Will Podcasters Be Replaced By AI? Google’s NotebookLM Points The Way

Are podcasters the next profession to be replaced by AI? What about radio and television pundits?

Between all the AI-generated horrible text articles and useless chatbots we encounter today, you will likely say, “Not a chance!” Before you dismiss the idea, though, I’ll share the results of an experiment I conducted.

Spoiler: we’re not there yet but we are far closer than you might guess.

Google’s NotebookLM AI Model

For the last few days, I’ve been experimenting with NotebookLM, an experimental AI model from Google. It’s a RAG-locked model, which means it uses only the information you upload. This makes it less prone to errors and hallucinations than AI models like ChatGPT, Claude, etc.

For example, I uploaded my books Friction and Brainfluence to a notebook and asked NotebookLM to compare them. In seconds, it generated a cogent comparison of the content and style of the two titles. I found its analysis to be spot-on. And, it was able to extract quotes without fabricating them.

Creating an Instant “Podcast”

Here’s the surprise. Recently, Google added a button to NotebookLM to create a podcast-like audio conversation between two people. Well, not really “people” – they are AI voices. There are no controls or settings, just a “Generate” button. It will use whatever you uploaded to a particular notebook: article(s), research papers, books…

The results were astonishing.

After a short wait, it produced a 15-minute conversation about the ideas in Friction. The combination of the script and audio were amazing. The language was casual and conversational. The two speakers played off each other, occasionally interrupting or adding a thoughtful “um.” Most importantly, they sounded perfectly human. If you listen to a short sample, you’d be hard-pressed to guess it was AI.

I continued my experiment by putting together a short video in which I paired a couple of avatars from another AI tool, HeyGen, with the Google audio. The result was, again, astonishing. The video of the conversation wasn’t perfect, but better than a lot of the human video you see on YouTube.

Because the Google audio is a single audio clip with both voices, I had to produce male and female versions of the video and then manually cut them to match the voices. Being able to export the two audio tracks separately would make video creation much easier.

From a content and audio standpoint, I’d say Google has successfully crossed the “uncanny valley.” The video version with avatars is very good, but a bit more detectable as AI.

I can imagine many use cases if Google turns what is currently a demo into a usable tool. They need to let users choose the number of speakers, age and gender, accents and languages, specify length and topic details, etc. Generating a video conversation with avatars inside NotebookLM (or wherever this tool ends up) would seem to be well within Google’s technical capability.

Even today, the audio summary feature in NotebookLM is useful. One colleague dumps long articles into it for an engaging audio summary. As my test shows, it can create an easy-to-digest book summary. And I can definitely see online articles offering an audio conversation version instead of, or in addition to, the straight audio narration version.

What would YOU use this for? Do you find this novel AI capability inspiring, creepy, or simply not very useful?