I investigate AI developments from two perspectives. First, what tools and applications will be personally beneficial? Second, what tools and applications can I write about that would be of value to those who read my posts? The following description fits both categories at the present time, but I expect it would end up more in the second category should Google move it out of beta into a paid category expected to be $20 a month. If it becomes part of the Google for Education suite, I can imagine it having great value to those with access to these services.
My present personal AI interest is in focusing AI on the resource material I have accumulated. What I mean by this is that I take notes and highlight while I read and have done so for 50 years. For the period of that time that I could read journal articles and books in digital form, I typically could export my notes and highlights and accumulate this personalized material. I have collected this content, but now I can use AI to target this content with retrieval augmented generation (RAG). I can “chat with this content” which offers me control over AI that I think is important and different from just interacting with the general knowledge base on which AI was trained. I want to generate insights and produce written products based on content and specific ideas within that content I have personally vetted. The broader application I see for what I am describing here involves an educator offering students access to content the educator has collected. An even broader application might focus on content collected by a team with a common interest.
This post described the use of Google’s NotebookLM as a tool suited to the implementation of this idea. I have described in a previous post how I get the content I want NotebookLM to focus on into that AI service in a previous post. Here I want to explain how “Mind Mapping”, a new capability of NotebookLM, can be used to explore a body of content.
So mind mapping is a way to identify the structure of ideas within content. I would have preferred Google called their implementation concept mapping, but this is not what they did. Concept maps can be a way for some to convey the structure of ideas to someone else or it can be a task in which someone creates a mind map to demonstrate how they see ideas to be related. The reason I would have preferred concept mapping is that the Google NotebookLM capability identifies concepts and then generates a simple structure of how these concepts are related. Think of it this way. I can feed in a large collection of information I have collected and then had NotebookLM show me how this content could be organized. In addition, it will provide summaries of the nodes that it has identified and allowed me then to explore the content I fed in that were judged to justify parts of the summary.
The following image shows NotebookLM already loaded with hundreds of notes and highlights (left hand column) and the button (red box) that will generate this first level of the mind map. To break one of the initial notes into subnodes, you click on the caret associated with a note.
Selecting one of the nodes will reveal a summary of the content making up that higher-level category. In the following image, the summary is based on the category Generative Aspect (red box).
Within the summary, you should be able to identify numbers that represent the source from the content referenced by that section of the summary. Selecting a number will display that note or section of original content.
One final feature is also quite useful. NotebookLM suggests questions related to the content displayed you might want to ask. It also provides a text box you can use to enter a prompt suggesting a question of your own.
Summary
NotebookLM now includes a mind mapping tool that identifies and organizes concepts from the content it has been fed. The nodes identified can be used to provide summaries of that content and to interact with that summary and the content on which the summary was based. To fully appreciate what this allows it may be useful to imagine that hundreds or thousands of notes could be submitted by a user and processed in this manner.
You must be logged in to post a comment.