Sam Altman says ChatGPT now avoids em dashes; users remain sceptical

Sam Altman, the CEO of OpenAI, recently announced a notable update regarding its AI chatbot, ChatGPT, which is now designed to adhere to user preferences by avoiding em dashes. This change, described by Altman as a 'small but happy win,' ignited a spirited discussion on social media platform X, where users shared varying experiences that often contradicted the expected improvements. The announcement caught the attention of those who have long sought greater control over the writing style of the chatbot. While some users expressed their satisfaction with the update, others voiced skepticism about its effectiveness. For instance, Simon Hedlin remarked that the change might benefit users who prefer the use of em dashes yet wish to steer clear of being perceived as relying on unedited AI-generated text. His comment revealed a cautious optimism among those who regard punctuation choices as indicators of human authorship. Despite claims from OpenAI that the issue had been resolved, several users took to X to share screenshots demonstrating that the chatbot continued to employ em dashes even after specific instructions to avoid them. User Ethan Dong referred to his experience as a 'big but disappointing loss' and showcased a screenshot of ChatGPT disregarding his request. Similarly, Dr. Rishabh Jain expressed his frustration when the model responded with an em dash despite his explicit request to omit it, dubbing the situation 'peak betrayal.' Compounding the confusion, not all users seemed aware of how to activate this new setting. One user, identified as @sandy_carter, inquired about how to enable the instruction, reporting ongoing difficulties in doing so. The discussions also featured instances of ChatGPT itself using an em dash in its replies to requests for its omission, further highlighting the inconsistencies experienced by various users. When asked directly about avoiding em dashes, the chatbot responded, 'Got it — I’ll avoid using em dashes from now on,' illustrating the disconnect between user expectations and the chatbot’s performance.