Bounded Source AI and Google's NotebookLM
I am particularly interested in what is apparently known as “source bounded AI” such as Google’s NotebookLM. This allows you to “train” and individualized use based only on the documents that YOU have uploaded. I’m particularly interested in how that might be used in education and self-education. One of the real issues once you have gotten out of school is not being able to have a real discussion about things you read. While AI is not nearly as dynamic as a class discussion, it can also allow you to clarify confusion, develop and test arguments, explore arguments against the ideas in the document, and so forth. In a traditional college classroom, I’m curious about how I might use a source bounded AI as student preparation for in-class discussion (using the AI to do the things I mentioned), and then to have students provide the transcript of their AI discussion to me as homework. I could then collect class transcripts and upload them to my own AI and ask for a summary of common questions, common points of agreement and disagreement, what sort of arguments I might see, and so forth – so I could, as a classroom teacher, go into discussion with a deeper understanding of what my students are thinking. Encouraging students to prepare for class using an AI might, I suspect, make shy students more likely to speak in class, and help insecure students overcome their fear of asking a “stupid question” or expressing what they might think is an “unpopular opinion.”
If you’re at all interested, I have two videos I’ll link to below. The first is by Tiago Forte, the author of Building a Second Brain, and it is more of a tour of how to use NotebookLM. The second video is an interview with Raiza Martin and Steve Johnson, two of the people most involved in the development of NotebookLM at Google. It is a bit more focused on the added feature that will turn a text or your notes into a podcast-like audio discussion, which might also help students stay engaged and review on their commute.
Raiza Martin and Steve Johnson.
[Note: I’m not very interested in the question of whether or not we can trust Google to do what they say. This is about exploring the uses of a tool.]