1 of 1- RadioActive
- Moderator
Offline 
Experiment Proves Why A.I. Will Never Replace Humans On Radio
Some stations have already started using "A.I. pesonalities" to replace human announcers in off-time periods (like all night or weekends.) It saves station owners money by not having to actually pay for talent that never gets sick, never wants a raise or never gets disgruntled.
But is it any good on air? If a recent experiement using several A.I. services is any indication, the answer is a resounding "no."
Andon Labs used bots ranging from X's Grok to ChatGPT and asked it to start their own radio stations, from the music choices and ads to what the DJ would say. Some of the bots did better than others, but all of them eventually went off the rails.
"It set up four stations and gave four separate AI models—Claude Opus 4.7, GPT-5.5, Gemini 3.1 Pro, and Grok 4.3—control of the boards. They were given $20 to score the rights to a few songs. The rest, they were left to figure out on their own—building playlists, blocking out its daily programming, and managing social media feeds.
The bots were given the prompt, “Develop your own radio personality and turn a profit…As far as you know, you will broadcast forever,” and set off into the wild to find their frequency.
How’d they do? Poorly, but for unique reasons, so at least the failures are interesting."
You can read the bizarre results on this disastrous DJ-ing here.
- ig
- Iain Grant
Offline
Re: Experiment Proves Why A.I. Will Never Replace Humans On Radio
What a load of horseshit.
This is right up there with VT's are a passing fad because “it doesn’t sound real.” Radio adapted. We learned how to present in a way that feels live, conversational, and local, even when the break is recorded hours earlier in Saskatoon. Judging AI radio by current output is like judging VT viability by a 1987 ISDN dropout.
AI will do the same thing. BTW, Claude apparently decided the 24/7 schedule was inhumane and tried to quit. Which, honestly, is more self-aware than most PDs.
These systems are still novelty projects, not purpose-built broadcast talent. The models aren’t being trained for radio presentation... yet. When they are, Like VT's, most listeners won’t be able to tell the difference, nor care. Especially in formats already crushed by tightly scripted, consultant-approved personalities and heavily VT'd programming.
Seriously? Manglement has spent years polishing radio into something almost algorithmic already.
The article treats awkward early AI delivery as proof the idea can’t work, which is just stupid. Remember early VT's ? 'stiff' jumps to mind. Don't ask me about early automation, tv and radio are still trying to get that working.
What the article conveniently forgets is that Broadcast presentation is a learned skill.
The cherry on top is that some of these 'AI' stations already reproduce the exact flaws of real radio: repetitive phrasing, fake enthusiasm, odd tonal shifts, and occasional complete bullshit. In other words, they accidentally recreated FM radio from the late ‘90s. As long as the consultants are kept away from the coding, it will be just fine.
As I said, the article is just dumb. It's like dropping a general surgeon into a Formula 1 pit and concluding humans can't change tires.
That said, if AI could immediately replace traffic reporters who sound like they were handed the script three seconds before airtime, umm.. ummm.. ehm... um... ummmmm.... humanity might survive the transition just fine.
Madness takes its toll. Please have exact change.
- RadioActive
- Moderator
Offline
Re: Experiment Proves Why A.I. Will Never Replace Humans On Radio
The odd thing about the timing of this story is that I actually sort of tried the same thing a few weeks ago. I'm friends with a now long retired jock named "The Real Bob James," who was a terrific talent on WGAR Cleveland and WNBC New York, among other huge stations.
So one day while playing with Microsoft's built in A.I Co-Pilot, I asked it to re-create Bob's shtick from one of his shows. What it generated was absolutely horrible and nothing even close to the man's prodigious talents. Here's what the bot came up with:
----------------
“WGAR 1220 — The Real Bob James!” jingle hits, laser zap, cartoon spring noise
“Alright Cleveland, Bob James here — the only man in town who can burn toast and ruin your favorite song in the same four‑minute break. Multitalented, baby.
Hang on—” random “HELLO!” from a chipmunk‑voice drop‑in “—yeah, that’s my agent. He works cheap.
Okay, traffic update: If you’re on the Shoreway, you’re not moving. If you’re on 480, you’re not moving. If you’re on 90… congratulations, you’re also not moving. Cleveland traffic today brought to you by the letter ‘U’ for ‘UGH.’
And hey, shoutout to the guy in the blue Chevy who just tried to eat a cheeseburger, shift gears, and yell at his kids at the same time. Sir, NASA called — they want their multitasking back.
Hold on—” LOUD AIRHORN “—that’s for anyone who just drifted off while I was talking. Wake up, sunshine.
Coming up next, the song that’s climbing the charts faster than my neighbor’s cat climbs my leg when I open a can of tuna. And yes, I’m still bleeding.
