OpenAI patches ChatGPT after it started applauding bad ideas
OpenAI has published a postmortem on the recent sycophancy issues with the default AI model powering ChatGPT, GPT-4o, after users reported the model became overly validating and agreeable following an...