It’s experimental for now, but Google’s NotebookLM is not only designed to be an AI research assistant, you can even create podcasts from text!

Available here in a current experimental mode, NotebookLM is designed to be “your personalized AI research assistant powered by Google’s most capable model, Gemini 1.5 Pro”. The idea is that you upload the documents that are central to your projects and NotebookLM “instantly becomes an expert in the information that matters most to you”.

So, I decided to test it to see how it works. I went to the NotebookLM initial page and clicked on the link to “Try NotebookLM”.

NotebookLM gives you a few example notebooks to check out to get started to see what it can create – each with multiple sources. When you’re ready to create your own notebook, you just click the New Notebook square in the upper right corner.

You can then upload one or more sources from which to start building the notebook. Potential sources include: 1) files from Google Drive (e.g., Google Docs or Google Slides), 2) links to a website or a YouTube video, and 3) pasted text. It will also support file types including PDF, text and even mp3 audio files!

Once you begin uploading sources, NotebookLM will create a summary of what you’ve uploaded. It will then enable you to create: 1) FAQ comprised of questions about the content of the sources, 2) Study Guide with a list of questions and the answers to those questions at the end, 3) Table of Contents for the content that you uploaded, 4) Timeline of Events within the uploaded sources, and 5) Briefing Document examining key themes and insights from the uploaded sources.

Here’s what the FAQ looks like:

One of the most interesting things it enables you to do is to create podcasts from the text you uploaded. NotebookLM calls it a “deep dive conversation”, but the discussion sure sounds like a podcast to me! When you click on the “Generate” button represented in the screen shot above (be prepared, it takes a few minutes), it will generate an audio “discussion” between two “hosts” about the topic.

While the “discussion” is totally AI-generated, it sounds like real people having a conversation about the topic – they reference key points about the sources, attempt to analyze them and do so in a plain language, conversational style that (I assume) is designed to be at a level that the layperson can understand. The “hosts” even use colloquialisms, like “throw the baby out with the bath water”! It’s wild! 😮

I tried a couple of examples. The first was my post last week on the AI-generated case law order created by California Magistrate Judge Allison Goddard for our case law panel at Relativity Fest. In that instance, I uploaded the blog post in as text and Judge Goddard’s “order” via a link to the PDF. It generated about a 7 minute “discussion” of that content, which is available here (the “hosts” pronunciation of “Goddard” is priceless!). Here it is:

I also created a second example, using the “Blackberry lesson” post I wrote about a year ago, comparing the failure to embrace generative AI adoption to the fall of the Blackberry when the iPhone was introduced. I simply uploaded the link to the blog post. This time, it generated a whopping 14-minute discussion of the topic! Here it is:

While the discussion for this topic was quite a bit more lengthy, that wasn’t always a good thing. While loading the link to the blog post caused the AI algorithm to credit me for the post (and mention my full name five times), it stated some things that I didn’t say in the post, like that I was specifically advocating for “explainable AI”, which I never mentioned.

When I went back and re-did the Judge Goddard exercise (this time, with links to both the blog post and her order), the generated video was 16+ minutes! So, it’s either getting longer as a matter of course, or it’s longer when you link to sources as opposed to pasting in text from those sources. Will need to test more.

Regardless, NotebookLM is very interesting in terms of its ability to create useful materials based on uploaded sources, and its ability to create podcasts from text! Those “podcasts” may be one of the more interesting content it generates! While the quality isn’t there to replace actual podcasters (yet), who knows what will happen when the quality improves? The possibilities are endless!

So, what do you think? Are you concerned about NotebookLM’s capabilities to create podcasts from text? Please share any comments you might have or if you’d like to know more about a particular topic.

Image created using GPT-4o’s Image Creator Powered by DALL-E, using the term “two robot podcasters”.

Disclaimer: The views represented herein are exclusively the views of the authors and speakers themselves, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.