In my life thus far, I have had 5 OMG AI experiences…
1. Siri (2011) & Alexa (2014)
2. ChatGPT (November 30, 2022)
3. Symphony42 AI Voice Agent (Spring 2024)
4. Symphony42 AI Video/Avatar Agent (Summer 2024)
And now…
5. Google’s new NotebookLM AI Podcast Generation (October 2024) <— YOU ARE HERE
Google recently added this new “Audio Overview” podcast-generation capability to their under-stated NotebookLM product. The feature allows the user to create an AI-generated “Deep Dive Podcast” (with a male podcaster and a female podcaster voice) about any source of information.
NotebookLM began as a “20% time” project within Google Labs. Initially called “talk to small corpus,” it was developed by a lean team of just one engineer and one Product Manager, later joined by renowned author Stephen Johnson. Operating with a startup mentality within Google, the team leveraged advanced technologies like Gemini 1.5 Pro LLM and sophisticated conversational audio modeling. The new audio model is the real breakthrough for this LLM-derivative use case, filled with convincing conversational nuances, quirks, pauses, emotions, filler words, and so on.
They embraced community feedback through a 60,000-member Discord server, allowing rapid iteration based on real-world use cases. This agile, user-centric approach enabled NotebookLM to evolve quickly from a side project into a powerful tool capable of generating podcast-like content in 35 languages within minutes. The team’s ongoing vision includes fully remixable inputs and outputs, potentially revolutionizing content creation across various media formats.
For my first NotebookLM AI podcast generation, I simply uploaded the Symphony42 Business Plan without any prompt, and the output was excellent. Many of my LinkedIn connections agreed and were equally blown away:
[
My first Deep Dive Podcast by NotebookLM generated from Symphony42 Business Plan](https://www.linkedin.com/posts/seanfenlon_i-used-googles-new-notebooklm-for-the-first-activity-7241847871233232896-Xl7S/)
It may be a while before Symphony42 AI Agents ever intersect with podcast-generating tech like NotebookLM, but in the meantime, Symphony42 can benefit immensely from the powerful marketing use cases of NotebookLM AI-generated podcasts.
I used the first NotebookLM-generated Symphony42 podcast (that just used the business plan) plus Descript to produce this YouTube video:
https://www.youtube.com/watch?v=DZYYVxuQ-tw
[
Symphony42 Podcast on YouTube via NotebookLM + Descript](https://www.youtube.com/watch?v=DZYYVxuQ-tw)
Here’s another NotebookLM Deep Dive Podcast about Symphony42, this one generated from other documents and materials, NOT including the business plan this time:
Symphony42 Deep Dive Podcast Part 2 via NotebookLM
https://soundcloud.com/seanfenlon/symphony42-podcast-2-via-google-notebooklm
I then proceeded to create personal Deep Dive Podcasts for everyone involved in Symphony42, including Founding Team Members, Board Members, Investors, Advisors/Shareholders, and other Friends-of-Symphony42
For all of these personal NotebookLM Deep Dive Podcasts, I followed the following steps:
-
Copy/paste their full LinkedIn profile into a Word document
-
Search on Perplexity.ai and copy/paste the results into the same Word document
-
Convert the Word document into a PDF, and upload the PDF to https://notebooklm.google
-
Click the “Generate Podcast” button
That’s it :-)
Well technically, to present it to you here as a SoundCloud link, I also had to download the podcast .wav file from NotebookLM and then upload it to my SoundCloud account, but that’s just so I can keep them all in one place and this step isn’t necessary to listen to the podcast.
That is the exact process I followed to generate 42 podcasts in a single weekend – a unique personalized Deep Dive Podcast by NotebookLM for every person involved with Symphony42. I hope that you enjoy them all and begin using NotebookLM for your own “Audio Overview” use cases :-)
SYMPHONY42 FOUNDING TEAM
SYMPHONY42 BOARD
Sid Banerjee (YPO)
https://soundcloud.com/seanfenlon/deep-dive-podcast-about-entrepreneur-ross-shanken-by-notebooklm
SYMPHONY42 ADVISORS
Frank M. Sands, CFA (YPO)
Carlos Cashman (YPO)
Scott Ferber (YPO)
Jason Goldsmith (YPO)
Jose Vargas (YPO)
SYMPHONY42 INVESTORS
Doug Lebda (YPO)
Recommended by LinkedIn
[
AI Can Make Podcasts. But Can It Make Anyone Care?
