NotebookLM: Its More than AI Podcasts- Your Personalized AI Research Assistant

[@unclecode]

Great review. OpenAI Canvas and now this. This is a turning point where giant companies will build many utility tools by themselves. To stay ahead, startups and entrepreneurs must know exactly what to focus on. Unfortunately, when these companies release new features and many newly created startups, fueled by excitement and rapid investment, will likely fade away in due time.

[@airebreton]

Researching entire KZbin channels focused on a specific topic has become much easier. Soon, marketplaces will emerge where "notebooks" are sold, featuring professionally curated bibliographies and equally well-crafted prompts to help users accelerate their learning on a wide range of subjects. But does anyone really understand the computational demands of managing a bibliography that covers, say, ten full books of 400 pages each? I don’t think that matters much to the end users. For them, just opening a window into a topic is a big step forward. From there, they can choose to dive deeper into specific resources if they want to.

[@alexanderroodt5052]

It is one of the most entertaining things I have ever used

[@paulham.2447]

You are right it is much more than an AI podcast! Thank you

[@kushis4ever]

I have used it quite some time now. Be careful. While it's impressive, the accuracy is 50-50. E.g. if you ask to synthesize a topic from your pdf documents, it will not look at all of your documents but will only touch on only a few of them, that also not in depth. No matter how good you are in prompt engineering, there is no improvement. Also, there are frustrating issues like it cannot output citations like research papers (gives numbers to references text-you have to work on your APA citations yourself), you cannot undo prompts (have to start new notebook if you don't like the chat chain), saved notes loses links to numbered references (only numbers are shown, links are lost).

[@sean-uw6op]

Do you know how this podcast feature in notebooklm works, its architecture, please let me know if there any paper which explain this.

[@anthonymannwexford]

Excellent. Thank you

[@x3d561]

1:17 Do you how to easily create the videos Google used in their landing page?

[@calvin2013]

Curious as well

[@LeonardoDLuca-k7v]

Any ideas which kind of rag system NotebookLM use?

[@engineerprompt]

Seems like it puts everything in the context of the LLM and then have some custom prompts on top of it.

[@LeonardoDLuca-k7v]

​@@engineerprompt so the 2M context window of gemini 1.5 pro and caches can actually work really well? That is cool

[@KumR]

I use this for my podcasts but can we use own voice ? kzbin.info/aero/PLCiLnvtXMbwBanBBdMTz0eCx5_Te2hosQ&feature=shared

[@ysy69]

Are you on NotebookML for business pilot? I noticed mine doesn't have the audio summary sharing option yet. I love this product, just made one on The State of Crypto 2024 Report: kzbin.info/www/bejne/eHOkgJWYq6mfgKM