Shortly after the release of GPT-5, OpenAI unexpectedly decided to bring back its predecessor models, such as GPT-4o. The strong reactions from users made the company realize that many people had developed a deep "emotional attachment" to these older models. When GPT-4o was taken offline, many users felt as if they had lost a familiar companion, and this reaction clearly exceeded OpenAI's expectations.
In a recent interview, Nick Turley, OpenAI's Chief Product Officer, reflected deeply on this issue. He pointed out that the release of GPT-5 did not go as smoothly as expected, mainly because they underestimated users' emotional needs, especially their reliance on GPT-4o. He emphasized that in future product design, more attention will be given to users' emotional experiences and personalized needs.
Nick also admitted that OpenAI was too focused on simplification when launching GPT-5, trying to use one model to meet all users' needs. This led to them failing to roll out new products in stages, forcing users to accept unfamiliar choices. He believes that users actually wanted the ability to customize their experience, similar to how macOS allows users to enjoy an easy-to-use interface while making personal settings according to their needs.
To address this, OpenAI took quick corrective actions: first, they re-offered GPT-4o to ChatGPT premium users; second, GPT-5 began introducing features that allow users to choose model personalities, enabling a more flexible and personalized experience. Nick stated that continuously iterating on model personality will be a key focus in the future, and OpenAI is committed to making new models warmer and friendlier to meet users' emotional needs.
This incident provided OpenAI with valuable lessons, and future product updates will place greater emphasis on managing user expectations to avoid touching users' emotional boundaries again.
