Hello NotebookLM: Google Docs Gets a Research Assistant

After teasing the AI application at its developer conference in May, Google is finally rolling out NotebookLM, an experimental language model designed to be a virtual research assistant.

Previously called Project Tailwind, NotebookLM can summarize facts, explain complex ideas and brainstorm — all based on the content that users selected. At its core is Google’s large language model PaLM 2.

There is waitlist to try it out, but only for U.S. users for now.

Compared to ChatGPT, NotebookLM can be ‘grounded’ by specific documents or notes – starting with select Google Docs files, with other formats coming later. Notably, NotebookLM only has access to the source material that users selected.

By being grounded, NotebookLM learns from the users’ content to create outlines, yield insights, and answer questions about the documents. For example, a science student can upload class notes and ask questions about them.

Image: Google

As it is grounded in a user’s content, NotebookLM reduces the model's ability to hallucinate – or generate incorrect facts or figures, according to Google. However, the team behind it stressed that users should still fact-check the AI’s responses against original source materials. Google said NotebookLM provides citations for its answers to make it easier to do so.