supersquirrel@lemmy.ca to TechTakes@awful.systemsEnglish · 5 days agoA teen tried to navigate Canadian mountains with Claude. His trek ended in an emergency rescuewww.theguardian.comexternal-linkmessage-square12linkfedilinkarrow-up1114arrow-down10
arrow-up1114arrow-down1external-linkA teen tried to navigate Canadian mountains with Claude. His trek ended in an emergency rescuewww.theguardian.comsupersquirrel@lemmy.ca to TechTakes@awful.systemsEnglish · 5 days agomessage-square12linkfedilink
minus-squaretardigrade@scribe.disroot.orglinkfedilinkEnglisharrow-up16·5 days agoI recently read about a similar story with Deepseek. Don’t understand why someone blindly trusts a LLM.
minus-squareMoonrise2473@feddit.itlinkfedilinkEnglisharrow-up2·4 days agoIt also happened with Gemini https://www.sfgate.com/bayarea/article/mount-shasta-rescue-ai-22414497.php It’s becoming the new “our latest model also “accidentally” hacked someone”
minus-squaretardigrade@scribe.disroot.orglinkfedilinkEnglisharrow-up3·4 days agoOne major point in this respect is that these things are mainly reported in the West about Western LLMs, while in China similar stories related to Chinese LLMs are censored to avoid negative reporting on the country.
I recently read about a similar story with Deepseek. Don’t understand why someone blindly trusts a LLM.
It also happened with Gemini https://www.sfgate.com/bayarea/article/mount-shasta-rescue-ai-22414497.php
It’s becoming the new “our latest model also “accidentally” hacked someone”
One major point in this respect is that these things are mainly reported in the West about Western LLMs, while in China similar stories related to Chinese LLMs are censored to avoid negative reporting on the country.