I’ve written about AI podcasts before, but until now, I didn’t really appreciate the degree to which there’s a rather inverse relationship between the time and effort required to create them (almost none at all) and the quality they’re able to achieve (these things are getting really, really good).
Case in point: I spent all of 30 seconds using Google’s buzzy new NotebookLM tool to create an extremely realistic-sounding podcast episode derived from several recent articles I’ve written. The two AI-generated hosts you’ll hear in the episode riff, quite convincingly, on each topic in my articles with the realistic banter you’d expect from actual human hosts — so much so that the result is, honestly, the most impressive AI-created thing I think I’ve encountered to date.
I’ll give you the podcast episode first and then show you how I made it. Click the play button below:
Pretty incredible, right? And again, I have to stress that it took almost no time at all to create what you just heard. It was super-easy to feed my articles into NotebookLM (try it yourself here), which allows you to upload anything from your own text notes to URLs as well as YouTube videos.
In my case, I simply inserted the URLs to four recent articles of mine from the past several days. They included:
- Article #1: My preview of a new documentary about the music of film composer John Williams, coming to Disney+ on Nov. 1
- Article #2: Everything to know about Territory, a new Netflix series coming later this month that’s set in the Australian Outback and sounds like it’ll basically be Netflix’s version of Yellowstone
- Article #3: Confirmation from Apple that its detective series Sugar, starring Colin Farrell, has been renewed for a second season
- Article #4: And, finally, my recap of what an awful year it’s been at the box office, with a long and growing string of expensive flops up to and including the new Joker sequel
Having tried it out, I can’t believe NotebookLM isn’t generating more buzz than it is right now. This thing is scary good, and even generated a — I know I sound like a broken record here — very convincing-sounding podcast episode based on one user simply uploading the words “poop” and “fart” over and over again. Listen to that result below, and marvel at the AI hosts musing out loud about whether we can find meaning in repetition, even by repeating words as stupid as these:
The product was reportedly created with the help of author Steven Johnson (via Wired). Google says that the app can easily help you summarize your notes, allow you to search specific things in notes you’ve uploaded, and even create outlines for you using your notes.
Among other things you can do with NotebookLM: It will answer detailed questions or surface key insights about documents you’ve uploaded, convert complex material into easy-to-understand formats like FAQs or briefing docs, and add resources to a notebook that are shareable with your org to create a group knowledge base. Pretty impressive all around.
