ChatGPT ‘off the rails’ as AI starts ‘threatening users’ who worry chatbot is ‘sentient.’
Some users were left wondering whether the Chatbot had gained some sort of sentience as it warned the “AI is in the room” with its users. “Scary” lines of dialogue from the bot were found within streams of nonsense.
A flurry of announcements from OpenAI soon followed as they confirmed maintenance work was underway on the tech tool. Teams as of today are “continuing to monitor the situation”.
Screenshots of the out-there robot later circulated on Twitter/X, with one interaction showing the bot put together “the stages of ether in music time”, before repeating “Happy listening!” dozens of ftimes.
Another saw the bot flicker between Spanish and English responses as it promised to speak “in a more grounded lingua”, before slipping into complete Spanish. It kicked back into English to suggest a “grape-turn-tooth over a mind-ocean jello type?”
h/t Stephen Green