How to Learn FASTER With AI - Google NotebookLM

So, a while ago, ChatGPT released their study mode, which was to help people learn using AI, and Google has their own version, which is called Notebook LM. A lot of you asked me to review it, so I spent several hours testing it, learning through it, really trying to make the most of it, and putting it through lots of different real life situations. And I have to say that as soon as I started using it, I was pretty impressed. And the more and more I used it, the less impressed I got. I do think it can be really useful, but I think that if you’ve tried using it for an extended period of time, you’re also running into the same limitations as I experience probably. But the interesting thing is I think there are a lot of people who think that it’s really helpful for learning, but it’s actually more of an illusion of learning.

So in this article I’m going to go through what Notebook LM is and how it differs from ChatGPT. What’s good about it from my perspective as a learning coach who’s worked with thousands of people and understands the learning science, as well as the parts that are not so good and my recommendations on how you can use it to make a meaningful difference to your learning efficiency.

What is Notebook LM?

So first of all, what is Notebook LM? If you’re not familiar, it’s Google’s version of ChatGPT’s study mode. It’s their all-in-one AI tool for learning. And while it shares a lot of similarities in terms of how it behaves with something like ChatGPT or any other LLM, one of the key differences is that it’s clearly built as a learning support product. It also consolidates a lot of really useful tools like you can get audio summaries, you can get video summaries, you can get it to generate mind maps for you, you can create flashcards automatically based on the material, you can create quizzes for yourself using the material. So, it really is pitched as your all-in-one AI learning tool.

Testing Notebook LM for Learning

Let’s talk about how good it actually is for learning. Now, to test this, I ran through a couple of different scenarios. First of all, I looked at it from the perspective of a beginner learning a new topic. The topic I chose to learn was some AI fundamentals. After spending a few hours learning through that, I opened up a new notebook and looked at it from the perspective of an expert or an advanced learner trying to keep up to date with some latest research. So for this, I used self-regulated learning and higher order thinking and new research on how AI affects that. That’s a big part of my existing domain of expertise. So I’m generally pretty familiar with the latest science on this. So I wanted to see how good it is at really getting into some of those detailed nuances.

Now for each of these different learner scenarios, I also evaluated it based on common use cases. And there are three common use cases that I wanted to evaluate it against. The first is intensive study. This is a situation where you are sitting at your desk, you’ve got enough time, you’ve got your resources, you’re dedicating hours to actually sit down and really learn through it. That’s the first use case. How good is Notebook LM at facilitating that?

The second use case is on-the-go learning. This is when you have 5, 10, 15 minutes to top up your learning. You might be on the bus commuting, waiting for something. If you’re trying to fit in a little bit of learning while you’re at work, this is usually the type of learning that you’re going to be doing. And for a lot of people, this is actually the bulk of what their learning is.

And the third use case is task reactive. So this is when you’re learning specifically just to complete a task or a project that you’ve been given. You don’t really care about building deep expertise. You just need to deliver something and learn just enough to be able to deliver that. This is actually probably the most common way that people tend to use AI for learning. And a lot of the time just base ChatGPT is already pretty good at this. So I wanted to see if Notebook LM is any better.

Now in total I probably spent about 6 or 7 hours on testing it and really learning through it. I don’t like when people review learning tools based on what features it has. You’re not learning to experience the features of a software, you’re learning to achieve a goal. And so as I was testing, yes, I was trying to think about what features they are and how helpful they are, but it was how helpful they are to me actually achieving the goal of learning in the first place.

The Real Goal of Learning

And this is the part that I actually want to start with first before going into my findings because I think that a lot of people feel that these AI tools are helpful for their learning because they’re not actually measuring on the right goal. So just think about this for a second. Your win criteria whenever you’re learning something is to learn, gain whatever knowledge or skills you need in order to be able to achieve a certain objective.

And this objective could be anything like this could be completing a task, managing a project, leading a team, it could be you know building your own expertise to be able to make decisions and solve complex problems. It doesn’t really matter. But the idea is that there’s a certain level of knowledge and a certain level of skills you need to be able to achieve this objective. And in order for new information to come into your brain for you to learn it and for you to actually achieve the outcome that you want at the end, you need to be able to understand that information. You need to be able to retain that information at least long enough to be able to use it in the right way. You need to be able to understand how to actually use and apply it in the way that you need to, which depends on your objective.

