Google announced on Thursday the expansion of NotebookLM, its AI-driven note-taking assistant, to over 200 new countries, six months after its initial launch in the U.S. The platform, utilizing Google’s multimodal LLM Gemini 1.5 Pro, now includes new features and supports additional languages, enhancing its capability to generate summaries and answer questions based on user documents.
The newly supported countries for NotebookLM encompass Australia, Brazil, Canada, India, and the U.K., along with 208 other countries and territories. Google has increased the interface language support to 108 languages, such as Arabic, Assamese, Bengali, Cantonese, Chinese, Dutch, French, German, Hindi, and Hinglish. The platform also supports sources and chat in 38 languages, including Arabic, Bengali, Chinese (simplified and traditional), Dutch, French, German, Hindi, Japanese, and Spanish.
Originally introduced as Project Tailwind at Google I/O in 2023, NotebookLM debuted to a selective group in June last year. It employs AI to create summaries and respond to questions from documents, transcripts, notes, and other user-uploaded sources. Unlike traditional AI chatbots like ChatGPT, which may generate information not strictly related to the provided source, NotebookLM adheres closely to the user’s data.
Google has extended NotebookLM’s capabilities to source content from Google Slides and web URLs, in addition to the existing support for Google Docs, PDFs, and text files. This enables users to create notes or query the content within documents or online material, whether in image or text form.
Some early adopters in the U.S. had anticipated compatibility with traditional note-taking apps like Evernote and Google Keep. However, Raiza Martin, senior product manager for AI at Google Labs, explained in a virtual roundtable this week that the focus is currently on refining the product’s core value before introducing new integrations.
“Down the road, you’ll hopefully see these types of integrations,” she mentioned.
Google has also introduced inline citations to facilitate verifying AI-generated responses and reviewing original texts for better context. Previously, citations were positioned below the responses generated by the assistant.
Additionally, the platform includes Notebook Guide, which assists in converting content into various formats like FAQs, briefing documents, or study guides.
Steven Johnson, Google Labs’ editorial director, noted that NotebookLM was created with the input of authors, students, and educators, and early adopters have integrated its source-grounding features into their research and writing processes.
The platform has also been employed for creating hyperlocal newsletters, summarizing interview transcripts, preparing grant proposals, and managing descriptions of fantasy worlds.
Martin emphasized that Google does not use any data uploaded to NotebookLM for training its algorithms.
“We often get this question because users want to use it with work or school documents,” she said. “Your data stays private to you.”
During its Google I/O 2024 keynote in May, Google demonstrated an early prototype of Audio Overviews for NotebookLM, utilizing the Gemini model to scan uploaded materials and produce a podcast-like discussion. Gemini 1.5 Pro enables NotebookLM to accommodate up to 50 sources per notebook, with a capacity of 500,000 words per source.
The worldwide expansion of NotebookLM is set to position it against numerous platforms that provide similar GenAI tools for tasks like answering questions and summarizing PDFs. Many of these platforms come with a fee, but Google offers this service for free, leveraging its substantial resources.
