I Listened To AI Podcast For 10 Days So You Don’t Have To

Here are my review and thoughts on Google's NotebookLM used as a tool to help me study.

my newsletter

notebookLM

This video is supported by the kind Patrons & YouTube Members:
🙏Andrew Lescelius, Ben Shaener, Chris LeDoux, Miguilim, Deagan, FiFaŁ, Robert Zawiasa, Marcelo Ferreira, Owen Ingraham, Daddy Wen, Tony Jimenez, Panther Modern, Jake Disco, Demilson Quintao, Penumbraa, Shuhong Chen, Hongbo Men, happi nyuu nyaa, Carol Lo, Mose Sakashita, Miguel, Bandera, Gennaro Schiano, gunwoo, Ravid Freedman, Mert Seftali, Mrityunjay, Richárd Nagyfi, Timo Steiner, Henrik G Sundt, projectAnthony, Brigham Hall, Kyle Hudson, Kalila, Jef Come, Jvari Williams, Tien Tien, BIll Mangrum, owned, Janne Kytölä, SO, Richárd Nagyfi, Hector, Drexon, Claxvii 177th, Inferencer, Michael Brenner, Akkusativ, Oleg Wock, FantomBloth, Thipok Tham, Clayton Ford, Theo, Handenon, Diego Silva, mayssam, Kadhai Pesalam, Tim Schulz, jiye, Anushka, Henrik Sundt

[Video Editor] bhargavesque

0:00 Intro & context
4:07 Review

[IsaacFoster..]

I put my college notes in it the podcasters were like,
"uhh, yeah... These are some difficult topics... Not for starters for sure"

[AAjax]

I'm pretty sure it's 100% meant to just be engaging and entertaining, and a way to first dip your toe into the water. It's a podcast for normies not acquainted with the subject, not a lecture from an expert. There's a heck of a lot of slop on most human podcasts too.
If you have more in-depth questions, that's what the rest of NotebookLM is there for.

[olivetree]

It's good for little wikipedia deep dives when I am driving to work

[gamagama]

god theres gonna be even more ai youtube slop isnt there

[JohnLewis-old]

Conclusion: AI summary sounds human, but it's still a summary. It's hard to take large papers and give reasonable, detail rich summaries.

[reinerheiner]

Honestly I don't use the podcast feature at all but notebooklm has some other great features. Such as when giving it a source, that it completely restricts itself to said source. Or, that it provides sources for everything it talks about, so you always know what part of the source it is using. This makes it a pretty great tool for talking to, working with, and learning information from sources like pdfs. Or for quickly finding where in the source information is located that cant easily be found by using keywords. And the LLM can be used in many other ways as well of course.

[IvarDaigon]

when an "AI" says things like "They are teacing AI to be more like us" that is the clearest example of how AI systems are not self aware.

[skybasee]

this is exactly what I wanted as someone to review some AI technology that potentially help us, thank you for helping us test them.

[VorticalGab]

I use it to check if a paper is worth taking a deeper look into. Its good and it is fun to listen during walks and washing the dishes.

[FuZZbaLLbee]

That’s wild!

Uhu uhu, imagine that

[YT-gvcz]

I've tried generating a podcast based solely on *that* nlab article on Hegel's Science of Logic, which I've been using as a baseline for "arcane mind-fuckery that might actually make perfect sense to a super intelligence". The podcast expectedly fails to get into any concrete details, but it does have a stellar opening: "Have you ever opened a dense philosophy book and say, I really wish there's more quantum physics in it", which does capture precisely the vibe. So I guess I will give it credit for that.

[OperationDarkside]

This is probably good for what I'm using chatgpt right now. Getting a rough overview of DC circuits. You can't die from a 12V sub 1A circuit, that only powers a fan with a solar panel.

[AB-wfek]

I actually found a bunch of joke research papers and prompted them to be in a state of paniced anguish in front of a condescending and belligerent audience, which was pretty entertaining.

[sori]

Have you tried to prompt it to go more into detail? I am a medstudent and here is my go to prompt and it massively improved the output and also sometimes went up to 30 minutes of runtime! (Albeit repeating a lot of things sometimes…) „ you are two medical professionals making a podcast for doctors and medicine students. Focus on the chapter Item 24 – Principales complications de la grossesse
and review the chapter in detail. Come up with mnemonics and tell stories about patients to help remember the facts. Mention the important french medical terms used in the book.“ (they also skewer the french terms in pronounciation which is quite funny)

[caroxic-tokyo]

I find this super fascinating!
Actually once I uploaded a bookled from the EU with like goals and such.

It generated a Podcast that was 36 min long! Interestingly they split it into 3 Parts. So after like 10 minutes they said "Well this is only the first part. See you at the next part!" - kind of and that for 3 times. Each time focussing on different aspects.

[festinuz]

I was not prepared for the "artifact" at 11:02, had to re-listen half a dozen of times until i could stop laughing

[chess]

It would be interesting to see how many of these problems you could get rid of with the new prompting you can do, and how you can adjust what it talks about.

[BryanBortz]

Hey, whatever makes this video unique from your others, keep it up. I appreciate the assessment

[theAItechtoolbox]

Super interesting review of Notebook LM's capabilities! It's great to hear both the positives, like speeding up initial research, and the drawbacks, like its tendency to oversimplify or hype things up. AI-generated podcasts feel like a glimpse into the future of personalized learning, but clearly, they’ve got some refining to do. Love the transparency and detailed testing you put into this—curious if you'd see improvements with updates or alternatives in the future!

[VaibhavShewale]

well you can add the prompt before generating the audio and that helps alot