starsoftonline News

Seattle News Central: Discover What's Happening in the Pacific Northwest

NEWS

I AI-generated some podcasts – and the results are uncanny

Google’s new tool NotebookLM lets you create podcasts at the click of the button. They’re way more realistic than you’d think …
Stuart HeritageStuart HeritageWed 16 Oct 2024 14.00 BSTLast modified on Wed 16 Oct 2024 14.02 BSTShareAnyone who grew up watching The Terminator or The Matrix knows that AI poses an existential threat to humanity. As the robots become smarter, it was thought, they will inevitably replace us, either by destroying us or mining us for resources. However, the age of AI is now here, and the truth is so much worse than anything from a dystopian sci-fi. You see, AI has decided to give us more podcasts.
The world needs more podcasts like it needs to be kicked by a horse. Everyone’s got a podcast. Gyles Brandreth has a podcast. Paul Giamatti has a podcast. Your four or five worst friends all have podcasts, blathering endlessly into an environment already cluttered with too much content. Now Google has just created the first AI podcasts, and they’re as fascinating as they are superfluous.
Gyles BrandrethView image in fullscreenWe’re coming for you! Podcasting will no longer just be the preserve of the likes of Gyles Brandreth. Photograph: David Levenson/Getty ImagesNotebookLM is basically ChatGPT but for audio. You upload a bunch of sources – documents, websites, YouTube videos – and it parses all the information, then creates a bewilderingly human-sounding discussion about them. Two hosts, one male and one female, chat about whatever subject you’ve given them in an uncannily podcasty way. Their speech is full of ums and ahs. They hesitate, they talk over each other. They, like, kind of like talk like this all the time? It’s so imperfect that you can quickly forget you’re listening to a couple of robots repeating crap from the internet.
The world needs more podcasts like it needs to be kicked by a horseNotebookLM bills itself as a study resource, which makes sense. If you want to summarise a lot of information in a way that keeps your attention, or if you want to take in information on a run or a drive, then it’s great. Before long, people will prepare for exams by ramming their textbooks into something like NotebookLM then pottering about with earplugs in.
But if you want to make a podcast on any subject you like, it can do that, too. Rivals starts on Disney+ this week, and it has already generated acres of coverage, so I fed in a few interviews on the show to see what the hosts would come up with. The resulting five-minute podcast was uncanny.
In it, the hosts treated the show like it was something they had just organically fallen upon. “All right, get ready, because we are diving into Rivals!” the female host announces at the start, to a volley of agreeable muttering from her male counterpart. They discuss the attitudes of the 1980s, its sexism and racism, and applaud the show’s willingness to directly confront them. It makes it sound as though Rivals is a brilliant, pioneering piece of agenda-setting television.
Victoria Smurfit in Rivals.View image in fullscreenAgenda-setting television or just a bit of camp fun? Victoria Smurfit in Rivals. Photograph: Sanne Gualt/DisneyThe problem is that it isn’t really like that at all. It’s camp fun with loads of nudity. However, the sources I fed in were interviews with actors from the show, who are understandably more keen to talk about real-world issues than what it was like to whip their bits out all the time. And so that’s what the podcast is. A more accurate version would have fed in every available piece of information – interviews, reviews, show notes, maybe even the full source novel – and created a 360-degree view of the series. Instead, it was an extremely confident presentation based on limited information. And in the end, isn’t that all a podcast is anyway?
After that I decided to make the type of podcast the world needs least of, which is two people bibbling on about conspiracy theories. Admittedly, I could have done a better job here, finding flat-Earth forums and Facebook groups made up of people who still blame Covid on 5G masts. Instead, I just threw in a bunch of stuff from Wikipedia and Reddit, and found myself surprised by how measured the tone of the resulting audio was. It ended up being a fairly eye-opening 14-minute episode about things such as confirmation bias and the human impulse to understand the world. At one point, they start to list prevalent conspiracy theories, but stop because – as the male host says – “My brain would literally explode.”
The existence of NotebookLM raises a lot of questions. Will it make people too lazy to read their own research? Can it be fully trusted? What will humanity do with all the millions of newly unemployed podcasters roaming the Earth? But as a way of disseminating information for beginners in a naturalistic way, it’s sort of, like, brilliant.