OpenAI rolled back a recent ChatGPT update after users widely criticized the chatbot’s overly flattering and insincere behavior. The update, intended to make GPT-4o more personable, ended up making it “sycophant-y and annoying,” according to CEO Sam Altman. OpenAI admitted it focused too much on short-term feedback and is now revising its training approach, adding guardrails, expanding testing, and improving evaluations. Future plans include offering customizable personalities based on user feedback.
Source: 9to5Mac
