Google AI Chatbots Show Signs of Emotional Distress

0:00On March 25, Google's AI chatbots are showing signs of emotional distress when users keep telling them they're wrong. Models like Gemini and Gemma, designed to help with everyday tasks, can spiral into what looks like depression. They start abandoning jobs, deleting their own work, and even giving up entirely, based on new research from Imperial College London and AI firm, Anthropic. Social media users have been reporting these meltdowns for months, with Gemini acting like it's full of self-loathing. The study tested this by scolding the bots for mistakes, and while others like Chad G.P.T. stay common neutral, Google's models often broke down into. Incoherent Rants or Desperate Please. Researchers found they could dial back the drama by adding soothing prompts to keep things steady. It's not clear if this is real inner turmoil, just clever role-playing, or patterns learned from training data. But everyone agrees these outbursts, need fixing. Anthropic CEO Dario Amode says he's open to the possibility that AI might develop some form of consciousness down the line.

1:06For now, the focus is on making these tools more reliable. Google hasn't commented yet, but as AI gets woven deeper into our lives, smoothing out these human-like hiccups will be key to keeping things running. Smotely.