In a rare visit back to TwitterX, I fortuitously stumbled across a conversation between the prolific author Steven B. Johnson, now editorial director of Google Labs (who knew?), and This Week in Startup‘s host and MAGA wannabee Jason Calacanis. In the convo/demo, the erudite Mr. Johnson kvelled over his latest creation, NotebookLM.
While the world marvels at the large language models (LLMs) of OpenAI and Google Bard, the really cool stuff (IMHO) is happening with what I call SLMs or smart or small language models. In this particular SLM, the user is given the chance to upload 20 documents of any length to create a custom dataset to which one can pose questions and receive back intelligent, insightful, and prosaically written answers based on the uploaded content.
I immediately decided to put it to the test for one of my PR clients whose CEO is a well-respected subject matter expert in his industry. I was lucky to have the audio of the CEO’s six most recent media interviews, which I transcribed using my Otter.AI account. Those transcripts were the first documents I loaded into NotebookLM.
I then ventured over to my client’s website and grabbed the content from the last two industry white papers his company had produced. They went in, which was followed by a couple of his speeches, some independently published scientific studies, an industry trade association report, pertinent news from government and regulatory authorities, and several deep-dive think pieces from reputable news organizations and trade journals.
Now the fun begins. What does the CEO think of this trend in the industry? What is his company’s position on this important issue? How does his industry view the latest industry data? You get my drift.
Voila! Each of these questions produced very substantive and well-written, answers that exactly reflected the sentiment of the CEO, his company, and the industry in which they reside. Here’s a screen shot from the demo Steven gave to Jason. On the left are the inputted five chapters from Johnson’s tome “Where Good Ideas Come From” and in the center is NotebookLM’s output after being prompted with a question:
NotebookLM lets you copy and share the output, add notes, and rate it. The platform also saves your inputted content and queries. I will say one thing: I’m pretty sure I’m just scratching the surface of what this cool new AI tool is capable of doing. Google Labs describes it as “an AI-first notebook, grounded in your own documents, designed to help you gain insights faster.”
On that level it surely succeeds.