The owner of a shore-facing restaurant in South Wales potentially saved the lives of two swimmers who ventured out into unsafe waters — after consulting ChatGPT to find tide times.
As local news outlet WalesOnline reports, the swimmers attempted to wade back from Sully Island, a rocky outcropping thousands of feet offshore, shortly before the tide came back in. The restaurant owner, Gordon Hadfield, got out his megaphone and yelled at them to turn back, which was likely lifesaving advice.
The region has the second-highest tidal range in the world, at roughly 50 feet. In the summer, people are known to get caught out by rapidly rising tides. The narrow causeway to Sully Island can see tides rising at up to eight miles per hour.
It’s only the latest instance of outdoor enthusiasts landing themselves in grave danger thanks to AI chatbots. The tech has garnered a reputation for spitting out hallucinations, which in certain circumstances can have devastating consequences. While AI companies argue that the tech can rival the intelligence of “PhD-level” academics, tools like ChatGPT still struggle with the very basics and more often than not spit out garbled and completely made-up information.
“I made the mistake of using ChatGPT for research to see when the low tide was,” one of the swimmers told WalesOnline. “It said 9:30 am. So we were out on that side, and then as soon as you come back over, it was literally completely different.”
“A lesson learned for me,” he added.
Netizens were taken aback by the carelessness on display.
“I wouldn’t trust ChatGPT to explain how to run a bath!” one user wrote in a Bluesky post.
“What amazes me is there are plenty of resources to get tide times, they are even still printed on paper, so why use ChatGPT to give you a tide time?” another user wrote. “Seems absolutely insane to me as a swimmer.”
In a separate incident, two hikers near Vancouver, British Columbia, had to call in a rescue team earlier this year after being caught unaware that higher altitudes would still be snowy in the spring and not appropriate for flat-soled sneakers. They reportedly followed the advice of ChatGPT to plan their excursion.
Earlier this month, the BBC reported that AI was sending unsuspecting tourists to potentially dangerous geographic locations in search of nonexistent landmarks.
“It doesn’t know the difference between travel advice, directions or recipes,” Carnegie Mellon University distinguished professor in machine learning Rayid Ghani told the broadcaster at the time. “It just knows words. So, it keeps spitting out words that make whatever it’s telling you sound realistic, and that’s where a lot of the underlying issues come from.”
Fortunately, the latest incident didn’t end in any preventable deaths.
“It’s fantastic that the local community and people like Gordon Hadfield are knowledgeable of the dangers,” a spokesperson for the Royal National Lifeboat Institution told WalesOnline. “The volunteers and people like Gordon being vigilant and providing early warning has almost certainly saved lives.”
More on ChatGPT: AI Endangering Tourists by Sending Them to Nonexistent Landmarks in Hazardous Locations
The post Two People Almost Drown After Asking ChatGPT When Low Tide Was appeared first on Futurism.
