Simon Willison:

Audio Overview is a fun new feature of Google’s NotebookLM which is getting a lot of attention right now. It generates a one-off custom podcast against content you provide, where two AI hosts start up a “deep dive” discussion about the collected content. These last around ten minutes and are very podcast, with an astonishingly convincing audio back-and-forth conversation.

Here’s an example podcast created by feeding in an earlier version of this article (prior to creating this example).

I listened to the whole 15-minute podcast this morning. It was, indeed, surprisingly effective. It remains somewhere in the uncanny valley, but not at all in a creepy way. Just more in a “this is a bit vapid and phony” way. I think that if you played this example podcast for a non-technical person who isn’t informed at all about the current state of generative AI, that they would assume for the first few minutes, without question, that this was a recorded podcast between two actual humans, and that they might actually learn a few things about generative AI. But given that the “conversation” is literally about creating artificial podcasts like this very example, I wonder how many would, by the end, suspect that they were in fact listening to an AI-generated podcast? It’s quite meta — which the male voice on the podcast even says during the episode.

But ultimately the conversation has all the flavor of a bowl of unseasoned white rice. Give it a listen, though. It’s remarkable.

Update: Jiminy Christ, listen to this one, where the prompt was a document with nothing more than the words poop and fart repeated over and over.