OpenAI is updating its latest model, GPT-5, to be warmer and friendlier after a rocky launch that some users found disappointing compared to GPT-4o. The company aims to address complaints through subtle changes intended to make the model more approachable. New conversational touches like 'Good question' and 'Great start' are part of the update, and internal tests have shown there is no rise in sycophancy. Executives acknowledged the issues during a recent dinner with journalists while attempting to shift focus to future plans beyond GPT-5.
OpenAI is updating GPT-5 to be warmer and friendlier, addressing user complaints about its initial launch being bumpier than expected.
The update includes subtle changes aimed at making GPT-5 more approachable, incorporating small, genuine conversational touches like 'Good question'.
Internal tests indicate that there will be no increase in sycophancy compared to the previous version of GPT-5.
Despite executives' focus on future plans beyond GPT-5, the rocky launch and user feedback remained a significant concern during the recent dinner.
Collection
[
|
...
]