M.G. Siegler:

Like everyone else, I had heard all the hype around NotebookLM, the new AI product from Google in which you give it content to ingest and it spits out a... podcast. But not just a dry, computer-voice overview podcast, an almost humorously human-sounding one. With cohosts. But it wasn't until I gave it a few of my own recent posts to work with that I truly appreciated the output.

Listening to these, I had the weird feeling that, "wow, they really got that point." Wait. Who is they?!

To be clear, the outputs aren't flawless. And actually, after listening for a few minutes, you can pretty clearly understand what they're doing with the back-and-forth banter and some of the "reactions" start to sound quite repetitive. But even that yields a weird effect where you just kind of think that one of the "hosts" is a bit dumber than the other. And that's the way it is in many actual podcasts! And actually, that should be easy enough to clean up, Google just needs a lot more variety in reactions – including, sometimes, no reaction.

There are also a few factual errors here and there – but that almost makes me wonder if I simply couldn't have explained what I was talking about in my original post in a more clear manner!

Anyway, color me quite impressed. The product itself is still too obtuse with the podcasting feature buried in the "Notebook guide" area when it really should be front-and-center. I get that this is just one feature of the product, but really, for most people now hearing about NotebookLM, it is the product.