Dwell 95.5, a radio station in Portland, Oregon, introduced Tuesday that its noon host Ashley Elzinga will broadcast a cloned model of her voice — aka “AI Ashley” — to listeners each day from 10 a.m. to three p.m. They’ll be utilizing Futuri Media’s “RadioGPT,” an AI-powered device that makes use of ChatGPT to generate a script primarily based on trending information and reads it with an artificial voice.
The radio station famous that AI Ashley isn’t absolutely changing “conventional Ashley,” which has been an enormous concern for lots of native DJs. Radio stations have lower down their broadcasts in recent times because of rising prices or in a shift to embrace tech like AI. In 2020, iHeart Media restructured its group, shedding a whole bunch of individuals nationwide as a result of it wished to spend money on synthetic intelligence.
Dwell 95.5’s mother or father firm, Alpha Media, assured us that Elzinga wouldn’t be shedding her job and would nonetheless be paid the identical wage.
“It’s a hybrid state of affairs the place we’ll have conventional Ashley on throughout some segments, and we’ll have AI Ashley on throughout different segments,” Phil Becker, Alpha Media EVP of Content material, defined to TechCrunch. “In an occasion the place AI Ashley could be broadcasting, the standard Ashley is likely to be doing one thing in the neighborhood, managing social posts or engaged on digital property or the opposite components that include the job.”
Becker additionally famous that Alpha Media isn’t utilizing RadioGPT to avoid wasting prices. It’s meant to be an environment friendly device for radio hosts to have of their toolset.
Listeners could also be cautious of AI DJs, since they’ve tuned into native radio personalities for years and don’t essentially need to take heed to a robotic-sounding voice on their method to work within the morning. That’s why Dwell 95.5 skilled the AI to sound like an present radio host that the viewers has already linked with.
As seen in a tweet the place Elzinga reveals off her AI DJ counterpart, the voice is just like the actual factor however very apparent that it’s AI. The artificial voice additionally introduces itself as AI Ashley, so listeners know that an AI is talking and never an actual human.
“One of many absolute most essential elements of that is that we’re clear with the listener. It’s not our intent to ever deceive anyone,” added Becker.
Twitter customers expressed their disappointment within the new choice. One consumer even mentioned the AI voice was disrespectful to the whole radio career.
Whereas Dwell 95.5 is among the first radio stations to make use of RadioGPT, it definitely received’t be the final. Alpha Media’s KUFO Freedom 970, one other Portland-based radio station, can even be utilizing the AI tech.
Alpha Media oversees greater than 200 radio stations within the U.S. The corporate claims to be the primary radio broadcasting firm on this planet to have an AI DJ.
One other Futuri associate is Rogers Sports activities & Media, which owns 55 stations and over 29 podcasts in Canada.
Picture Credit: RadioGPT
When Futuri launched the beta model of RadioGPT in February, individuals questioned the ramifications of the AI tech. One main concern was if there was a danger that AI DJs may unfold misinformation. Reality-checking isn’t ChatGPT’s robust go well with and it has been identified to “hallucinate,” that means it confidently offers customers with faux info.
Bing and Google each have AI chatbots that customers have criticized for making factual errors.
ChatGPT depends on the assistance of human moderation, so it’s essential that radio stations utilizing RadioGPT have actual individuals checking the generated content material.
Futuri’s tech scans over 250,000 information sources in addition to tales on Fb, Twitter and Instagram to establish which matters are trending regionally. RadioGPT will take these trending tales after which use GPT-3 to create a script. Earlier than the AI voices relay the data to listeners in that native space, Becker claims {that a} workforce of human moderators critiques and edits the content material earlier than it goes out on the air.
“You’ll nonetheless need to have the unique content material creators touching it, checking it, proving it, all of these kinds of issues,” Becker mentioned.
Plus, Futuri CEO and founder Daniel Anstandig claims that safeguards are in place to forestall hallucinatory content material from going out over the air.
“Our AI is programmed to cross-check and scale back the prevalence of speculative or inaccurate content material, as is usually known as ‘hallucination,’” Anstandig claims. RadioGPT additionally “confirms that any generated content material doesn’t include any offensive materials,” he added.