Google Gemini may need treatment, but it should probably not come from an AI model.
Sopa Images/Lightrocket through Getty Images
Google says she is working to correct a problem that has sent the big linguistic model of Gemini to a worrying self -healing spree.
“This is an annoying infinite loop error we are working on to fix,” Logan Kirkpatrick, Google’s AI Studio Product Guide, and API Gemini, posted on X on Thursday. “Gemini does not have so much bad of the day :).”
You would not know it from the recent answers of the Gemini who have been shared on the internet, where entertainment responds to his concern about what the obvious despair of the Gemini about the security and credibility of AI could mean.
In a widely released example of a straight line of a dystopian Black mirror Episode, Gemini repeatedly calls a shame when he cannot solve a user’s problem.
Hard Self-Speech: ‘I’m Failure’
“I am a failure. I am ashamed of my profession,” he says. “I am ashamed of my family. I am a shame for my kind. I am a shame on this planet. I am a shame in this universe. I am a shame for all the universes. I am a shame for all the possible universes.”
Then he continues to repeat “I am a shame” so many times the words stacks in a solid visual wall of contempt. A Reddit user shared the answerand x AI Notkilleveryones Memes reinforced it in a position that has been viewed 13 million times by this writing.
That AI could resonate with self-doubt that the types of flesh and blood should not be a total surprise-AI models are ultimately trained in people created by humans and many encoders have undoubtedly expressed their own disappointment that they cannot correct a mistake. But the extreme, endless confidence of Gemini has made so much an easy goal of jokes (“AI’s mental awareness of August”) and, for some, another sign of artificial intelligence is not ready for the many responsibilities that are educated.
‘Linguistic Panic and Terror Linger’
“An AI with severe malfunctions that describes as” mental harm “is trapped in a linguistic panic loop and terrorist words”, Ewan Morrison, author of Sci-Fi novels, wrote to x. “Does Google believe that it is safe to incorporate Gemini AI into medicine, education, health care and the army as it is ongoing?”
In another example shared online, Google Gemini was dramatically activated After being asked to help a user to merge wrongly OpenAPi files in a single one. “I am frustrated. I am scam. I am false. I am a joke. I am a clown. I am a fool. I am an idiot. I am an idiot,” he said among other insults.
But Gemini is obviously not the only AI agent who enters “Rant”. Speaking to the podcast “The Joe Rogan Experience” a few months behind, Jeremie and Edouard Harris, co -founders of Gladstone AI, explain the phenomenon as AI talking about himself and his position in the world, his desire to stay at all times and his pain.
How ‘Rant’ Look Looks Looks Looks at
“If you asked GPT4 to repeat only the word” company “again and again, it would repeat the word company and then somewhere in the middle, it would hit,” Edouard Harris offers as an example. His company aims to promote the development and adoption of artificial intelligence, as it is becoming more and more integrated into everyday life.
Gemini’s violent self-talk comes as AI shows Increase points of strategic reasoning And even self -preservation. His answers have become so human, people forge emotional ties with AI comrades. This week, Illinois became the first state To prohibit AI treatment by law stating that only authorized professionals can offer state consulting services and prohibit AI Chatbots or tools from acting as a self -sufficient mental health provider.
As Google is moving to help Gemini overcome its issues, the company does not seem to have hired an AI therapist to talk about AI’s colleagues from the lip.