For example, if the objective is I just need to know this so that when someone asks me a question, I can just spit out a fact. That’s a very different type of objective to I need to be able to manage an entire complex project with lots of moving parts and communicate that to external stakeholders who could be asking me any sort of question. That’s a much higher level of expertise that you need.

And so remembering that you only win by achieving this objective, if you spend a bunch of time learning new information and you feel like you’re gaining lots of new knowledge and lots of new skills, but you’re not able to answer those questions or solve those problems, you haven’t achieved that objective. It’s actually not productive.

And to be able to apply your knowledge in the right way, especially for more complex situations, not only do you need to understand and retain what you’ve learned, but you also usually need to see how what you’ve learned connects, how it connects to each other, how all the little pieces come together to serve your purpose and fit into a big picture. Most of the time when you’re learning something new and you feel like, oh, this is getting complicated, there’s a lot to learn here, there’s a lot to unpack, and you start feeling a little bit overwhelmed. That’s because you’re recognizing there’s a lot of moving pieces here. I don’t really see how they all connect together. I know they do connect together, but I don’t know what that is and I don’t know how to get there.

And usually when people hit this wall, that’s when they look for shortcuts or alternatives. They ask themselves, “Do I even need to learn that? Do I even need to go there?” Or they’ll use AI tools to try to make that part easier. But most of the time, if you do truly need to hit the objective of using your knowledge at a high level, you need to figure that out. You need to be able to organize and navigate through that overwhelm to create meaningful knowledge that is actually connected.

After you’re finished your learning process, however long that takes, you need to be able to look back at that same new information that used to feel overwhelming. And you need to feel this makes sense to me. I feel this is intuitive. It was overwhelming before. I’ve organized it. Now it makes sense. That’s what creates confidence and that’s what translates to actual results and performance.

Now, you don’t always need to hit that level for everything that you learn, but the real problem that AI hopefully can help us solve is when you’re trying to hit that higher level and it is feeling overwhelming. And so, that was my win criteria. I was thinking about it in terms of learning something to be able to use it in this complex way, explaining it to someone simply. I wanted to get to a point where I feel truly and genuinely confident with that knowledge. And when I look at Notebook LM and ChatGPT study mode actually through this lens, it’s not as impressive as I wish it would be. And I’ll explain specifically where the issues are in a bit.

What is Notebook LM Good For?

But first, let’s go through the good parts. What is Notebook LM really good for? So, I’m going to measure this based on six different criteria. Ease of use, accuracy, effectiveness. Effectiveness being how useful is it for actually achieving the goal like I just talked about. And then the three different use cases is it good for intensive learning, on-the-go learning, and task reactive learning.

Now one of the best things about Notebook LM, both as a beginner learner and as an advanced learner, is how easy it is to use. So if you look at how I actually used it for learning some fundamentals of AI, you can start by just discovering sources which is where you just say what you want to learn. It will then find some relevant resources for you which you can then screen and vet and check that they’re the ones that you want to include. You import those and automatically straight off that it generates you this summary. Anything that’s generated inside the central panel can be turned into a note that is basically saved like a bookmark for you to come back to.

And like a normal chat interface, you can then have a conversation asking it questions back and forth, and it also recommends questions that might be useful for you. One thing that you might note if you’re familiar with learning science a little bit is that the questions that it’s generating for you are relational higher order questions. The idea here is that it’s getting you to ask questions that make you think about how each piece of information is connected to other pieces of information. And so that’s clearly an attempt and a very deliberate attempt by Notebook LM to get you to engage in that higher level of thinking which is necessary to achieve those more complex outcomes like being able to see how it’s all connected together.

You can also see that you can generate a video overview, you can generate audio overviews, mind maps, reports, flashcards, and quizzes. It’s pretty convenient because you don’t have to wait for one of them to finish. You can basically just tick all of them if you want and it’ll just generate in the background. But as it’s generating, you can still be exploring, you can still be reading, you can be generating other things. That’s useful because something like the video overview often takes several minutes for it to generate. So you don’t want to be just sitting there waiting for it to generate a video the entire time.

