NotebookLM founder: ‘AI is for brainstorming, not for writing’

Steven Johnson is a writer, veteran technology researcher, and author of many best-selling books on innovation. He is currently the co-founder and Director of Product Management of the NotebookLM project – Google’s “AI notebook” tool with millions of users, distinguished by its in-depth knowledge processing capabilities. Last year, Vietnam was in the top 10 in terms of users and most viewed documents of the tool.

On the sidelines of the Google I/O conference taking place in the US at the end of May, Steven Johnson shared with VnExpress about the future of notes and writing thinking in the AI ​​era.

– How was the journey to turn NotebookLM from an internal experiment into a popular product?

– Honestly, when I first came to Google, I just thought this was a fun experiment. I told myself, I have the opportunity to meet interesting people and touch the latest technology in the world, that’s enough. The chances of NotebookLM actually being released are slim, as Google has thousands of prototype products being built at any given time.

The turning point came in the fall of 2022. We took a few early, very rough prototypes and showed them to Josh Woodward (Vice President of Google, in charge of Google Labs) and CEO Sundar Pichai. Immediately after experiencing it, Sundar said he was very impressed with the product’s capabilities. Josh came back to tell me that the CEO was really fascinated by this idea, and I just woke up and thought, “Oh, so we’re actually going to make a real product here.”

Looking at the number of millions of people globally using NotebookLM, I still feel extremely amazed.

– As a writer, you have to consider every word in your work, but you created a tool to summarize and process information, helping people not need to read. What will the future of this industry be like, will we write for humans, or to serve machines?

– A very deep question and something I ponder every day as someone who has spent most of his career writing.

When I started building NotebookLM, my big dream was to turn it into a tool to help me become a better writer. I want it to support data research and spark ideas. But the core goal and vision from day one never changed: to create better, deeper, and more engaging books for human readers.

We are moving towards a world where, to some extent, writing must accommodate both the reader and the agent (AI agent). But intellectually, the only thing I crave is human connection. I use AI to amplify ideas, to fully understand complex documents and explain them back to readers. But my goal is to talk to people, not machines. So I always use AI mainly for my readers.

– Many people use AI to create content very quickly, as a result, the text tends to be stereotyped and similar to each other. What is the solution to this problem?

– The root nature and initial idea of ​​NotebookLM is to personalize the artificial intelligence you are working with. You give it your project’s own documents, clearly explaining the goals and expectations, and the AI ​​must have the ability to self-adjust to adapt to that need and information flow. The next step is to expand your possibilities to your creative style and personal voice.

I can take an example from myself. I’m writing a new book using NotebookLM and importing a lot of data about me, what I’ve published, and how I want to launch the new project. I still write directly, but when I ask the AI ​​to ask critical questions or brainstorm ideas, it responds with a keen understanding of me.

It no longer gives a generic, stereotypical answer, but actually matches my core worldview and vision for the work. Hopefully this will become the standard for those who use AI to create.

– It’s called “Notebook”, but the writing and note-taking features have not been placed in a central position, do you have a different vision for the product?

– This is quite a funny fact. The tool is named Notebook, you can still write and take notes, but that feature is not really in a direct position. For students, note-taking is a core and inseparable part of the learning and research journey. But we don’t want NotebookLM to stop at just a reading and summarizing tool. The goal is to turn it into a leading AI-powered note-taking space, an optimal interactive surface for writing, alongside inbuilt data analytics. Users will definitely see strong upgrades on this front in the near future.

– In the AI ​​agent trend, how will NotebookLM change?

– The upcoming experience on NotebookLM brings the feeling that you are truly working with an agent – a partner who understands and knows how to flexibly apply all the tools available inside your notebook, from creating documents, building output products to research tools.

Let’s look at the reality: if you want to do deep research in NotebookLM, you still have to actively enter special commands specifically for the in-depth query feature. But in the next step, the AI ​​agent will automatically understand what you have done in Notebook up to that point. It automatically scans and recognizes where your source documents are lacking, what data pieces are missing in the essay you want to write, so it can proactively search and compensate. In the coming time, users will experience a series of breakthrough functions of this type. Really interesting.