NewstalkZB’s fast-talking breakfast host Mike Hosking was once again crowned king of the airwaves in August when the latest radio ratings figures were released.
Hosking’s show is a juggernaut in FM radio land. But just as Paul Holmes’ era as ZB’s star host came to an end in 2008, so too will Hosking’s once he finally tires of the early-morning starts five days a week. But in the era of AI, will it ever have to end?
A couple of weeks ago at a conference in San Francisco, I watched Grammy Award-winning frontman of the Black Eyed Peas music group and tech innovator Will.i.am show off his new AI-powered radio station, Raidio.FYI. It features a series of AI-generated radio DJs, each with their own distinct personality and accent. The one I’ve been listening to on FYI, FYIona, has an East London accent and plays a mix of electronica and ambient soundscapes.
AI hosts aren’t new. The generative AI chatbots that first appeared two years ago with the debut of ChatGPT do an increasingly good job of reading out text, copying a newsreader or podcast host’s cadence and flow. Music steaming platform Spotify is testing its AI DJ with a subset of users. The host spends 15 seconds or so introducing the song with a bit of trivia about the artist, a nice addition to the music-only personalised radio stations Spotify has offered for years.
But with FYI, Will.i.am is taking the concept further, creating a series of streaming channels hosted by AI DJs and allowing interaction with listeners.
“The request line has always been awesome for radio when the callers get to call in and talk to the DJ or talk to the guests, but they’re limited to one person at a time,” the musician told Billboard magazine.
In the FYI app, available for free on iPhones and Android smartphones, you can tap the microphone icon on the screen to talk to the DJ. You can ask FYIona about the songs she is playing, or about current affairs. Sitting behind FYI are a series of “mega prompts”, information loaded in about the music featured in the show, topics of discussion and featured guests. A large language model draws on that information to assemble it into conversational passages and announcements. FYIona can access web-based news sources to comment on topical stories you ask about.
It isn’t as smooth and convincing as a human DJ – yet. But it’s listenable, and you can see where internet radio will head with a few more upgrades to the algorithms and tweaking of the language models.
By the time Hosking is considering hanging up his microphone, there’s a good chance ZB bosses will have “HoskGPT” in the works. Transcripts of Hosking’s thousands of past radio shows could be used to train it, emulating his take on current affairs, mimicking his Mike’s Minute segments. HoskGPT could hold thousands of talkback radio chats with listeners simultaneously, with ZB’s endless ad interludes, of course.
Is this the future of radio? Will.i.am thinks so. Gen Z reaches for streaming apps and TikTok before turning on a radio. Maybe RNZ’s slightly desperate efforts to appeal to younger listeners should embrace AI DJs and interactive music shows, too.
The next step, according to Will.i.am, is to generate all of the music with AI, too. That could keep costs down. But he’s run into a problem. The musician is a financial backer of Udio, an AI music generation platform that is being sued by major record companies for copyright infringement. Udio and its rival, Suno, are accused of harvesting millions of songs to train their systems, which allow songs to be created with text prompts. The case continues, but radio is unlikely to lose its human touch just yet.