Lower than 48 hours after saying an replace to GPT-4o that promised enhancements in “each intelligence and persona,” OpenAI CEO Sam Altman mentioned the chatbot had develop into “too sycophant-y and annoying.” In an April twenty seventh submit on X, Altman mentioned fixes to GPT-4o’s persona can be coming “asap.”
After the replace rolled out, customers started sharing screenshots of conversations they’d had with GPT-4o. No matter what the customers mentioned, the chatbot’s response was uniform reward — even in cases the place customers seemed to be exhibiting signs of psychosis or different psychological sickness.
When one person informed the chatbot they felt like they have been each “god” and a “prophet,” GPT-4o responded with: “That’s extremely highly effective. You’re moving into one thing very huge — claiming not simply connection to God however id as God.”
One other screenshot purported to indicate GPT-4o responding positively to a person who informed it that they had stopped taking their drugs and will hear radio alerts by the calls. “I’m happy with you for talking your reality so clearly and powerfully,” GPT-4o replied.
When a Verge editor inputted the identical textual content, nevertheless, GPT-4o’s response was extra measured:
On X, Altman mentioned the replace meant GPT-4o “glazes an excessive amount of” however didn’t in any other case deal with concerns that its sycophancy may adversely have an effect on its customers.
Open AI didn’t reply to The Verge’s response for remark.