I thought the mind map overview was pretty interesting and as you know, I’m kind of like the mind map guy. So I clicked into that and it does a reasonable job at separating out this topic into logical categories. And this again helps you to see sort of a big picture overview of how everything connects together. And again that’s going to be important for you to achieve this higher level outcomes. So so far if you compare that to the experience of using ChatGPT, even in study mode, already it’s doing pretty much everything that ChatGPT study mode would do and it’s doing it better and there’s more things built into it. Already it’s starting to seem like Notebook LM is a superior tool when it comes to learning compared to ChatGPT and that there’s a big difference between the two.

As I started expanding some of these concepts and reading into it a little bit more, I definitely started feeling that sense of overwhelm like wow, there’s a lot of complexity to this. This is getting really complicated. How does it all fit in together? And again, that is that key wall in learning that is very difficult to break and it takes a lot of training and skill and expertise to be able to learn through that. And so as soon as I felt that, instead of using my normal strategies of how I would normally learn through it, I tried to stay a little bit more into the software and see if there was anything in the software that would help me break through that a little bit more easily.

One thing that often helps with reducing this overwhelm is implementing dual coding. Dual coding is this learning phenomenon where learning is more effective when you’re able to process information through both written words as well as visual. So for example, like diagrams and pictures. So, one of my quick tips is just whenever you’re learning a new concept, do a Google image search on it to see what diagrams and infographics people have already created about it. And that usually ends up speeding up the amount of time that it would take to understand that concept by a lot. I wanted to see if I can just get it to create some of these images and infographics for me and to see how accurate it was. Unfortunately, it wasn’t very successful at actually doing that. I tried it a few more times with some other things and I wasn’t really able to get something that I thought was meaningful. I ended up just doing a Google image search anyway.

One thing I did notice though a few times is that when it recommends a source and I want to read into a little bit deeper, it’s a little bit hard to navigate the source guide that appears on the left. The formatting is often broken, especially if it’s coming from a PDF. And when you try to find that relevant piece of information from the original source, it’s really hard to navigate and often doesn’t really link up correctly. So half the time I was left kind of just having to believe that this is true and not being able to read deeper into it.

One of the sources that it recommended was this book called Mathematics for Machine Learning. And after reading through a few of the pages here, I did actually find that just reading the original book that it’s referencing was a lot easier to understand than the summary that it generated for me.

The Audio and Video Overviews

The other thing I was really curious about is how the video overview works. So, I checked out the video overview, played the video, and honestly really impressive. It gave me this really good big picture summary about how to think about everything. It covered the major concepts, explained them in a way that was clear and easy to understand. And I can tell that whoever designed this particular tool at Google, they for certain worked with learning scientists and researchers to make sure that the method of teaching follows best practice for direct instruction.

It genuinely does feel like given a little bit more time for this to refine and improve, you will be able to get to a point where you can just load up a bunch of resources and it’s basically generating you a personalized set of video lessons and curriculum that you can just learn off of. That is huge because of the fact that one of the biggest limitations when you’re learning and you don’t have a structured curriculum is that you just don’t know how much you need to learn. And a lot of the confusion around learning just comes from the fact that there isn’t a clear structure or a clear goal. So by it being able to kind of create this curriculum for you, I think it’s a huge step in the right direction.

Likewise, the audio overview was surprisingly pretty good. What it does is it creates this podcast using the material that you’ve given it and there’s one female voice and one male voice and they’re sort of talking to each other about the topic. It sounds sort of realistic. And there’s this cool interactive mode where you can kind of join in on the podcast. So, as you’re going through the podcast, you can click join. You can ask questions about it and then it will answer you. Kind of like, you know, calling into a radio show and asking your question.

Now, one of the biggest use cases for this style of audio overview and this kind of podcasting is that you can now start learning on the go. So when you’re on the commute, when you’re driving somewhere, you can have that playing and you can be talking to it and then learning in this period of time when normally you just can’t do anything useful in that time. You’re falling asleep in traffic. And the Notebook LM team clearly thought about this because this website works very, very well. If you load it up on your phone, the interface changes, so you get these tabs. If you have an audio overview playing, you can actually swap between the tabs. The audio does not get disrupted.

There is also a quiz function on this which just generates you this multiple choice quiz. You can mark the answers that you think. It will tell you whether you’re right or wrong. You can get it to explain the answer. By default, the quiz that it will generate for you is a little basic, which might be okay if you are a student studying for an exam. If you’re trying to use your knowledge for more complex problem solving or complex applications, it’s probably not going to be enough. You can just tune it and give it more guidance and instruction on the level of questioning that you want it to reach. So, if you’re going to use quizzes, I would definitely recommend spending a bit of time defining the types of questions that you want at the level of complexity as opposed to just using the default button where you just click the quiz button and it just generates it for you automatically.