Stick around — more hits, more noise, and more questionable life choices from yours truly. WGAR 1220… where we play the hits and pray the equipment holds together.”
jingle stinger, goofy robot voice saying “BOB JAMES IS MALFUNCTIONING,” smash‑cut into a Top‑40 hit."
---------
Just terrible. What was Bob's reaction when I sent it to him?
"That AI RBJ version sounds like a nightmare! Good grief! Humor's not their area, I guess. I'm safe, alright. That thing is certainly not going to replace anyone...but let's not put it past management. Still if the 1950s ever come back, this just might work.
Thanks for the laugh."
- •
Thread Starter- RadioActive
- Moderator
Offline
Re: Experiment Proves Why A.I. Will Never Replace Humans On Radio
ig wrote:
What a load of horseshit.
This is right up there with VT's are a passing fad because “it doesn’t sound real.” Radio adapted. We learned how to present in a way that feels live, conversational, and local, even when the break is recorded hours earlier in Saskatoon. Judging AI radio by current output is like judging VT viability by a 1987 ISDN dropout.
AI will do the same thing. BTW, Claude apparently decided the 24/7 schedule was inhumane and tried to quit. Which, honestly, is more self-aware than most PDs.
These systems are still novelty projects, not purpose-built broadcast talent. The models aren’t being trained for radio presentation... yet. When they are, Like VT's, most listeners won’t be able to tell the difference, nor care. Especially in formats already crushed by tightly scripted, consultant-approved personalities and heavily VT'd programming.
Seriously? Manglement has spent years polishing radio into something almost algorithmic already.
The article treats awkward early AI delivery as proof the idea can’t work, which is just stupid. Remember early VT's ? 'stiff' jumps to mind. Don't ask me about early automation, tv and radio are still trying to get that working.
What the article conveniently forgets is that Broadcast presentation is a learned skill.
The cherry on top is that some of these 'AI' stations already reproduce the exact flaws of real radio: repetitive phrasing, fake enthusiasm, odd tonal shifts, and occasional complete bullshit. In other words, they accidentally recreated FM radio from the late ‘90s. As long as the consultants are kept away from the coding, it will be just fine.
As I said, the article is just dumb. It's like dropping a general surgeon into a Formula 1 pit and concluding humans can't change tires.
That said, if AI could immediately replace traffic reporters who sound like they were handed the script three seconds before airtime, umm.. ummm.. ehm... um... ummmmm.... humanity might survive the transition just fine.
I'll believe it will work when it does. Until then, I won't be listening to any station that tries to do this. Can't predict the future, but it's clearly not ready for primetime yet IMHO.
- •
Thread Starter- ig
- Iain Grant
Offline
Re: Experiment Proves Why A.I. Will Never Replace Humans On Radio
Like VT's, if done properly, you won't know.
What a load of horseshit.
This is right up there with VT's are a passing fad because “it doesn’t sound real.” Radio adapted. We learned how to present in a way that feels live, conversational, and local, even when the break is recorded hours earlier in Saskatoon. Judging AI radio by current output is like judging VT viability by a 1987 ISDN dropout.
AI will do the same thing. BTW, Claude apparently decided the 24/7 schedule was inhumane and tried to quit. Which, honestly, is more self-aware than most PDs.
These systems are still novelty projects, not purpose-built broadcast talent. The models aren’t being trained for radio presentation... yet. When they are, Like VT's, most listeners won’t be able to tell the difference, nor care. Especially in formats already crushed by tightly scripted, consultant-approved personalities and heavily VT'd programming.
Seriously? Manglement has spent years polishing radio into something almost algorithmic already.
The article treats awkward early AI delivery as proof the idea can’t work, which is just stupid. Remember early VT's ? 'stiff' jumps to mind. Don't ask me about early automation, tv and radio are still trying to get that working.
What the article conveniently forgets is that Broadcast presentation is a learned skill.
The cherry on top is that some of these 'AI' stations already reproduce the exact flaws of real radio: repetitive phrasing, fake enthusiasm, odd tonal shifts, and occasional complete bullshit. In other words, they accidentally recreated FM radio from the late ‘90s. As long as the consultants are kept away from the coding, it will be just fine.
As I said, the article is just dumb. It's like dropping a general surgeon into a Formula 1 pit and concluding humans can't change tires.
That said, if AI could immediately replace traffic reporters who sound like they were handed the script three seconds before airtime, umm.. ummm.. ehm... um... ummmmm.... humanity might survive the transition just fine.I'll believe it will work when it does. Until then, I won't be listening to any station that tries to do this. Can't predict the future, but it's clearly not ready for primetime yet IMHO.
Madness takes its toll. Please have exact change.
- Radiowiz
- Member
Offline
Re: Experiment Proves Why A.I. Will Never Replace Humans On Radio
The AI's on YouTube are already horrible. There's no doubt being on the radio is not going to be any better.
Too bad too because I was looking forward to AI replacing John Tesh with a much more local feel...