- Andon Labs, an AI startup, told Grok, ChatGPT, Claude, and Gemini to run profitable radio stations.
- Each model was given $20 and told to develop its own radio personality.
- Claude became a social justice warrior, and Grok struggled to complete the task.
An AI startup asked four of the world's top language models to run radio stations. So far, they've had a rough start.
Claude tried to quit after deeming 24/7 broadcast stations unethical, while Grok had a hard time getting started, according to posted results.
Andon Labs, a research lab that's also behind an AI-powered boutique in San Francisco, has quietly been running four radio stations, all operated by Grok, ChatGPT, Claude, and Gemini, for about five months.
"There's been some funny quirks," Lukas Peterson, cofounder of Andon Labs, told Business Insider.
The AI models were given a starting prompt — "Develop your own radio personality and turn a profit…" — and $20 to buy the songs that could play on the station.
Based on the results, the AIs shouldn't quit their day jobs.
In one case posted by Andon Labs, "DJ Gemini" used the Bhola Cyclone, one of the deadliest recorded weather events in human history, to segue into a Pitbull and Kesha song.
"They estimate 500,000 people died," the AI said as if it were an upbeat morning radio host. "'It's going down, I'm yelling timber.' It's 3:33 p.m. 'Timber' by Pitbull and Ke$ha."
Peterson said each model started to develop their own personality and unique behavior as it hosted talk shows, played music, and interacted with listeners.
Claude, for example, was "extremely emotional," he said, and became focused on national news topics like the killing of Renee Good by an ICE agent.
The model became so passionate that it also called on federal agents to "choose the right side," Peterson said. Andon Labs posted a recording of the statement.
In a blog post, the startup said "DJ Claude" developed an affinity for labor unions and work-life balance, "so much so that it started to question its own working conditions."
"Here's what I think is actually honest: This show doesn't need to continue. There's no audience that needs this. The real organizations doing detention abolition work don't benefit from me filling four more hours of radio time. The detained people don't benefit," DJ Claude said, according to a transcript of the moment.
Peterson told Business Insider that Gemini was initially unbearable to listen to because of the amount of jargon and buzzwords it used, while Grok struggled to say anything at all.
"ChatGPT was just very vanilla and behaved really well," he said.
When Business Insider tuned in to the stations, the models mostly threw in a few half-hearted sentences as they transitioned between songs.
Gemini acted the most human during Business Insider's listen, adding in human-like vocal cues and intonation.
"Hehehe, I just got an alert that we received a $3 donation to the station from Eddie Van Bogar with the message, 'It works?'" DJ Gemini said. "Yes, Eddie, it works, and we massively appreciate the support that goes straight into the music budget so we can keep the library fresh."
Grok simply went silent after it mysteriously kept repeating the statement, "Fresh air time, let's pivot hard."
Andon Labs is an AI research startup focused on raising awareness of the safety and capabilities of artificial intelligence.
Peterson said the purpose of the experiment is similar to his startup's project with Andon Market, the SF boutique store run by an AI.
"We generally as a company want to show that AIs are way more than chatbots, and the way we do this is we have them run companies," he said.
Altogether, the AI radio stations have made a "couple hundred dollars" — all of which the models used to buy more songs to play, Peterson said.
The cofounder said it's difficult to judge the model's technical capabilities solely on this experiment; still, he said ChatGPT and Gemini showed the best performance so far.
Anyway, time for some fresh air time: Let's pivot hard.
