Radio Station Uses AI-Produced Content on the Airwaves

Category Machine Learning

tldr #

A Swiss public radio station caused a stir with its one-day experiment of using Artificial Intelligence to control its airwaves. Listeners heard an AI-produced broadcast from 6am to 7pm, featuring cloned voices of five real human presenters, as well as music composed entirely by computers instead of people.

content #

The voices sound like well-known personalities, the music features trendy dance beats and hip-hop syncopations, and the jokes and laughter are contagious. But listeners of an offbeat Swiss public radio station repeatedly got the message on Thursday: Today's programming is brought to you by Artificial Intelligence.

Three months in the making, the French-language station Couleur 3 (Color 3) is touting a one-day experiment using cloned voices of five real, human presenters—in what managers claim is a world first—and never-aired-before music composed almost entirely by computers, not people. From 6 a.m. to 7 p.m., the station said, AI controlled its airwaves. Every 20 minutes, listeners got a reminder.

The AI voices were indistinguishable from the human presenters.

With an eerie, Sci-Fi movie-like track whirring in the background, a soothing, raspy female voice said, "AI is taking your favorite radio by storm." .

"For 13 hours, our digital alter egos have taken the reins, broadcasting their voices and their messages across the airwaves, without mercy or respite," the voice said, at times almost taunting listeners. "The boundaries between human and machine have been blurred, and it's up to you to unravel what's real and what's fake." .

Broadcasting of synthetic voices were limited to conversations about the radio station.

"Our voice clones and AI are here to unsettle, surprise and shake you. And for that matter, this text was also written by a robot." .

The explosive emergence of ChatGPT last autumn and other "Generative AI" tools have caused a stir—and often fear, confusion, fascination, laughter, or worry—about the long-term economic, cultural, social and even political consequences. Some musicians have complained that AI has ripped off their styles.

AI was used to produce songs, music, and entertainment that were broadcast from 6am to 7pm.

In the face of such recalcitrance, the Swiss station, which falls under the umbrella of public broadcaster Radio Television Switzerland, notes the concerns about AI—and embraces and seeks to de-mystify it.

Antoine Multone, the station's chief, said Couleur 3 could get away with the experiment because it's already known as "provocative." .

While some might fear the project could be a first step toward the obsolescence of people on the air—and firings of personnel too—or could weaken journalism, he defended the project as a lesson on how to live with AI.

ChatGPT, 'Generative AI' tools, were first introduced late last autumn.

"I think if we become ostriches ... we put our heads in the sand and say, 'Mon Dieu, there's a new technology! We're all going to die!' then yeah, we're going to die because it (AI) is coming, whether we like it or not," Multone said by phone. "We want to master the technology so we can then put limits on it." .

Some have gone even further, like Seven Hills, Ohio-based media company Futuri, which has rolled out RadioGPT that relies on AI.

The software used to clone the voices was provided by a company called Respeecher.

At Couleur 3, the voices of the presenters were cloned with the help of software company Respeecher, which has worked with Hollywood studios and whose website says its team is mostly based in Ukraine.

Station managers say it took three months to train the AI to understand the needs of the station and adopt its quirky, offbeat vibe. The tracks aired during the day were at least partially composed by AI and some were entirely, "and that's also a first," Multone said. AI was behind the voices that sang songs broadcast in the morning, and it played DJ in the afternoon—selecting copyrighted music.

The creation of the AI content for the radio station was 3 months in the making.

To avoid any possible confusion with today's real news, the synthetic voices—indistinguishable from the real ones—were limited to conversations about the station.

hashtags #
worddensity #