I tried not to reflect too hard on that, knowing it would lead to further anxiety and self-doubt regarding my abilities as a writer and interviewer. If NotebookLM was faster than me AND more perceptive than me, what was the need for me?
Yes, I was experiencing anxiety and self-doubt. Who wouldn’t be?
During our interview, Johnson (a bestselling author of 13 popular science books) had tried to reassure me, as he has probably tried to reassure himself, that everything was going to be all right. However, looming over his reassurances was the uncomfortable fact that the bulk of his income now comes not from writing but from Google.
NotebookLM told me Johnson had used terms like “curatorial” and “quality control” to describe how writers could work with AI, and that while NotebookLM can generate possibilities, human writers still play a crucial role in evaluating, refining and shaping them.
These are facts I probably could have found myself, by going back through the transcript but why would I?
However, contra NotebookLM, Johnson hadn’t used the term “quality control” – I had. And I hadn’t meant it as a positive: no self-respecting writer wants to be reduced to standing at the end of the AI production line checking for defective sentences.
NotebookLM continued:
“Johnson suggests that AI tools like NotebookLM can empower writers by handling more routine tasks, allowing them to focus on higher-level creative and analytical processes.”
It concluded: “Greg’s anxieties and Johnson’s nuanced responses capture the complex interplay of fear, excitement and adaptation that characterises this ongoing technological shift.”
I thought that this was a pretty good summary. In time, I might have been able to write something of comparable quality, but why would I bother?
My opening line for this article was going to be: “I used Google’s new AI-powered research tool to research Google’s new AI-powered research tool for my story about Google’s new AI-powered research tool”. But when I entered that line into NotebookLM, it advised against it.
I had prepared for the interview with Johnson by googling “NotebookLM” and uploading everything I found to NotebookLM. My interview prep took about 10 minutes. By that time, NotebookLM knew more about NotebookLM than you, I, or any other person – even smart ones – could ever know about any single subject.
Not only that, but it had near-perfect recall of everything it had learned. Not only that, but I could ask it things like: “What are the five most interesting things about NotebookLM” and it instantly told me (e.g. “NotebookLM is a source-grounded AI-first notebook designed to help users understand things by gaining insights from their own documents.” Users can upload a variety of sources, including Google Docs, PDFs, websites, and YouTube videos. The AI then assesses and makes connections between those sources using Google’s Gemini 1.5 multimodal capabilities … ”
Had I asked it to produce this article based only on that material, it would have, and I have no doubt it would have done a good job.
But I didn’t do that, and not just because the Herald’s AI policy prohibits it.
Referring to the information I had given it, NotebookLM told me it has the “potential to automate some tasks traditionally performed by writers, such as research, summarisation and even idea generation”.
It suggested that, during the interview with Johnson, I: “Pose the question: Does he believe that AI will ultimately make human writers obsolete, or will it simply change the nature of the work we do?”
“We”? Who’s “we”?
I asked NotebookLM to write a story outline for me, based on the materials I had given it about itself.
It suggested a four-part story, titled “The Writer’s Dilemma”, centred on “Greg Bruce”, a writer struggling with writer’s block, who turns to Google’s NotebookLM, “hoping it can help him unlock his creative potential”, even though he fears it will take his job.
The outline started with Greg using NotebookLM as a “thought partner” rather than a replacement. He uses it to analyse old journals, articles, and even his acclaimed 2022 book, Rugby Head. It helps him discover new connections and insights, comes up with potential themes and structures, and even critiques his drafts.
Midway through the writing, however, he starts to wonder if the book is truly original (it’s not. It’s a total rip-off of his acclaimed 2022 book Rugby Head).
But, in the story’s final act, he overcomes his fears and begins embracing the “curatorial” and “quality control” aspects of co-working with NotebookLM, apparently reassured by the fact AI can’t replicate human intuition or emotional depth in storytelling.
In the poignant ending, he reflects on the unexpected ways the AI has helped him better understand himself and his writing. He finds peace in his new role in the world. He appears either untroubled by, unaware of, or uninterested in, the fact he’s produced a work of shameless self-plagiarism.
I had to admit it was a compelling outline. I wondered what NotebookLM would do if asked to turn it into a story.
I wrote: “Work it up into something featuring human intuition and emotional depth.”
And to its credit, it tried.