GONE are the days of computers as mere tools. NotebookLM, an AI-powered research tool co-created by Steven Johnson and Google Labs, is evolving into a true thought partner. It allows users to upload their own documents and interact with various sources through features such as chat, summarization, study guide creation, and audio overviews. I had an interesting conversation with Johnson at the Google Philippines Headquarters on the origins and the collaborative nature of NotebookLM.
As a renowned author, co-creator of NotebookLM, and editorial director of Google Labs, Johnson has always been fascinated by the intersection of technology and thought. As someone who has written 14 books, Johnson’s research methods have always involved meticulously organizing quotes and notes.
Over the years, he compiled a searchable database of 8,000 quotes from his reading materials. This commitment to structured information led him to dream of a tool to store data and intelligently assist in the creative process. The opportunity to turn this dream into reality came when Google Labs approached him in 2022. The team, already familiar with his work, invited Johnson to collaborate on developing an AI-powered research tool. Thus, NotebookLM was born, designed to help users manage and interact with their sources in a meaningful way, powered by advanced language models.
Johnson worked closely with Raiza Martin, a senior product manager for AI at Google Labs, who leads the team behind NotebookLM. Together, they envisioned a tool that would evolve alongside advancements in AI. Johnson’s unique approach to note-taking, particularly his extensive quote collection, directly influenced the design and functionality of NotebookLM.
Each NotebookLM notebook has a capacity of 25 million words, roughly equivalent to 40 books, due to the 500,000-word limit for each source file. While a single notebook has a limit of 50 source files, users can create multiple notebooks for different projects. Earlier versions of the tool were limited by the context window — the number of words the AI could focus on. Still, the introduction of Google’s Gemini model expanded this capacity dramatically.
Today, NotebookLM can process up to 1.5 million words simultaneously, allowing for much deeper interactions with complex materials. For Johnson, this functionality is personal. In one of his notebooks, titled “Next Book,” he stores ideas and sources for potential projects. The AI helps him brainstorm by suggesting chapter structures and offering insights based on the documents he uploads. According to Johnson, it feels like “a personalized AI” built to complement his workflow.
One of the most surprising breakthroughs with NotebookLM was the introduction of AI-generated discussions or audio overviews. These audio conversations are known as Deep Dives. Initially developed by another team within Google Labs, this feature allowed users to listen to simulated conversations based on their research materials.
The audio tool, which quickly gained popularity on platforms like TikTok, enabled users to engage with their content in a conversational format, much like listening to a podcast. When testing the Deep Dives on one of my notebooks, the facts were accurate, but some of the host’s insights were not. While this highlights that the technology is still under development, it doesn’t diminish the potential of NotebookLM across various applications.
Some of the most promising use cases are creating audio study guides for auditory learners and enhancing performance reviews with AI-generated feedback. It’s also compelling to turn resumes into engaging narratives, analyze business documents, generate podcasts, and provide personal finance insights. Its ability to transform information into various formats makes it especially useful for different learning styles, and user feedback is shaping its future.
A key aspect of NotebookLM is its emphasis on privacy. Unlike other AI models, NotebookLM does not use user data to train its algorithms. Johnson emphasized that this decision was critical for maintaining trust, particularly for those working with sensitive or copyrighted materials. With enhancements in language processing, handwriting recognition, and multi-modal capabilities, the future of AI-assisted research looks promising. Whether through text-based analysis or audio conversations, NotebookLM offers a glimpse into how AI can transform how we create, learn and understand complex information.
Recognizing the importance of a diverse range of perspectives, Johnson embraced a collaborative approach to developing NotebookLM. He also emphasized the importance of user feedback, actively engaging with the community on Discord to understand how people were using the tool and what improvements they wanted. “I don’t think of my word processor as a collaborator, but I think of NotebookLM as my collaborator in some way,” Johnson added.
What was once a passive process of gathering and organizing information is now an active collaboration between humans and AI. As NotebookLM evolves, its potential to assist writers, researchers and learners in new, dynamic ways will only expand.
Listen to the audio overview of my interview and two secondary sources generated on NotebookLM on my blog, techiegadgets.com.
Be the first to comment