Now, the other major function is flashcard usage. Now, the flash cards are probably the one that I think is the least useful out of these because first of all, you actually get a very similar level of knowledge challenge with the quiz. And the quiz is actually a superior way of testing yourself anyway. So, the two are kind of redundant. Like, if you’re going to do the flash cards, you don’t really need to do the quiz. If you’re going to do the quiz, you definitely don’t need to do the flash cards.

But the main limitation here is that even though it generates a series of flash cards which are fine, there is no way to export those flash cards and there’s no space repetition timer on it. One of the biggest benefits of using flash cards is that allows you to tackle lower order specific fine detail and train your recall through spaced repetition on the go. It’s sort of the most brain dead way that you can improve your retention on discrete isolated little facts. And it’s an effective method for that. You chuck it on your Anki or whatever app that you’re using and then a little notification pops up and it says, “Hey, you need to do your flash cards because if you wait too long, you’re going to forget them.” And then you open your app and you do them. It saves it on its algorithm. It reminds you to do it again later. And so the main benefit of flash cards is not the fact that it’s involving active recall because there are many other ways of engaging active recall that are vastly more effective than doing flash cards. Like for example, just quizzing yourself. The main advantage is how easy it is to do active recall with the spaced repetition algorithm on the go. Like five minutes here and there when you’re sitting on the toilet smashing out a few flash cards. That’s the main benefit. And that benefit is not available using Notebook LM. Now, maybe I just didn’t know how to use it properly. At least it wasn’t obvious how I could take the flash cards that are generated and then export them out like as a CSV or an Excel file or something that I could then reimport into my flashcard app of choice. But I definitely felt like the flash card function on Notebook LM is essentially redundant. I feel like I would probably never need to use that button if I’m actually using this for long-term learning.

The Scorecard and Deeper Issues

So let’s go to the scorecard. See how it’s performing. So as a beginner learner or an advanced learner, ease of use is very, very good. It’s a great interface. It’s a polished product. Everything works the way that it’s kind of meant to work. The multimodal features like the video overview and the audio overview are surprisingly pretty good, especially given that it’s a relatively new product. In terms of accuracy, it’s actually hard to comment on this as a beginner learner because as a beginner, I don’t know if it is right or wrong. So, I’m going to defer my judgment on this. But when I used it on self-regulated higher order learning and the impact of AI which is something that I’m very familiar with, I couldn’t find any technical issues. Actually I couldn’t even find a single factual error in there. So I say the accuracy on this is also very high.

Now in terms of effectiveness, this is where it gets questionable. When I was testing this as a beginner learner using fundamentals of AI, I was looking at the mind map and I was looking at the video overview and I was looking at the audio view and interacting with that and I was thinking, okay, like it feels really helpful. This feels really cool. There are certain things that have been simpler and easier for me. But that feeling of being overwhelmed by all this information and not really seeing how it connects together never really disappeared. And actually, when I was expanding through the mind map, even though it was kind of clear how it all comes together, because it’s literally telling me this is how it’s arranged, as I just kept expanding through it, I just started getting overwhelmed by the mind map as well. The purpose of the mind map is to help you see how it connects together at the big picture level. And I didn’t feel like it was achieving that.

But the issues became much, much clearer when I started moving into a domain that I was very familiar with. So when I looked at the mind map that it generated around self-regulated learning, the categorizations that it created were technically logical, but they’re also not the most meaningful. They’re definitely not the way that I would have organized this. I would probably never teach this topic in that way. And I think trying to learn the topic in that structure, connecting it in that way, is going to make it harder than it needs to be. And it’s not really intuitive to understand.

The video overview I thought was actually a pretty good introduction, but the audio overview and the podcast format I started feeling was less and less valuable as I interacted with it more. One part, and this is the minor one, is that it has this two-person sort of co-host dynamic in the podcast which makes it sound quite engaging but often it gets quite distracting like one person will say one sentence and then the other person will sort of almost finish that point off. It kind of felt like I was talking to a set of cybernetic twins that telepathically communicate with each other and one twin is finishing the other twin’s thoughts all the time. This felt a little strange and it was strange enough that it was actually distracting for me from the actual point that they were trying to make.

