's interesting. I'm curious looking at the citations, which tell me some of the specific ways that it did that. For example, not using titles or headings that are clever or punny.

because that is interesting or entertaining but not useful is a great piece of advice for this kind of writing. The next use case for writing I want to show you is really earlier in the ideation process. Let's say you're not even sure what you want to write in the first place.

You just kind of have some messy notes and you're looking for a collaborator that can help you give them some meaning and some structure. I'll go ahead and create a new notebook, which is a new article. Let's say I don't even know what it's going to be about.

I'm gonna go ahead and add a source, go to Drive, Let's go back to Readwise, which contains all the notes and all the books that I've read. And what I'm going to do is pick one, let's say, that I've read recently, such as A History of the Index. And what I'm going to ask it is, please suggest related ideas or avenues of research related to this book.

So it's giving me ideas from indexing prehistory to alternative indexing systems. This book is about indexing the impact of digital indexing. Each one of these is potentially a rabbit hole that I could run after.

The one that jumps out at me is historical evolution. And just seeing that reminds me that I have another book that I've taken notes on in the past, which I believe was called Glut. There it is. So let's add that in, which is more about the past and about how our relationship to information has changed over time. So now I have two sources I can ask, please find connections or relationships between the ideas in these two books.

Let's see if it found anything interesting. So this is the one that strikes me. I would like to write a piece about the democratization of knowledge, how access to knowledge and ways of structuring knowledge has become more accessible over time.

So I'm going to say, please suggest a few ideas for articles I could write about the democratization of knowledge. through history. And these all kind of strike me as somewhat generic.

What I think I would like to do is use a specific story, use a specific example. And what comes to mind, I don't know why, is a book I read about paleontology. Let's see if I can find it. Yeah, it was called Fossil Men.

Let's add that in here. And now I'm going to say, suggest the title of an article about the democratization of knowledge using an example. From paleontology or fossils.

Okay, that's an interesting title, but now I'm gonna say expand on that title with an outline of key points this piece could contain. Interesting. Fossils in the history of life, fossils in human origins.

So it's really going hard on this idea of fossils, but kind of weaving together or weaving through it different ideas about democratizing knowledge. This would obviously require more development. element I'd want to go in and look at the sources that it found, maybe ask some of these follow-up questions.

Let's say I'm satisfied with this initial outline for a piece. I could even go so far as to say, please write the opening lines of this article, making them as attention-grabbing as possible. Like a hidden world awaits, a chronicle of life frozen in time.

This is fantastic. Just notice what we've done. We went from not Not even knowing what I wanted to write about, really just a single source that I had been reading recently, all the way to the opening lines of what sounds like a pretty interesting piece if you ask me, through a completely conversational back and forth collaborative exercise with an artificial intelligence is totally amazing.

And that's a very important point that Tiago points out here, that people really need to understand how to approach AI. So for us as well, it's really just... to get started so you're no longer there's no longer an excuse that you're sitting on a blank canvas on a blank piece of paper and you get writer's block or something like that there are hundreds of ways to get your thought process going and then you go down a rabbit hole in a conversational style with ai and i think this is really the superpower of ai this is how we use it with chat GPT on a daily basis.

And the results don't need to be perfect, obviously. Otherwise, we would be swamped with only AI responses. But it gives you an idea.

So when I approach something like this, I usually always ask for, give me five titles or give me even 20 titles, okay? And then out of these 20 titles, there's one that really caught my attention. And then I ask, give me five more variations of this specific title. And therefore...

I can keep refining and then I exchange one or two words myself and therefore I still have the feeling that I developed this result together with AI but it's still original it still feels human the moment I use this content then and I think this is really key how we can use the AI in this case. Now you may have noticed that what allowed this, what made this possible, was that I had all my notes and highlights from every book and article I've read over the past few years already in Google Drive. But you can do that too, and I wanna show you how.

It's possible with a really powerful tool called Readwise. Check out our in-depth video on how to use Readwise for more information. But what you can do is go to Readwise, which is a... paid service and once you create an account and log in you can click dashboard here if you go to the import section you want to make sure that under the connected section right here you have your Kindle ebooks or your relator app or social media or whatever is the source that you want to be able to save in Google Drive Which you can then use in notebook LM. So for me, the two main ones are Kindle right here and Readwise reader, which I use to read online content.

Let's head back to the dashboard. And then the second thing to look at is under export. You want to be sure that Google docs is listed here under the connected section. Normally it will appear down here. It's really amazing that allows you the export into the.

G drive, this not only allows the notebook alarm to look at this, but also again, I mentioned ClickUp Brain, but even inside G drive, you can search for the information and it will find the information you're looking for. As it looks into the content of these notebooks, I think this is really the way it should go. So that's exactly what I said that I can now throw the stuff into my read later app, which was shown there as well. The Readwise Reader read later app.

and it connects to your readwise and therefore it saves then your highlights from even youtube videos and other things into g drive this is really amazing you have to do is click connect sign into your google docs account the same account that you use to create the notebook lm account and then google docs will appear right here if you click into the google docs settings the key feature that you want to pay attention to is this third one right here if this is not checked Then Readwise will send all the excerpts from each source to its own individual Google Doc. So the excerpts that I saved from one book will go to one Google Doc. That makes it really easy to go through like you saw and select only the sources that you want to use in Notebook LM. But if this is turned on, what it will do is it will send all the highlights to one, or maybe if you have many highlights, just a few Google Docs.

so that you can fit it within that 20 notebook limit. What that allows you to do is essentially have a conversation with your entire history of highlights from whatever sources you imported into Readwise. This is really powerful, guys. That's really powerful.

