I posted this photo of the Moon and Venus back on Jan 6. Shortly thereafter I was contacted by a former student who had used a new AI tool to convert that post into an audio podcast. The tool, built by Google, is called NotebookLM. You can read more about it here.
I’ve dabbled with AI in the past, and even used it experimentally to write a post about The True Shape Of Raindrops. I was not impressed. But that was over a year ago and AI has become increasingly capable of generating realistic output. The podcast linked to below runs around 10 minutes. You won’t have to listen to all of it to get my point. But first, if you haven’t already, read my Jan 6 post. It’s only 437 words. Reading it will better prepare you for what you are about to hear:
https://sky-lights.org/wp-content/uploads/2025/01/Moon-Venus.mp3
Those were not real people in that podcast. Those were entirely synthesized voices. The input to NotebookLM was the text of my post, plus the text on my About page for this website. From that content it created a realistic podcast discussing the key features of my original post. Unlike the raindrop post, this time the AI was spot on and scientifically accurate. Well of course it was … I wrote it.
What struck me the most was the realistic banter between the two podcasters. The back-and-forth exchanges, the vocal inflections, the interruptions, the chuckles, the use of colloquialisms … to be honest, the experience was quite unnerving. What does AI portend for content creators like me? Or graphic artists? Or musicians? Or actors?
If someone wants to steal text from your blog, turn it into an audio podcast, and slap it on their YouTube channel, how would you know? They would surely change the title, and you can’t really “search” audio.
I am nonetheless ready to embrace AI as a tool for content creation. I believe it can be useful as an adjunct to real creativity. But I am concerned about copyright infringement. I am also concerned about the potential for AI to help create misinformation — that’s already a problem and sure to get worse. Fake graphics and videos are becoming harder to identify, and the cloned voices of well-known people are being used to create fake audio. It’s a brave new world. Be careful out there.
Next Week in Sky Lights ⇒ Extreme Weather Attribution Science