Technology giant Google is behind numerous tools, many of which do not reach a particularly broad audience. This was the case with Notebook LM, a service for keeping track of notes and documents, which allows users to ask questions about the content via a chatbot.
It's a tool that lets you understand your sources, says Raiza Martin at Google Labs.
"Becomes humble"
But when Google released an audio function a few weeks ago, the service went viral. Notebook LM can generate an audio file from almost any document, consisting of two podcast hosts discussing the content of the material. Users have tested creating podcast episodes from everything from their receipts to complex reports. Whatever is input, the tool tries to create an interesting conversation. Raiza Martin says she is surprised by the impact, even though internally, they also think the function is cool.
But you become humble when you realize that many others also think it's very cool, she says.
The podcast hosts pause, change tone, and comment on the content in a human-like way. Exactly how it works is a secret.
But a lot of thought has been put into what type of audio we think is the most interesting. And that's how we decided on this conversation style, with two hosts giving an overview of your content and trying to make it as interesting as possible, she says.
Considering the realistic result, it's a surprisingly small team working on Notebook LM. The work began with Martin and a few others at the end of 2022. By May last year, there were three engineers, and the group grew to eight before the audio function was released.
Continued development
Now the group has grown, and work is underway to add more languages. It will also be possible to control the podcast more, and within a few months, users will be able to ask their own questions in real-time. Google states that the uploaded material is not used to train the AI model and that the files are protected.
But is there a risk that the service will replace real podcasters? Raiza Martin says that the development is being closely monitored. So far, it's mostly users uploading their CVs – or creating podcasts about their website.
It's about people creating audio from things that wouldn't have been audio otherwise, she says.
Notebook LM is a service where users can upload up to 50 files. Additionally, users can get summaries and ask questions about the documents, using Google's AI model Gemini 1.5.
So far, audio files, YouTube links, PDFs, text files, Docs, and Slides are supported.
The service is free, and Google states that they do not train the AI model on the uploaded material.
Notebook LM has a Discord channel with 60,000 users, where all members of the development team also participate.
