This Google AI Tool Can Turn Your Research Into a 'Podcast'

By | 10:12 AM Leave a Comment

When it comes to Google's AI tools, you might think first about the Gemini assistant app, or maybe some of the cool tricks you can do in Google Photos. However, Google has another powerful artificial intelligence tool that's great for collecting and collating notes on any kind of topic. It's called NotebookLM.

The aim of NotebookLM (the LM stands for Language Model, of course) is to help you make sense of whatever it is you're researching. It'll lend a hand by creating summaries, answering questions about the documents you've gathered, and linking together data points where required. As is normal with this type of AI, it's still considered an experimental feature, and you need to double-check everything it produces for hallucinations.

That said, it's a potentially useful tool, and just got another new feature: Audio Overviews. Essentially, this feature can create short podcasts, with AI podcast hosts that summarize all the information you've brought together. Whether you're researching 19th century literature or which new phone to buy, the Audio Overview will present it in a friendly, accessible way—and the results can be pretty impressive.

You can access NotebookLM, and the Gemini models underpinning it, for free using your Google account. Give it a try with your own research notes, and be prepared to be amazed at how natural the resulting clips sound.

Getting started with NotebookLM

Google NotebookLM
Adding sources to NotebookLM. Credit: Lifehacker

If you're completely new to NotebookLM, you can sign in with your Google account to get started. You'll see some example projects (called notebooks) included for you to have a play around with. Open the Invention Of The Lightbulb one, for example, and try asking the AI questions about the work of Joseph Swan, Thomas Edison, and others.

The question-and-answer routine will be familiar if you've already used something like ChatGPT, Gemini, or Copilot—but here, the training data used for the model is only the sources you've uploaded to your notebook, rather than the wider web. You can point NotebookLM to web links though, as well as add plain text, PDFs, Google Docs, and Google Slides (each notebook can have up to 50 sources).

To start putting together your own notebook, click the NotebookLM heading (top left), then select New notebook. You'll be asked to start adding sources to your new notebook; you can upload documents from your computer, pick them out from Google Drive, point NotebookLM towards web links, and paste in text from the clipboard. You can add more sources later by clicking the + (plus) button next to Sources on the left-hand panel.

As soon as you start adding sources, NotebookLM will start summarizing them for you, then suggest questions you might want to ask. You can click individual sources to view them separately, and follow the onscreen links to put together a FAQ, a study guide, a table of contents, a timeline, or a briefing doc for everything you've collected.

Producing your Audio Overview

Google NotebookLM
Your podcast appears on the notebook overview. Credit: Lifehacker

For the purposes of this guide, I uploaded a PDF of s recent study on how smartphone apps could be developed to measure blood pressure, with no cuff required. I also added the accompanying press release as a web link, to give NotebookLM a bit more material to work with.

Right away the AI suggested question prompts around how the app utilizes existing smartphone technology, what the potential benefits might be, and what the challenges are in developing this tech—and on the whole, NotebookLM came up with responses that made sense and that were accurate based on the material provided. (As you use the app, the answers you get come with citations that take you back to the sources of that information.)

To produce your podcast (or Audio Overview), you need to open the Notebook guide (there'll be a link on the right if you're on the chat screen), then click Generate on the right. As of now, there are no settings you can play around with—you get the same two podcast hosts every time, one male voice and one female voice, talking in English. After a few minutes of thinking, the audio starts playing, and you can download the file too if needed.

Impressive, but not without some caveats

In my test, the Audio Overview did a fine job of summing up the blood pressure app research with a high level of accuracy, producing what sounds like a natural podcast conversation between two people—though it did miss some nuance, and was a little generic in places. As a tech demo and a summarization tool it's really impressive, but if details truly matter in your work, I wouldn't rely on AI entirely at this point.


from LifeHacker https://ift.tt/tuMDpwa

0 comments:

Post a Comment