Today’s episode topic is on Google’s newly-released (and frankly sensational) product NotebookLM. All you need is a Google login, which is as easy as having a Gmail account. Use of NotebookLM is likewise totally free.
The user interface for NotebookLM is simple and intuitive. You provide it with resources — such as documents you upload or links to websites. You can provide up to 50 resources, which is a pretty vast amount. It’s not obvious to me that there’s a file-size limits for the resources you upload, but inspired by Dr. Brett Tully (guest in #533) who messaged me a NotebookLM-generated podcast conversation about his doctoral dissertation, I similarly uploaded my 12MB, 200-page doctoral dissertation into NotebookLM. (I wanted to upload my book Deep Learning Illustrated, but the manuscript PDF I had immediately to hand is 236 MB and it turns out NotebookLM has a limit of 200MB per document, oh well.)
NotebookLM allows you to converse with whatever resources you upload; it even suggests starter questions, based on the specific resources you upload, to get your chatbot-style conversation with the resources going. It also is pre-configured to automatically generate helpful documents from all the resources you provide like:
Table of Contents
Timeline of Events
Briefing Doc
Study Guide
FAQ
So, NotebookLM is handy for all kinds of summarization or interactive conversation that might be invaluable for understanding or analyzing vast amounts of information. The most impressive aspect, however — the aspect of NotebookLM that has been generating all of the buzz — is the podcast episodes it can create from your resources. Within a few minutes, NotebookLM created an 11-minute podcast episode about my 200-page doctoral dissertation and 99.9% of it is indistinguishable from a real podcast. On top of that, the podcast conversation between two hosts is extremely compelling. I think my PhD is super super dry… but listening to this conversation, the NotebookLM “podcast hosts” manage to make it seem interesting and practical. The “hosts” also cleverly use analogies (that I never had in my thesis itself!) to make complex concepts easier to understand. The hype is no joke and, to prove it to you, here’s the entirety of the 11-minute podcast episode NotebookLM generated based on my PhD. Note that NotebookLM didn’t output a transcript, it actually rendered this audio with just a few minutes of processing.
I’d love to hear your thoughts on the NotebookLM-generated content… did it blow your mind too? Or were there aspects that could improve?
The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.