Google Notebook LM Demo & Review
Notebook LM is a new free, AI-powered research assistant that helps users organize, analyze, and create outputs from their research materials. We first had a chance to check it out last year when it was demoed at Edtech week. The AI for Education team has had a lot of fun playing around with the tool now that’s been released out of beta, and can really see the potential value for schools.
In this video, Amanda showcases its key features for educators, including the ability to consolidate information from up to 50 different sources—such as Google Docs, PDFs, and text files—into organized notebooks. Among the notable functionalities demonstrated were the ability to summarize uploaded documents, and a chat feature that cites sources in responses, ultimately aiding in the creation of study aids like quizzes and FAQs. Amanda also reviews the podcast generation feature, which produces dialogues in an NPR-style format but comes with a warning regarding potential inaccuracies.
Key Highlights:
Recently moved from beta to experiment mode, now available to everyone for free
Uses Gemini, a large language model with a 1.5 million token window
Allows users to synthesize and interact with large amounts of content
Can create notebooks using various input methods (Google Docs, links, pasted text)
Supports up to 50 sources in various formats (PDF, text, markdown, audio)
Can process audio from YouTube videos
Has a citation feature, which links information to specific sources
Still prone to some hallucinations and misinformation
May sometimes flatten or exaggerate differences between documents
Use Cases for Educators:
Creating study guides, FAQs, table of contents, briefing docs, and timelines on research papers, curriculum, and other sources
Generating podcast-like conversations to reinforce important content
Help for students studying or researchers synthesizing information from multiple sources
Check out the whole video review for more!
-
Hi everyone, it's Amanda Bickerstaff, the CEO and co founder of AI for Education. Really excited to be here with you today. It's been a while since I've done a video, but I really am enjoying notebook LM and so is our team. It's also something that we're seeing to have potentially a lot of value in schools. And so notebook LM is a Google experiment. It's actually something that we first saw almost a year ago today at an event during Ed tech week. And it's really exciting. Just over, I guess, a couple weeks ago, it went out of beta into experiment mode, and now everyone can try it. And what's really interesting is that notebook LM is using Gemini, the large language model. It's 1.5 million token window version.
So it has this ability to have an enormous amount of content and context to give you this opportunity to start to make sense of all the documents that you have and really start talking to them and synthesizing them. While it is working, absolutely to lower hallucinations or misinformation, it's still going to mess some things up. Remember, early stages of all generative AI technology. But I want to show you a couple of things to get started. So first of all, it's free, which is super awesome. And then you're going to have to make sure you look at the terms of service. But what you see is that you can create notebooks. And so these are going to be examples that they've created for you.I don't think we need to talk about mushrooms, but what you can do is you can create a new notebook and I'm going to create a new book a couple different ways. You can do it through Google Docs, you can do it through links, you can do it through pasting text as well. It allows you to do up to 50 sources, which is pretty cool. And you can see that it has all different types of. So you can do PDF, text files, markdown or audio even. And now, so you can actually upload. They have a connection now where you can do the audio from a YouTube video or just from even this video. So what I do is I'm going to actually upload. You guys know us that we love doing all things education and guidelines. And so for AI.
And so these are going to be some examples of some guidelines that we've either helped create or really believe in. And so I've got both North Carolinas piece that was done by DPI there and Vera and shout out to all of our people in the southeast impacted by the hurricane. It's a Park Hill guidelines, which are pretty simple, that were done with our wonderful partner out in Kansas City. And then I'm also going to put in Chicago's guidance, which was released a couple of like last month. And what you can do now is they're going to be uploaded and they're going to be essentially parsed. And so the large language model starting to like parse all this information.
