Google NotebookLM has been open in a tab on my computer for about a month. I’ve been meaning to play with it further, but I just haven’t had the time (and still don’t really have the time!). Then again, what is a Friday afternoon for? I can see a lot of potential (potential!) for academic work, especially given Nokebook LM’s potential to draw upon a more extensive range of sources than I’ve previously been able to with ChatGPT. This short blog captures my reflections in action.
As I head to Google NotebookLM, I am looking forward to giving this a whirl and I start by spooling up a new notebook. While there appears to be a 300-source limit, I purposefully started with just one: my thesis! I have several reasons for this, but it is primarily (1) something I own, and so there are no copyright issues and (2) it is something I really know. I’ll be able to judge accuracy without reading and re-reading this thing quickly. My thesis was a five-year (part-time) journey, and so I recall it very well.
The NotebookLM audio ‘deep dive’
I’m immediately drawn to the ‘Audio overview’ as soon as it’s uploaded. I can’t help mashing the button over here, and patiently waiting three or so minutes for this to generate. I have to admit – it wasn’t at all what I expected. As I hit play, I heard what I could only describe as a podcast episode about my thesis. As I’m writing, I’m about two-thirds of the way through this,s and so far, it is pretty good:
The ‘deep dive’ audio from NotebookLM (listen above!) shows a very conversational and dialogic approach to the overview. It makes it much more engaging than a monotone overview, and the fact that the two voices are in dialogue does make it a lot easier to listen to. I’m particularly impressed with the tone of these voices, and it’s almost impossible to distinguish this as an AI product. That’s kind of scary, especially for something I’ve only used for about 20 minutes!
As I finished listening to the audio, my first reflection was that it was highly correct. Nothing in the recording is incorrect or wrong, except for the constant reference to ‘researchers’ and research, as though drawing upon multiple articles and authors. (How annoying! My thesis was a lot of MY work, not that of others!). Being ‘correct’, however, doesn’t mean this deep dive is an accurate summary. While there was nothing incorrect here, I fear the ‘deep dive’ wasn’t always that deep. Some areas of the work were delved into in great detail—others, not so much. The overall structure was a bit odd, and there was an overwhelming focus on the role of Library staff. While this was an element of my thesis, I don’t think the ‘deep dive’ emphasis really reflects the full text that well. It also bypasses great swathes of research and methodology, focusing instead on the contribution and findings of the work. While okay for an overview, this lacked the situated context found in the document itself. This would certainly make me worry about ‘trusting’ it to overview something else I was less familiar with.
NotebookLM Notes
Underneath the ‘audio overview,’ I next see a range of buttons that can generate study guides, briefing documents, FAQs, and a timeline. I couldn’t really help mashing them all. Here are my reflections:
- The briefing document:
The briefing document provides a brief overview of the whole thesis section by section. I suspect this was mainly driven by headings and provided an alright overview. I don’t think this is a ‘briefing document’ by any means, and it wouldn’t make much sense without having known the original document. That said, it formed an alright set of notes, and even now, this works well as a reminder! - The study guide
This wasn’t quite what I expected. The document generated provided a series of short-form questions and answers, followed by some essay-format questions. This might be useful if I were to deepen my learning or prepare for an exam. That said, some of the things it focused on were a little bizarre! I’m not sure this works for the higher education context. That said, if I had uploaded all of my GCSE notes, I could imagine it useful for that learning and examination style! - The timeline
In fairness, this didn’t make much sense in the context of my document. I can’t be surprised that the output was a little zany – though it did exactly what it had set out to do. It seemed to look for chronological aspects of each chapter and draw them out. The funny bit was the ‘cast of characters and brief bios’. It drew out names from my acknowledgements and a random list of reference authors (mostly the theorists and geographers who had substantial careers and publications). - The FAQ
Like much of the other documents, this was all fairly accurate but just a little odd. Functional, yes, but I’m not sure any single one of the questions would really be ‘frequent’.
Final reflections
As my emails are calling, I’ll leave my experiment for the day. From what I have done so far, I can see a lot of potential for this as a ‘notebook’ – and who would have thought based on the name 😅. For a document that I know, this tool was able to rapidly help me explore it further, but also ‘jog my memory’ on bits of it. The reason I could trust this was because it was something I knew. It is clear NotebookLM couldn’t be trusted to really summarise work or build notes from it where you haven’t read the source itself. That said – where you know the source, it is a brilliant way to organise your thinking and engagement further.
I can’t wait to play with this some more, especially with further sources in play. Watch this space.
Recent Comments