The NotebookLM: An AI Love Story of Learning with no Ryan Gosling

Read time:
5 minutes

Gary Davies

I wrote you 365 letters. I wrote you every day for a year.

It’s the only Notebook quote I’m going to use in this blog and you can bet Allie would be signing up for a NotebookLM trial after reading this blog – what better tool to provide an overview of all that text.

This tale of love is going to be focused on my love of learning and AI, and the journey I’ve been on over the last few months with NotebookLM, an incredible AI powered tool from Google. It allows you to provide a series of data sources pertaining to a particular topic and then allows you to generate summaries, guides and notes while providing an interface to quiz on the content.

If that wasn’t impressive enough it added the power of creating podcasts, then being able to write prompts for them before the most recent release (in beta) which allows you to converse as part of the podcast.

If you haven’t come across NotebookLM yet, you will definitely have heard the voices from the podcasts as their popularity has risen across social media, and I hope that after reading this you will find the tool as useful as I have.

Starting Simple

NotebookLM allows a variety of data sources to be used including URLs, files, text and even youtube videos. In the gif alongside I have taken the opportunity to feed it the last technical blog from Data Intellect by our CTO Jonny Press and ask some questions. As you can see from this NotebookLM provided some prompt questions which I've used in addition to picking one of my own.

A Quick Learner

If we then navigate to the right hand panel which is out of view from the previous gif you will see the Studio and Notes section. Now don't get too excited just yet, I'm leaving you in suspense for now as we skip past the "Generate Audio" and focus on the Notes section.

From here I can quickly generate a study guide with sample questions, answers and even a glossary of terms. I can create an FAQ or a briefing doc. We'll come back to the timeline a little later.

Ramping up, a Data Intellect use case

That gives a little taster into the potential and I’m sure everyone is thinking what funky things can you throw at it…well I had the same and did the “analyse these song lyrics” and “summarise this text I can’t be bothered reading but need to know”..yes it can be fun but I also have to remember I’m working and there is a purpose to playing with AI tools.

For me it is about identifying opportunities to leverage AI and AI solutions to bring about efficiencies in our business at Data Intellect and therefore I started to have a thoughts on how could we use this to help our people. Also when it comes to NotebookLM it needs to be something that I can’t just ask any LLM on the market and thats leveraging something in or about Data Intellect that isn’t out there for consumption.

This sparked some trains of thought:

Could we use this to help our folks get answers to Data Intellect questions? Be it subject matter, company related or more general
Could we use this to help folks learn kdb+?

And from this I went off trying a variety of things and putting the tool to the test, before settling on “an AI assistant to the people team”.

PAT, Assistant to the People Team

Inside Data Intellect we have generated substantial documentation to assist our people and their development as we look to realise the potential within. These documents include:

Consulting Progression Framework: A guide to all the roles in the company and the skills needed to progress your career into those roles
Consulting Skills Guide: A Data Intellect guide to the soft skills around consultancy, living our values and our vision
Data Intellect People Leader Guides: A two part guide into the how and why of being a people leader
Policy Handbook: The guide every company needs but people only usually read it when they need it
Our onboarding workbook

These documents presented a good grounding for the NotebookLM but given it also took URLs I fed it our website too give it a little more background on the company.

I spent 10 minutes setting it up and feeding the docs/urls into it (including finding the docs) and I immediately prompted “I’ve just joined Data Intellect in New York as a grad, what do I need to know” – the response shared the key points on company value, vision and the key overview of the handbook, but it was the last bit that got me – You should probably reach out to Joyce Tiu as she heads up North America. My immediate reaction was I didn’t think we’d that in the notes and had to click on the reference, yes that’s right it provides references and links to where it got the information to make statements, and it turns out it took that from the website. So straight away I knew it had used the website, the progression framework and the onboarding workbook.

This resulted in me excitedly doing demos around the company to show this tool and it became my go to place of questions – in some cases I shared output first to showcase its power. During this time Google also launched the Podcast functionality, and then the ability to tweak it via prompts, both of which produced an excellent insight into the life at Data Intellect.

And so PAT, the People-Leader Assistant Tool was born – and its currently in Beta testing at Data Intellect.

Below shows a screenshot of PAT, note everything in the chat you see was auto-generated from NotebookLM, I have not tweaked it.

But Gary, what about your avid kdb+ readers

Of course I also added kdb+ content in it, using a combo of internal and publicly available information, an example of which you can see alongside (Using only public information, keeping the secret sauce for when you join us).

In this case its useful for a study guide and its great for getting quick information out, as a partnering guide, but it won't assist with code generation or reviews.

So, what next?

NotebookLM has become a key part of my work ecosystem, its a tool I use regularly and I intend for PAT to go beyond beta and be released to our people team for usage.

Beyond that, we have a number of people internally who are sandboxing out ideas, one particularly excites me….Learning & Development, why? because we have lots of internal training across our performance software engineering, data and time series streams and this presents a new medium with the study guide and podcasting capabilities.

Where's my audio

I quickly skipped past the "Generate Audio" earlier but I think it is worth showcasing it's power. As such I have in fact fed this blog to NotebookLM and had it generate a short podcast.

Check out below

Listen in