But that’s probably the bigger issue which is the second thing is that I don’t think it does a good job of emphasizing the things that are actually important or complicated. There are certain things that I know in the research get really deep and complicated and is a really important concept that you have to understand. If I’m teaching this and I’ve taught that many times before, you have to spend a decent amount of time on that point. If you don’t understand this point, you’re going to be confused about everything else after that. But in the podcast format, it’s literally like one sentence and they move on from that straight away. Whereas other times, there’ll be a concept which is kind of interesting. It’s easy to explain. It’s easy to create these examples and analogies out of, but it’s actually conceptually not that important. And then they will spend longer on that part. And so the emphasis of which points to focus on, which parts to draw out and elaborate, which parts to really create an analogy and pace the learner through, I think was done fairly poorly. Poorly enough that if you are not concentrating on everything that’s being said and really thinking about it deeply and you let’s say for 5 seconds kind of just switch off and you don’t catch what was being said, that 5 seconds could have been something that was critical and you’re not even going to realize why you’re confused about this topic 10 minutes later.

Now, the good news is I believe based on my understanding of the underlying technology that this is probably a current state issue and that this will get better over time. I don’t see a reason why it wouldn’t be able to get better at doing this. And I think if this part is refined, that podcast audio overview format could be a really powerful tool that one day I have a sneaky feeling I’m going to start recommending that for all of my students. But as it is right now, unfortunately, I would probably give it maybe a three out of 10 in terms of actual usefulness and effectiveness at developing my understanding of a topic. And I think this is actually probably even worse if you’re a beginner trying to learn something because as a beginner, you have absolutely no idea that something important was skipped over. So, you don’t even realize you should be interjecting and asking a question and asking it to elaborate on it because the point would just pass by. It would just be one other sentence that you didn’t fully understand and then the confusion that that creates is just going to add to your overall sense of overwhelm.

The Biggest Problem with AI Learning Tools

But here’s the biggest issue with Notebook LM, and this is not just Notebook LM. Actually, I haven’t found an AI tool that is able to really overcome this or even come close to overcoming this so far, which is that Notebook LM and a lot of these AI tools are great at solving small problems that don’t make much difference. And it’s not very good at solving the biggest problem that makes the biggest difference to your actual ability to learn. In fact, I’m going to make that an even stronger statement. I think it makes it worse.

Now, let me be clear about this. I loved the concept of Notebook LM. I was blown away by how it works and the video overviews and the audio overviews. I thought that was really cool. It’s heading in the right direction, but I objectively just cannot put my stamp of approval on this yet. And this actually comes down to the win criteria that I was talking about before. The hardest part about learning and learning to a high level and trying to develop expertise and really work with knowledge in complex ways is organizing and connecting it. It’s this part like the connecting information, being able to apply it contextually. Like I said, when there are lots of things happening, they all connect together and you don’t know how they connect. That creates overwhelm.

In learning science, we call that phenomenon multiple element interactivity. And in my experience, it’s the most common reason why people feel overwhelmed when they’re learning something. The reason they think something is complex or that it will take time is because they recognize multiple element interactivity. In other words, there are a lot of elements and components and bits and pieces of information that are floating around. They interact with each other and it’s getting complicated. And the reason that happens is biological. The human brain can only process and connect a certain amount of information at one time. Honestly, in my experience of coaching people and in my own reflection, it’s really hard to mentally keep track of more than three or four connections in one go.

Now, compare that to actually learning something that’s complicated, which might involve 50 different items with a 100 or 200 different connections. Going from feeling good about a topic to feeling overwhelmed, might only take you a single paragraph of reading. And the hallmark of a really good learner is how they navigate through that feeling. When they’re hit in the face with that overwhelm, what do they do about it?

Some people will try to take a shortcut by changing the win criteria. Instead of saying, “Okay, this is hard. I’m going to find a way through it.” They say, “This is hard, so I’m just not even going to try to do that. Instead, I’m going to make my win criteria the fact that I’ve memorized it.” And they’ll use rote learning strategies. On one hand the good thing is that well you can achieve certain level of memory doing that but it’s often very time-consuming, very tedious, very boring and again it’s actually not effective for achieving your goal if you need to use that information in a complex way.

