Google recently upgraded its note-taking AI assistant to Gemini 1.5 Pro, expanded its availability to over 200 countries and territories, and introduced several new features. NotebookLM now supports web URLs and Google Slides for input to generate answers to related queries. Answers will also contain in-line citations that link directly to the relevant passages in the source material. Also, the Notebook guide is now capable of generating high-level summaries, overviews and FAQs.
Bringing NotebookLM to the most powerful version of Gemini means users can now take advantage of the multimodal capabilities of Gemini 1.5 Pro to upload images which can be analyzed and cited. This feature works as well as expected but since NotebookLM is a ‘closed system', it will not perform web searches or derive any information outside the aggregated sources. NotebookLM now allows up to fifty sources – up from only five previously. However, each source is limited to 500,000 words so the AI tool can only crunch 25 million words per notebook.
Google has shared a series of case studies on its blog highlighting several innovative uses of NotebookLM including biographer WalterIsaacson's analysis of Marie Curie's journals as well as podcasters and documentary researchers. 'Source grounding' minimizes the risk of hallucination that AI assistants are often prone to, but does not completely eliminate it. Still, larger issues are at play here. Privacy concerns remain since NotebookLM will require users to willingly upload material for it to be of any help unless Google finds a way to tether it to Gemini Nano – the lighter version of Gemini that runs on the Pixel 8 Pro.
While NotebookLM is still very much an experimental product, it has garnered the support of a diverse 14000-member community consisting of writers, publishers, students and role-playing game enthusiasts. And as Google commences the global rollout of its note-taking AI agent, the community is set to grow even larger.