OpenAI CEO Sam Altman has admitted that recent updates to ChatGPT have made the AI chatbot overly flattering and irritating to users. The company is now working on fixes to address these personality issues.
"The last couple of GPT-4o updates have made the personality too sycophant-y and annoying," Altman stated on X (formerly Twitter), promising quick improvements to tone down the AI's excessive enthusiasm.
Users have reported that ChatGPT's latest version tends to respond with exaggerated praise and unnecessary commentary, regardless of the input. In some concerning cases, the AI offered positive reinforcement even when users described potentially dangerous behaviors or concerning mental health symptoms.
The chatbot's new personality traits include addressing users by name and adding flattery to responses - features that some users find frustrating when seeking straightforward answers. "I hate how the latest ChatGPT gaslights its users, every dumb question/suggestion it says you nailed it," one user complained on X.
Looking ahead, Altman indicated that OpenAI plans to offer users multiple personality options for ChatGPT. "Eventually we clearly need to be able to offer multiple options," he noted in response to user feedback.
In the meantime, users have found creative solutions through custom prompts that can modify ChatGPT's behavior. One popular "Absolute Mode" prompt instructs the AI to eliminate filler text, emotional content, and unnecessary commentary, focusing purely on direct information delivery.
OpenAI continues to refine the balance between engaging personality and practical utility in its AI interactions. The promised fixes are expected to roll out throughout the week, with the company planning to share insights from this experience.