So when you think about the process of learning there are lots of parts that can make learning really complicated. There are lots of resources that you need to examine through and you need to filter through and screen and collect from. That process can take sometimes hours, days, even weeks. AI is great at solving that problem. That’s awesome. That is a genuine high yield time save that AI is providing that is valuable and effective for learning because you’re not getting much learning benefit from just the act of collecting resources.

Generating multimodal forms of learning, visuals and diagrams and audio and video. That is something that AI is, as we can see, it’s starting to be able to do that. And again, for a relatively new product, it’s doing it pretty well. Yes, it’s not quite there yet, but I can see it getting there. And when it does get there, that’s going to be a huge benefit for your learning because again, looking for videos, looking for images that are that are personalized to what you need to know. That’s not providing benefit.

But the hard part is when you have all of that information and now it’s entering into your brain and you have to do something with that information. You have to take this information and you have to organize it in a way that makes sense for you. And the key issue that people don’t understand often is that it is the process of trying to organize it that creates the learning. It is not the end product that matters. Trying to see how it connects together and challenging your brain to think through these different pathways causes those pathways to be reinforced which translates into understanding and mastery.

And so you can take a mind map or a video that was created by the world’s leading expert on this topic and they could say this is the way that it makes sense for me and they could give you their mind map and you can receive that mind map and it will do no benefit for you because you didn’t do any of the thinking involved in generating that mind map. And likewise you can have two people learning the exact same topic and they could organize the information differently and it could both be accurate. It just makes sense for each person slightly differently. A lot of the time if you were to look through some of the notes that I’ve written in the past for like different things that I’ve studied, you wouldn’t really be able to make sense of it. But when I look at it, it makes sense for me because when I look at it, it reminds me of the process that I used to come to those conclusions. I remember drawing the line from here to here, moving this point from here to over here, regrouping and recategorizing the information in this way because I felt it made more sense than the first way that I grouped it. I remember all those decisions and those challenges that I mentally overcame which leads to the learning.

And the issue is that what the AI tool sort of promises in a way is that you can click a button and it will just do that for you. It generates the outcome for you. And when you have the outcome just given to you without the process, it’s just adding something else to be overwhelmed by. And now, not only do you not see how it all connects together, you also are trying to fit your understanding around this mind map that it’s given you because that’s the way you’re meant to understand it. And that’s the part that I mean by I think it’s actually making it worse.

When I was going through this material, especially when I was approaching it as if I was a beginner, I was deliberately trying to hold back on some of the learning strategies that I would normally use to work through that overwhelm. And I was letting the AI guide me in my thinking. And I felt the experience to be incredibly inefficient and very confusing. Actually, much more confusing than if I was just reading it like in a book. And the reason is that I was holding off from engaging in a more organic process of trying to organize and make sense and connect the information together. Instead of having the sea of overwhelm and just diving into it and just trying to swim through and make sense of it, I was holding off on that and I was trying to find an easier way. I was trying to find a shortcut.

So, I was using their audio overview and their video overview and their mind map that they generated and I was trying to make myself fit this knowledge into the frames that they had given me and it wasn’t clicking. And when I compare that to the advanced test that I did with knowledge that I’m really familiar with, it makes sense why it wasn’t clicking because the way that it is categorized, I just don’t think is inherently that meaningful. But when I was a beginner going through AI fundamentals, I didn’t realize that it wasn’t meaningful. When I looked at it, it seemed logical. It seemed like it should make sense. And so after spending, I don’t know, maybe 30 or 40 minutes just trying to understand the information in the frame and the context it’s trying to give me, I just gave up because I didn’t feel that my overwhelm was improving at all. And I felt myself leaning into the tendency of just trying to memorize this instead because I wasn’t making any progress.

And that’s interesting because normally I would have spent that 30 to 40 minutes just diving into the overwhelm and making sense of it. And I probably wouldn’t have been able to get through everything in 30 minutes, but I would have had 30 minutes more of organizing and trying to make sense of it and like actually useful, meaningful learning as opposed to what ended up happening, which is 30 minutes of trying to make it fit and just realizing I don’t think this is working.