And what's really interesting is you have this notebook guide and what you're going to notice is that's really interesting because I have two versions of this now. It seems to be like you kind of have to pick the right content. So it's, Park Hill is so little in comparison to the North Carolina and Chicago guidance. And it's kind of getting, it's not even in the summary. And in fact, if you start creating things, it might pick up Park Hill or not. And so that's going to be example of like always think about the use case that you have in this case, maybe I shouldn't have if I didn't include North Carolina, maybe, sorry, Park Hill, it might be a better piece because it's comparing two larger kind of documents together.
And what you can do though, is you can start creating different things and so you can chat with the documents, so you can say what are the similarities between all three documents? And what's going to do is it's going to go, and what I love about this is it's actually going to give you answer and it's going to cite back to where it is within the document. So you can actually double check it because remember, everybody, always double check for hallucinations. So we have guidelines for flexible use, ethical considerations. And here we go. So now it's actually pulling up which source it's in. So this is in Park Hill. This is going to be in Chicago public schools. This is going to be the North Carolina guidance. You can double check the content to see where it's being pulled for.
We love that, especially if you're thinking about this with students that are potentially using this to help study, is that they can double check to make sure that it is accurate. You also can build study guides. You can build faq, table of contents, briefing docs and timelines. So let's do, let's actually do a faq. And it's going to take a little time to do that. So give it a little bit of patience. So I'm actually going to go back to the one that I've already created while we're waiting, because I want to show you something else. It's pretty interesting. I will say it get kind of confusing to use the UX. So always, like, hit this notebook guide if you don't know what to do. And what you'll notice is that I have here is I can generate.
Oh, this is actually the same one. So let's actually go back to the guidelines that I've already done, because this one's already been done. You can see I've done a couple notes, and what you'll notice is that, like, this is only comparing the two, and it kind of left out Park Hill on its own. But what I can do is I have created, essentially what's really fascinating that people love is that you can create a essentially a podcast. It sounds very npR. I am not a podcast person, and so I find the voices to be uncanny. But I know that, like, mandy and our team is, like, starting to give them, like, personalities, which is really funny. But what I'm gonna do is I'm gonna go to a portion of the this to give you guys an example of what's happening.
So let's listen along. Totally.
And that's where their sweet spot is, the e bike. You're still in control pedaling, but you get that extra help to conquer those hills and go further.
Okay, I'm with them so far.So AI is like, that e bike boost for learning. It's there to make us better, not replace us.
I like it. It's a really balanced way to look at it. Not all doom and gloom, but not ignoring the challenges either.
So, hey, guys. So, like, fascinating, right? What you see is it's always going to be two people, and they're going to take these kind of very NPR piece. I will say, if I had listened to this, because I know these documents very keenly, I could tell that they were trying to create the podcast is trying to create some kind of narrative that they're really different. And so it's kind of creating some hallucinations and flattening and also making more extreme the content within the documents. So what I would say is whenever you're starting using a new tool that's generative aihdenae, you always want to start with the first use case that you're very familiar with or you're expert in, so you can kind of test the limitations and start to understand how best to use them, because these are experiments, right?
And there's not a lot of really good indications of how to use this but what I will say, though, is there's some really cool things to do here. We've had Corey and our team created just like a funny, like, chipmunk themed podcast. We've had people maybe give up a little bit too much private information to create podcasts about themselves. We do not suggest using your real information, but you can also, like you see here if a student or even you are studying as an educator for an upcoming, you know, test or assessment, you could use this to build a study guide. You can use this to help you ask questions. The documents, you can help it identify key areas.
If you're doing working with a research paper and, you know, at this point, you're, like, completely out of time and out of, like, attention just to help you reassert that the stuff that you are using is the best stuff you can be using from these research papers. I think that is a really cool opportunity as well. But as, again, I hope that you guys enjoy as we walk through some of these new instances. So here you go. Here's that. AI literacies, you got quiz, you've got answer key, you've got essay questions. These are just some really awesome ways to start really working with generative AI in meaningful ways with these new tools. So I hope you enjoyed it and learned with us together. And, yeah, keep experimenting. That's the best thing you can do right now. Thanks, everybody.
