Body
Notebook LM is an experimental AI powered notebook from Google that is available through Google Workplace. It is powered by Gemini 1.5 pro. Google describes it as a virtual research assistant that can process PDFs, Google Docs, websites, and Google slides. It is able to turn complex documents and slide decks into engaging podcasts. You can select the voices, male or female, and the ages of the individuals on the podcast as they discuss the document or slide deck.
Notebook LM only generates responses from the data entered. It does not use the content for further AI training. It is available to anyone with a personal Google account. Only the user can share the data. Notebook LM can not.
Notebook LM does use a pre-trained model with data, Gemini Pro 1.5 however, it is only for its processing power to read, analyze, and process large amounts of user-entered text and to turn that text into analyses, summaries, test questions, test banks, PowerPoint presentations, podcasts, and other forms of output. As one individual notes,
"It's important to note that Notebook LM isn't like ChatGPT where the application is going to dip into a massive database of information in order to try to answer any question. It's far more focused and will only pull from the sources that I input into it. In other words, if I ask Notebook LM why the sky is blue, it's not going to know the answer unless I've provided it with specific articles or documents that explain that very thing." https://leonfurze.com/2024/09/16/hands-on-with-google-notebooklm/comment-page-1/
Notebook LM will only use the content you input to produce the output you desire without any model training.
Notebook LM does not share or store the sources you enter with anyone. It does allow you to share your notebook(s) with others, making it easy to collaborate on projects. Notebook LM is "source grounded" meaning that it can only draw on authoritative, private, and trusted information that the user has entered. It will provide citations for all of the information it generates so the user can easily verify the accuracy of the information. https://www.pocket-lint.com/notebooklm-with-gemini-1-5-pro/
Here's a brief article titled Thoughts on Google Notebook LM from an Instructional Designer and Learner October 19, 2024.
Recent faculty discussion focused on how faculty could use it to create quizzes and test questions from their uploaded class notes and slide decks. They also discussed how it is multimodal and can create customizable podcasts of their notes and slide decks that could be shared on Canvas. So mainly focusing on usage by faculty as an assistant in saving time yet still creating good quality teaching and assessment tools. From the notes, Joe Faith mentioned that "you can also chat with an LLM about the documents themselves, getting very specific context in a text environment."
Sources show how researchers can upload scholarly articles and then ask Notebook LM a question or questions, for example: "List all statements that discuss treatments for heart disease and cite them in APA format." Basically cherry-picking exactly what someone is looking for without having read any of the articles.
Someone who is looking to "cheat" can do so with almost any form of generative AI, however, Joe noted that sometimes the amount of time and effort to cheat with AI outweighs the perceived benefit of doing so. Another said that if a student has to know how to do this, that, and jump through hoops to cheat, they will see that it's faster to just do the assignment.
Comparison Notes about Google Gemini:
Gemini Pro 1.5 LLM: There are four flavors of Gemini: Pro, the best performing model which comes with Gemini Advanced for the $20 a month fee. Flash, the smallest and most cost-efficient multimodal mode which is the free Gemini. Nano, which is embedded on certain brand cell phones. Ultra, the largest LLM model designed for highly complex tasks.
The free version of Google Gemini is more like a friendly assistant similar to the free version of ChatGPT. Google notes that it will store data for 72 hours and for sharing with other Google services and third-party services that users interact with while using Gemini. (https://heydata.eu/en/magazine/is-your-data-safe-with-google-gemini).
Gemini Advanced (paid version):
A self-described "supercharged version" of the free Google Gemini. It is much more capable of handling Python coding, crafting detailed scripts, logical reasoning following nuanced instructions, creative collaboration, ability to upload and deeply analyze files, documents, and data. It can be used across Google apps and services, and can write and edit Google Docs, compose emails. Gemini Live allows for interaction with Google AI using voice commands spoken in natural language, sort of like an "Ask Gemini" feature. You can upload documents for analysis, summary, and visualization.
From Google's Gemini Accps Privacy Hub: "Please don’t enter confidential information in your conversations or any data you wouldn’t want a reviewer to see or Google to use to improve our products, services, and machine-learning technologies." (https://support.google.com/gemini/answer/13594961?hl=en)
Overall this statement and Gemini's other privacy statements are similar to privacy and data use statements used by other Generative AI applications such as ChatGPT.
Other sources:
https://concentric.ai/google-gemini-security-risks/
https://www.androidpolice.com/google-gemini-vs-gemini-advanced/
- John Boone