TL;DR
- Google is releasing its research-oriented AI tool globally, expanding availability beyond the US.
- The app now uses Gemini 1.5 Pro under the hood, allowing up to 500 sources per project.
- NotebookLM remains completely free to use, but it’s an experimental project under Google’s AI Labs initiative.
Google is expanding its research-oriented AI tool, NotebookLM, to more than 200 countries and territories beyond the US. Besides wider availability, the search giant has announced that NotebookLM will now be powered by Gemini 1.5 Pro under the hood. The latter is notably Google’s current flagship language model, otherwise only available if you pay $20 per month for a Gemini Advanced subscription.
The latest release of NotebookLM also gains a few new convenient features, such as the option to import new external sources like web URLs and Google Slides presentations. Once imported into a project (or notebook), NotebookLM can reference these sources to answer questions via a chatbot-style interface. Beyond that, it can also create supporting content such as FAQs, study guides, and timelines of events. You can export these as a pinned note within the project or draft your own for future reference.
In my testing, NotebookLM was not able to fetch up-to-date information about any recent world events. This is unlike Google’s Gemini chatbot, which can search the web even if you don’t explicitly ask it to. Put simply, NotebookLM will only consult its internal knowledge base plus any provided source material on top of that and draw the line here. To quell fears of AI hallucinations, responses within NotebookLM will also contain inline citations similar to how Microsoft’s Copilot links back to sources.
On that note, NotebookLM now supports adding up to 50 sources per notebook, a significant jump from the previous limit of just five. Each source document can include 500,000 words of text, which helps explain why Google has tapped its highest-end language model.
When Google announced Gemini 1.5 Pro earlier this year, it claimed the model showcased “a breakthrough in long-context understanding.” More specifically, the company claims it can handle up to one million tokens, longer than competing language models like GPT-4o. NotebookLM takes advantage of Gemini 1.5 Pro’s multimodal capabilities, meaning it can answer questions about graphs and images in its sources.
NotebookLM is one of Google’s many AI experiments currently under active development, but it’s on the more polished end of the spectrum. Another project, dubbed AI Overviews, sparked controversy after offering dangerous advice ranging from eating rocks to adding non-toxic glue to a pizza. Google launched AI Overviews to select beta testers in the US under “Search Generative Experience” last year. If NotebookLM follows the same path, it may become a mainline Google product in the coming months.
