Update makes Google's NotebookLM significantly more powerful and enables role-playing games

Whether it's the never-ending PowerPoint presentation from the last meeting, the hundred-page "summary" of the key points from a university lecture, or the transcript of the last court hearing, anyone wanting to search for specific information in such documents often had to invest a lot of time. Until now. Google's AI assistant, NotebookLM, eliminates many of these tedious tasks: a single command is all it takes, and the AI ​​searches through a wide variety of previously uploaded documents, often delivering results in just a few seconds.

Unlike Google's other AI assistants, NotebookLM does not focus on searching for data on the internet, but on analyzing uploaded documents. The AI is also designed to facilitate learning, for example, via audio summaries, flashcards and quizzes.

Update brings latest Gemini models to NotebookLM

Following the recent integration of Google's latest AI image creation tool, "Nano Banana", into the program, further changes have been made to the backend. Thanks to new Gemini models, the chat function has been significantly improved.

On the one hand, much longer inputs are now possible (up to one million tokens, equivalent to an eight times larger context window than before), and on the other hand, extensive, multi-level conversations are possible without NotebookLM losing context. Another practical feature is that the conversation history will now be saved, allowing conversations to be continued at a later time. However, chats between shared notebooks will remain private.

Another new feature is the way the system processes inputs. Instead of simply attempting to answer the user's question and thus potentially losing important context, the AI now searches the sources from multiple perspectives. This should enable much more nuanced answers.