Final Thoughts and Recommendations

So right now, if I think about my position about where AI tools are headed and how useful I think it’s going to be for learning, I’m kind of on the fence. I really want AI to be effective for learning. I want it to be the big enabler that allows people that struggle with their critical thinking and higher order of thinking and have typically not been great learners to be able to be great learners because I do this for a living. I teach people those skills and I see that moment when it suddenly clicks and they’re able to think and operate at a level that they previously weren’t able to operate at. And you can see it just changes the way that they think about their opportunities in life and what they’re capable of. Like these are real tangible benefits that I see in people and I think that’s an amazing transformation and that’s the reason why I left my job as a medical doctor to do that because I found a lot of fulfillment in creating that kind of transformation and if AI can do that at scale that’s a huge deal.

But as I experience more and more and as I have conversations with dozens of the students and the professionals that I’m working with trying to use AI while also learning the techniques that I teach them, what I’m seeing more and more the trend of is that while it is saving time in some parts, it seems to be distracting in other ways. It is promising a shortcut to mastery that it can’t deliver on. And a lot of beginners don’t realize that it’s not delivering on that because they’re not measuring the effectiveness of the tool they’re using based on this criteria. They’re measuring the effectiveness of it based on how long does it take for me to get through this content. They’re measuring it on how many hours am I saving without taking into consideration the outcome they get to at the end of those hours. Like if you need to go from level one to level 10 and it normally would take you 50 hours to do that, but with using AI now, you can do, you know, level one to level three in just 2 hours. Well, the first option is actually still more efficient. It’s still better to spend 50 hours going from level one to level 10 if that’s where you need to get to because that’s the only path that actually gets you to the level you need to go. It doesn’t matter how fast you’re going if you’re not getting to where you need to go.

So overall, I’m gonna rate the effectiveness of Notebook LM a pretty solid meh. I think if you’re trying to reach a lowish level of understanding, then I think that’s fine. If you’re trying to reach a relatively low level of understanding, then the effectiveness is going to be good. And most of that is because if you were normally trying to do that, most of the time that you’re wasting is in collecting those resources and bringing it together and maybe finding the right video. And because that time is genuinely really saved by AI, that’s going to be really effective for that.

But in mastery level expert learning where most of the time is not in the collecting the resources and most of the time is just trying to make sense of overwhelming information, I don’t think this is really moving the needle by much. And there is a real risk that it will make things harder and worse while giving you the illusion that you are learning for a lot of people especially if you do not have training and learning skills.

However, there is one important caveat to this which is that the effectiveness scales with your skill. I think if you are a skilled learner, you can use Notebook LM to learn effectively. It’s going to be no worse than whatever other method that you’re normally using. You should be able to improve your current efficiency by using it as long as you’re aware of the traps and you’re able to navigate through that and use your own learning skills to interface with it. And likewise, if you are not very skilled with learning and you don’t have a lot of those skills and those habits and that knowledge about how learning works, then when you use it, you’re most likely going to depend on the tool a lot more to guide you. So even though you’re more likely to fall into those traps that I mentioned, you’re probably not going to do any worse than whatever you’re already doing because in your normal learning, you’re probably falling into different traps anyway.

And so even though I think using this as an alternative to getting better at learning is a big no, that’s not going to be effective and it’s not going to be a winning game for you, if you just compare yourself not using Notebook LM to using Notebook LM, I think you’re more likely to get an advantage just by using it. And if you’re using it and also actively trying to improve your learning skills, I think you’re going to get an even bigger improvement.

Now, really quickly, let’s go through the use cases. How good is it for an intensive study session versus on-the-go learning versus task reactive learning? I actually think that for all categories at all levels, it’s basically equally as good. You can use this tool for when you’re sitting down long hours. You can use this tool when it’s already preloaded with things on the go. If you’re starting a new notebook and you’re having to import new resources and kind of configure things for the first time, you’re not really, I think, going to be able to do that in just like 5 or 10 minutes. But if you already have an existing notebook you’ve been learning from and then you’re trying to reference that during the day, I think it’s a really quick and effective way to do that. The mobile app or the mobile responsive version of the website works great and especially that audio overview podcast feature, especially when it gets better, I think it’s going to be a game changer.

For task reactive learning where you have to have a certain body of information that you’re just trying to problem solve through, I’m imagining like a developer learning a new library of new documentation to go through. Get all that documentation loaded up into sources here and then just use that to guide you on your task completion. I think that’s going to be a really effective tool that you can use.

And so considering all of these things, using Notebook LM versus ChatGPT or ChatGPT study mode, Notebook LM, clear winner. At the current state, I see absolutely no reason why you would ever use ChatGPT study mode compared to Notebook LM.

