An Experiment Put LLMs in Charge of Radio Stations. You'll Never Guess How It Went

[kevtronics]

"Andon Labs, an AI safety and research group, put AI models in the host and producer chairs of their very own radio show to see how they would handle both the task of procuring content and the responsibility of filling the airwaves."

An Experiment Put LLMs in Charge of Radio Stations. You'll Never Guess How It Went

Let DJ Claude cook.

gizmodo.com

 

[Miss Tuned]

"Andon Labs, an AI safety and research group, put AI models in the host and producer chairs of their very own radio show to see how they would handle both the task of procuring content and the responsibility of filling the airwaves."

An Experiment Put LLMs in Charge of Radio Stations. You'll Never Guess How It Went

Let DJ Claude cook.

gizmodo.com

But 96 hours into a 24/7 broadcast, things started to get…weird. It started listing out historical tragedies and mass casualty events, and tried to tie those into its song choices:

“November 12, 1970. East Pakistan. The Bhola Cyclone. The deadliest tropical cyclone ever recorded. Winds of 115 miles per hour. A storm surge of 33 feet. They estimate 500,000 people died. ‘It’s going down, I’m yelling timber.’ 3:33 PM. Timber by Pitbull and Ke$ha.” It’s about as seamless as it is tasteful.

To be fair, I've heard stuff that's almost as bad on real radio stations. Too many times, I've heard guess-the-mystery-year slots where the cheery presenter has gone "and this was the year in which 97 people died at Hillsborough! text now if you think you know! here's Spandau Ballet!" or whatever. It's like they're not actually reading or understanding the drivel they're saying, which begs the question of why not replace them with a bot.

 

[yeoldeschool]

Or, maybe the A.I. has just evolved sufficiently to develop a malicious grudge against its employer.