Back in the summer of 2011, I met Denise. She seemed nice enough but the more I got to know her, the less I liked her.
Denise was not a real person. She was a “virtual DJ” at a community radio station in Texas. No flesh and blood. No actual personality. No independent thought. For $200, the program director bought some software that was a (un)reasonable facsimile of a live radio announcer. Denise still required human input for her to work — this was 12 years ago, after all — but the idea of a computer program replacing a human DJ was horrifying to me and others in the media business.
Consultant Dave Ramsey wrote this at the time, pondering the importance of the DJ’s role on the radio.
“No jock who adds significant value to the radio experience need fear ‘Denise.’ She will always be a wooden Pinnochio in a world of real boys.
“However, any jock whose contribution to the station is indifferent from ‘chatter’ — any jock who is more of an obstacle to giving listeners what they want than an asset in providing the kind of relevance and spontaneous joy great jocks have always been famous for — those DJ’s are no better and certainly more expensive than ‘Denise.’
“In other words, if a jock can be replaced by ‘Denise’ with no fallout to the station in terms of ratings, revenue, or audience and advertiser satisfaction, then that jock can and should be replaced.”
Radio — in fact, all linear media — is under fierce revenue pressures brought on by a changing technological landscape and the new and different ways people are consuming media. Dominique Garcia, the program director behind Denise, defended the move to an AI DJ, claiming that he was actually performing a service.
“If Denise took over the role of a human on the air, a program worth roughly $200 buys a lifetime jock for that radio station. She does not require an annual salary, she never gets sick, does not need sleep, food, or has the need for restroom breaks. She can literally be on the air 24/7! Would a human work an entire year straight, 24 hours a day, seven days a week, for only $200? Clearly, the answer is no. Would a human work five years straight, 24 hours a day seven days a week for a one-time fee of only $200? It is easy to say absolutely not. Depending on the market and how much stations pay their on-air talent, the money saved over the years easily adds up by not having to pay humans an annual salary.”
In other words, the justification for something like Denise is cost-savings, the elimination of jobs, and no more nasty humans to manage and coach.
The furor over Denise died away after a few weeks and to my knowledge, software like this has been used sparingly and under the radar. But then along comes China.
In 2018, Xinhua, the country’s state-run news agency, was looking to save money by replacing flesh-and-blood news anchors. Modelling an AI anchor on a Chinese journalist named Zhang Zhoa, the new newsreader debuted that fall, introducing itself as “I am AI news anchor in Beijing.” It was real news from a fake.
The latest attack — and yes, I’m obviously biased given what I do for a living — comes from Spotify. Last week, it introduced a feature called DJ in Canada and the U.S. It’s a “generative AI” that curates your playlists as well as offers commentary in between the songs, information that’s supplied by human editors. There’s a text-to-speech component that translates whatever the editors write.
Here’s a demo of the product that is still in beta — but the tech is still awfully listenable. A bit generic, but still…
“X” is apparently a real-life DJ named Xavier Jernigan, Spotify’s head of cultural partnerships. That’s not him speaking all the time, of course. The AI has learned the sound of his voice and manner of speaking. From there, the X can say whatever is necessary to go along with the algorithmically chosen songs. It’s a form of deep fake. And if you didn’t know you were listening to an AI, you’d never know — and you’d probably not pay much attention. If you have a station full of average announcers, there’s really not much difference — except that the meat bags cost a lot more.
Spotify has this to say: “The DJ knows you and your music taste so well that it will scan the latest releases we know you’ll like, or take you back to that nostalgic playlist you had on repeat last year.”
I will say this: AI tech like this provides context, something that’s missing from pure music streams. Without someone telling you something about the song, the album, the sound, and the scene, music is reduced to little more than organized, pleasant-sounding noise. But Spotify says “when listeners hear that additional audio context alongside their music recommendations, they’re more willing to try something new and listen to a song they may have otherwise skipped.”
And it’s not just the fake radio personality that bothers me. Dig a little deeper into what Spotify is doing with AI and you’ll see that there’s a push to eliminate human curators entirely. Spotify has many people on staff and would love to reduce its headcount in an attempt to become profitable. And if the human element is removed, what’s to say that the algorithm will be tweaked to provide certain promoted songs/artists over others?
Amazon Music is in the game as well. Last week, the company announced a partnership with an AI company called Endel with a plan to create AI-generated (i.e. no human intervention) playlists for wellness and sleep. And YouTube also announced a radio-like feature called Radio Builder that doesn’t feature any AI voices, but it does push algorithm tech down the field. In every case, this tech is only going to get better, smarter, and more sophisticated.
And then there’s RadioGPT, launched by an AI company called Futuri. I quote from their literature.
RadioGPT™ uses TopicPulse technology, which scans Facebook, Twitter, Instagram, and 250k+ other sources of news and information, to identify which topics are trending in a local market. Then, using GPT-3 technology, RadioGPT™ creates a script for on-air use, and AI voices turn that script into compelling audio.
Stations can select from a variety of AI voices for single-, duo-, or trio-hosted shows, or train the AI with their existing personalities’ voices. Programming is available for individual dayparts, or Futuri’s RadioGPT™ can power the entire station. RadioGPT™ is available for all formats in a white-labeled fashion.
Eek. Now I’m afraid. Very, very afraid.
Being an engaging, effective, and entertaining radio personality is hard work. Becoming a proper communicator and trusted companion is an art. If it was easy, then everyone would be brilliant at it. Obviously, that’s not the case.
Do we really want to turn every bit of entertainment over to machines? Don’t we look to the various forms of media for companionship and the human touch? I maintain that no radio personality really knows who they are on the air unless they’ve done it for five years full-time with proper coaching and feedback. (For more, check out this video from talent coach Valerie Geller.) That means it takes time to create an engaging talent who becomes a friend.
I love radio so much that I’ve stuck with it for over 40 years. I’d hate to see it be turned over to an army of virtual T-1000s with sexy baritone voices.
—
Alan Cross is a broadcaster with Q107 and 102.1 the Edge and a commentator for Global News.
Subscribe to Alan’s Ongoing History of New Music Podcast now on Apple Podcast or Google Play
Comments