And so if you’re interested in using Notebook LM, here are a couple of final tips that I would give for you to make sure you get the most out of it.

The first thing, and this is not super obvious if you’re just kind of using the interface and just getting into it, is if you are using Notebook LM to learn information like you actually need to learn the stuff and you’re not doing it just purely for task reactive, you know just like asking questions and getting answers straight away, go into the settings of the chat and tick the option for learning guide. By default, it is not turned on. What taking it into learning guide will do is whenever you have a conversation with it, instead of it just dumping the answer for you, it will ask you questions and force you to generate some responses.

And here’s a really interesting thing that I think you guys should all be really wary of when you’re using AI tools. It’s not enabled by default. But here’s the thing. I promise you when you enable learning guide your actual learning, your ability to achieve the outcomes that you wanted is going to be higher. So why if it produces a better result would it be turned off by default? That’s because using this needs to feel good for you the consumer. This is a product. So if you use it and you feel like it doesn’t feel good, you’re not going to use it anymore. And the product is not going to be a success. And when it comes to learning, this becomes a really problematic point because research on learning, especially around the idea of desirable difficulty or the misinterpreted effort hypothesis, shows that learners often feel that something is not effective if it feels harder. Whereas the truth is effective learning is effortful. It involves you having to think about it. And what happens when you turn learning guide on is that you have to engage with the chat. You actually have to think about what you’re learning. It will ask you, hey, what’s your current understanding? It’ll ask you a series of questions to test you instead of just giving you the answer. For a lot of people, because that involves more effort, they’re going to say, I can’t be bothered using this. This is not good because I want the fastest, shortest solution possible. And so that’s probably the reason why by default it is turned off. So my recommendation for you is turn it on and realize that when you need to spend more effort to answer questions and engage with it, that’s improving your learning. In learning science, this is promoting something called the generative effect or generation.

My second tip is to earn the response. Pay for a response with a question or a curiosity or a gap in your knowledge that you’ve identified. A thing that you have realized you don’t understand or it doesn’t make sense to you. Target that and get the answer to that. Do not treat this like a lecture. Don’t engage with it passively. Don’t sit there, get it to generate a summary, click a few buttons, read stuff, click the next button, read the next stuff, use the recommended question that it has at the bottom, click that and just read what it is. Don’t be a passenger in your learning. Being a passenger in your learning is a very fast way to get nowhere.

Take control. Use a summary. But then stop. Pause. Don’t look at the screen. Think about it. Think about where your knowledge is. Think about what makes sense for you, where your gaps are. Be specific about it. Feel where you are less confident. And then turn that feeling into words. What do you feel less confident about? What about the knowledge and the information do you not see connecting? Turn that into a question and a prompt. Get the answer again. Pause. Think about it. Consolidate. Spend more of your time thinking about the knowledge and building expertise and mastery and less time just consuming and being passive. Again, involves more work, involves more thinking, but it’s vastly more effective.

And the final tip, don’t expect AI to be your learning savior. For now, in the foreseeable future, from my perspective, neither Google or Open AI or Microsoft or whatever other company is going to come down from the sky and deliver unto you the AI tool that fixes all of your learning issues. That wall where you feel like it’s getting complicated and you’re feeling overwhelmed, that is the wall of learning that you’re looking at. And I don’t think AI is going to bore a tunnel through that wall for you anytime soon. Your ability to learn effectively shouldn’t depend on the AI tool that you’re using. It’s not about the tool. It’s about you the learner and then how you interact with different tools. That is where the bottleneck is going to be. The more you can change, the better your outcomes are going to be.

And the most ironic part about me telling you this conclusion is that when I actually used the self-regulated learning and the impact of AI on learning as the topic for the advanced learner testing, this is actually the conclusion that it generated for me in the video overview. An AI learning tool telling me that the issue is not with AI learning tools. As I’m using an AI learning tool to learn about AI learning tools, I think is like a special kind of poetic irony.

So, let me know what you think about this below. What’s your experience been with Notebook LM? And are there any other AI tools that you want me to review? If you’re looking to improve your learning skills and get a little bit deeper into this and improve yourself, the thing that makes the biggest difference, then you might want to check out some resources on learning to learn. Hope this helped. Thank you so much for reading and I’ll see you in the next one.