Once you click here where it says start export to Google Docs, you'll want to leave it running for probably a few hours depending on how many highlights you have. But once it's finished, you'll see a Readwise folder appear in Drive. Alright.

So it's not syncing. I have to double check, but it would make sense that it keeps syncing because I don't want to get back into readwise every time and then start syncing. It's another friction point there. Probably it will automatically sync there. I know that was a lot.

I want to turn now to the other side of the story. There's always another side. Which is some of the limitations and pitfalls.

This is an experiment's brand new platform, which means it's a little rough around the edges. There are bugs. They're having to invent a completely new interface, which means there is a bit of a learning curve.

You have to figure out what the different buttons do and what the different parts of the window mean. Currently, you have to provide the sources to Notebook LM in the form of basically Google Docs or PDFs. It doesn't have the ability to directly connect to a note-taking app. Such as Evernote or Notion or even Google Keep.

You'd have to use something like Readwise as an intermediary to get your notes into Notebook LM. You're limited to only 20 sources at most. So if you have more sources than that that you want to use, you're going to have to do some manual copying and pasting.

That could become interesting for scientists, but they have loads and loads of papers they are working through. 20 is not... enough there obviously. ...to get them under 20. Just like a lot of other AI tools, Notebook LM has trouble doing math, reading PDFs that have messy formatting, and finding a lot of precise details if the sources that you're using are really large. I've also talked to the...

That's interesting, something that I mentioned earlier in this video, that this should be the superpower of Gemini, but maybe it's not part of the Notebook LM thing then yet. Notebook LM team. And they said fixes and improvements to a lot of those issues are coming soon. This is way more than a note-taking tool, way more than a writing tool, and it's way more than an organizational tool. Here's how I describe what Notebook LM is trying to do.

Until now, the stages of the creative process, like reading, researching, note-taking, and writing, were completely separate. We had to use completely different tools for each step, which means we were constantly having to switch between them and worry about things like formatting and compatibility and security and privacy no one software tool could do it all notebook lm i'm not sure what about notion q a where you can ask your database i just mentioned clicker brain asking your database what about mem where you can ask your whole knowledge base at the same time you can read it you read the things for researching i wonder well you do research on the sources that you add to notebook lm however Researching, in my opinion, is it should also be able to access the web and using and search the web and find relevant and latest information, maybe even access actual paper and articles and studies and things like that. So but the things that he is talking about is very revolutionary. It's already happening for more than a year now.

And we will just get used to this style of reading, researching, writing and note taking in one app. If it is notebook.lm, I don't know. The superpower of notebook.lm I see here, it's for free as it is part of the Google workspace. I think it would be really interesting if there are specific sources that you can add and log and then give your team access to it so everybody could ask questions about the specific knowledge base.

These are the things that we do in ClickUp right now, where we just We have all our SOPs, the standard operating procedures, the work instructions, but also the relevant projects and all the things that we worked on. And there you have also a chatbot that you can ask anything about your business knowledge base inside ClickUp and you will get relevant information and the sources as you have seen here with the citations and things like that. So as I said already, to me, Google is late to the game here when it comes to this innovative product here.

But it will become the new standard. That's the point. And other note taking apps like Evernote certainly have to catch up. When they introduced their AI search, it was still behind what other tools were already capable of.

And now let's see how quickly Google can adapt to this new pace there. Gets all those steps and folds them into a single integrated space so you can remain in that wonderful flow state and keep moving forward. They call it working at the speed of thought.

It's an AI collaborator trained on the data only you possess with your unique view of the world. I suggest signing up for a free account at notebooklm.google.com. Let me know what you think about Notebook LM in the comments.

Anything Google... I have trouble saying the word Google. How do you say that word?

Google. Google. Google. Google.

So it seems obviously that he talked about the Notion AI. If you want me to react to the other video here as well, especially in comparison now to the Notebook LM, as he pointed out all the advantages, I'm really curious about this one, but I won't watch it before I do an actual reaction video. So go to the comments below it if you'd like to see me reacting to the other video as well. I think it's more like goo like you say the GOO first Google good it's really interesting I mean I'm not a native speaker at all I'm overthinking all the pronunciations as well but it's good to see that native speakers even think about how to pronounce Google Google Google As it is just a made up word, right? Google.

I would maybe refer to the Google, but which would be then goggles. I'm not getting into this conversation either. Otherwise, I cannot speak a word anymore as I'm overthinking things.

How do you say that word, Alex? Google. Just like a normal person, basically. Like a normal person. Everybody, I really enjoyed this video.

Thank you very much for all the insights. Tiago was really interesting. I think it was very comprehensive. And well, I cannot wait to catch you guys up in the next one.

Transcript for:Tiago's Reaction Video on Notebook LM

Transcript for:
Tiago's Reaction Video on Notebook LM