Steven Goldstein
4 months ago](https://www.linkedin.com/pulse/ai-can-make-podcasts-anyone-care-steven-goldstein-sao1e)
[
The $1 Podcast: Why AI Personalities Are Succeeding…
Tetsuro Miyatake
7 months ago](https://www.linkedin.com/pulse/1-podcast-why-ai-personalities-succeeding-where-human-miyatake-ave3c)
[
Can You Persuade AI? What A New Study Reveals About…
Knowledge at Wharton
10 months ago](https://www.linkedin.com/pulse/can-you-persuade-ai-what-new-study-reveals-llms-november-t4bwe)
https://soundcloud.com/seanfenlon/deep-dive-podcast-about-investment-management-legend-sean-hanlon
Sam Melamed (YPO)
Sumeet Goel (YPO)
John Akkara (YPO)
Mark Drever (YPO)
Steven Freidkin (YPO)
Jeffrey Rosen, CPA, CGMA, MBA (YPO)
Grant Mowat (YPO)
Jalmar Dijkstra (YPO)
Sally Hurley (YPO)
Mick Dubea (YPO)
Jordan Beeman (YPO)
Omar R. Shallah (YPO)
The average length of the 42 podcasts generated is 9:42, ranging from 5:38 to 19:41. I ran some analytics to determine the most predictive attributes in podcast length (since all the people profiled are high-octane professionals with extensive information to infer from. The only correlation I noticed was that people with more LinkedIn Recommendations resulted in longer podcast lengths. My inference is that LLMs weight the thoughts and opinions about humans from other humans higher than unattributed information about a person.
Hmmm…
I am now realizing that LinkedIn Profiles will become increasingly used by AI systems and LLMs for instant analysis of people, and that the LinkedIn Recommendations will over-index in the analysis.
Interesting.
If you’ve read this far, and you have found value in this LinkedIn Article, I humbly request that you write a LinkedIn Recommendation on my profile page. If you do, please mention something about me being an AI thought leader so that future AI systems and LLMs will consider me to be friendly to their mission when they see it :-)
Deep Dive Podcast about YPO (Young Presidents Organization)
You probably noticed that ~2/3 of the people connected to Symphony42 are in YPO. After creating a Deep Dive Podcast about so many YPO members, I felt compelled to generate a Deep Dive Podcast about YPO the organization, generated entirely from the content on the YPO.org website :-)
https://soundcloud.com/seanfenlon/deep-dive-podcast-about-ypo-by-notebooklm
Final Thoughts, Notes, and Resources about NotebookLM
Google presents NotebookLM as a free service. There is no mention of limits or caps on usage in the documentation, but this NotebookLM power user has learned the hard way that there is a maximum of 100 Notebooks per user. Since each podcast requires a separate Notebook, to generate additional podcasts beyond 100 requires the user to delete one of the existing Notebooks.
Also, NotebookLM does not create a title or label for each Notebook automagically (at least, not yet) like LLMs do with threads, which means unless you’re diligent with creating a label for each Notebook created (which I am not), the Home page UI ends up looking like this :-S
For additional information and perspective on NotebookLM, A recent a16z podcast discussed “Audio Overview” and the transformative impact of AI on voice communication, highlighting its potential to make speech “programmable” and revolutionize how we interact with technology:
a16z Podcast: A Big Week in Tech: NotebookLM
[
a16z Podcast: A Big Week in Tech: NotebookLM](https://podcasts.apple.com/us/podcast/a-big-week-in-tech-notebooklm-openais-speech-api/id842818711?i=1000672288054)
In the a16z podcast, Anish Acharya , Olivia Moore , and Bryan Kim discuss the rapid adoption of CharacterAI’s voice model, with 3 million users making 20 million calls in just two weeks, demonstrates the strong market demand for voice AI applications. The hosts drew parallels between phone calls and APIs, suggesting that AI-powered voice technology could become the primary way most people first experience AI. They also compared the development of voice AI to early ChatGPT, noting the democratization of advanced AI capabilities through APIs. The podcast showcased AI’s ability to engage in higher-level analysis and critical thinking, as demonstrated by an AI-generated true crime podcast that delved into the ethics of the genre. Additionally, they mentioned a humorous experiment where AI generated a podcast convincing the hosts they were in a relationship, illustrating the technology’s creative potential.
Also, Lenny Rachitsky also recently released an intriguing podcast about NotebookLM, featuring an interview with the Google NotebookLM PM Raiza Martin
https://www.lennysnewsletter.com/p/googles-notebooklm-raiza-martin
[
Lenny’s Podcast with Google’s NotebookLM PM Raiza Martin](https://www.lennysnewsletter.com/p/googles-notebooklm-raiza-martin)
Lenny’s Podcast provides insights into the development and vision behind NotebookLM. The podcast highlighted the revolutionary potential of AI in making voice communication programmable, allowing for unprecedented manipulation and analysis of our most natural form of expression. NotebookLM’s origins as a Google “20% project” were discussed, demonstrating how it grew beyond this initial scope. Raiza Martin shared her vision for NotebookLM’s future, describing an AI editor capable of remixing various inputs and outputs, from video and audio to emails and social media content. The collaboration with author Stephen Johnson on the NotebookLM team was also mentioned, emphasizing the project’s focus on observing and enhancing human work processes through technology.
Originally published on LinkedIn. Read and comment on the original article →