NotebookLM for Researchers: Source-Grounded AI Workflows (Full Transcript)

[00:00:00] Speaker 1: Hey, ever since I did my ranking of the best AI for 2026, several of you have asked, well, tell me more about notebook LM. I didn't think that was going to be at the top of the list. And there's specific reasons why I recommend it so highly for academic purposes. And I'll go into that today. But let's start here with their own website, understand everything your research and thinking partner, you can already see notebook LM is postured differently from say, chat GPT or cloud, which I also use quite heavily. But here it is as an AI research partner, it does something valuable, upload your sources. I used to love different AIs for this to upload sources. But here, you can create your whole reference library that you can engage with and interact with right through notebook LM with all the sources, as it says here, it's very good at staying true to the sources, it doesn't hallucinate as much as the other AIs because you have limited its data set to work with. I think that's fantastic. Another great thing, the study guide really useful if you're trying to learn for a course. I mean, we focus at fast track more heavily on research. But if you want to learn something quickly, there's a body of literature where you don't understand something, this can also be a valuable use case, it also takes you to the source. So I sometimes call this citation 2.0 is that when it takes you the citation, you click and I'll show you this in a minute. And it goes right into the PDF and not just tells you which PDF or which article where that point was found, but where inside of it, it was found. So this is really fantastic. I love this. And this audio overview, I often say if you want to activate some dead spaces in your day and be more proactive, maybe you're on the metro or you're driving to work, you can listen to your literature, like a podcast or have an audio overview here, which is really fantastic to help you to learn on the go and customize your literature to engage with it and even converse with it while you're moving. So let's have a look at an actual notebook I created and what I've done with it. So here's an example of a notebook I've been working with that's looking at a range of different papers for research project on the causes and developmental causes of autism. And you see, I've got a whole bunch of sources here. And I can very easily drag in new sources, if need be, I can click on each source, and it's going to pull up the actual paper and I can then directly chat with it. It's going to give me an AI summary and other things to dig into it, which is also useful. But a lot of the features that you'll find here that you can make with it, some of them you may not use like an infographic, you'll see a ton of these now flooding the internet, flashcards, video overview, where they actually explains the sources to you as though it were a lecture, very cool use cases, great for accelerating learning. But what I love here is is this chat. So if I go ask this a question, it's going to give me the site and take me right into the paper in a highlight. Where is that? Where's that point coming from? So I can check it myself. And this is key. With AI, you always need to get in the habit of check, check that it's accurate, check what it's saying is true, check it always here. But this just makes it so much easier to check in a way that you can't as easily do at the moment with cloud or others. And that will be tempted to go outside your source of data. And you won't necessarily have curated the same quality as what you have uploaded to your library here. If you'd like our help in setting up AI enhanced workflows for your research, click link below and see if you'd be a good fit. Now one thing that it doesn't do well that I wish it could do better is engage with your reference manager to spit out formatted references for you. But really, I believe that should live with a Zotero, which is a great piece of kit. And just like notebook alum, 100% free and valuable to use. So let me give you an example here. Let's say what are the developmental causes of autism? And what is the strength of evidence for for them? This will dig into my 21 sources and ask it specifically those questions. So let's see as it sifts through the pages what it comes up with. While that searching, one thing I don't love searching the web for new sources, I recommend doing this in Google Scholar. And that's just that old fashioned Google Scholar search algorithm works. Well, it's tried tested and true. There are other integrations you can use for AI searches, but those just have partial coverage of the literature and you're kind of intentionally ending up with a blind spot. It's like seeing with one and a half eyes or something. So I don't love that. So I wouldn't use this feature so much in doing your research. Other things like mind map. I don't love these personally. They're like cute tools. If this helps you learn great, but for actual serious research, not something we've ever done. So here, this will give you some nice summaries. And again, you see this drawing on the spectrum of the studies, clustering it, this is going to be so much better than what I've seen researchers try to do that's created problems of hallucinating references of just ask chat GPT a question or Claude and have it answer that here, you're going to be dealing with real studies. And like I said, you can go check it. Look at this. This is so clear and so nice. And it can take me right to the paper. So I don't have to sift through the paper and try to extract that piece of information. If I'm trying to do a quick and dirty literature review for a paper, or I'm just trying to justify the existence of my paper, why my paper needs to exist, which is very common in your short introduction to a research paper, this is a fantastic way to save some time without cutting corners. So you can see here, it's really done a nice job clustering the different explanations. And I wouldn't put too much on this, but it gives me some ideas on what are the more strong explanations in this case. So trying to ask it, not just what evidence is, but also what the strength of the evidence is, will also give you kind of safeguard that AI is not just feeding you a bunch of junk, or at least force it to give you some sense of his confidence in the literature. Again, take take this part with a pinch of pinch of salt, because this is now the AI interpreting for you. But at least it's giving you a justification. At any rate, I hope you can see this is an incredibly powerful tool. I think you're gonna love it. And I think increasingly tools like this that preserve what humans do, but use what machines can do faster, like sifting through sources to find what you want through a large body of text. These are use cases of AI I highly recommend 100% ethical, you're not going to get yourself in hot water here, because you are going to be able to check it. So if you're using notebook LM, let me know comment below how you're using notebook in your own workflow would love to hear from you. And if you want to learn more about AI enhanced workflows that stay on the side of angels and won't get you into hot water. Click the link below to learn about our mentorship communities where we use AI all the time to save time and produce high quality research, not just AI driven slop. Alright, see you guys in the next video.