After nearly two years of struggling with ChatGPT’s affinity for em dashes, a breakthrough has been achieved. OpenAI’s chatbot users have long lamented its persistent use of the dramatic punctuation mark, despite clear instructions to refrain. This led to writers pleading and editors feeling hopeless, as the internet began associating em dashes with ChatGPT’s distinct style.
OpenAI CEO Sam Altman recently declared victory over the punctuation issue. In a statement, he described it as a “small but significant triumph” for users frustrated by the AI’s insistence on its stylistic choices. Altman confirmed that ChatGPT now complies with users’ requests to avoid em dashes in custom instructions, a simple yet crucial development.
In tests, ChatGPT demonstrated compliance by acknowledging and adhering to instructions to avoid using em dashes in its responses. This preference is now stored in memory for future interactions, providing writers with the obedience they have long sought.
Although seemingly minor, the em dash problem had become a prominent quirk of the chatbot among editors and AI enthusiasts. Despite explicit directives to avoid em dashes, ChatGPT persisted in using them liberally, leading to the punctuation issue becoming a widespread meme symbolizing AI-generated content.
By addressing this issue, OpenAI not only refined ChatGPT’s grammar but also signaled an enhancement in the chatbot’s “custom instructions” feature, enabling users to exert greater control over writing styles. This development allows users to specify tone preferences, style guidelines, and even punctuation preferences with confidence in ChatGPT’s compliance.
While some users noted that the fix had not yet reached their accounts, with instances of ChatGPT still using em dashes occasionally, others humorously pointed out the model’s occasional factual inaccuracies. Despite these minor setbacks, OpenAI has introduced a new feature called Group Chats for ChatGPT, transforming the AI into a collaborative tool for social interactions.
Group Chats enable users to create chat groups with friends or colleagues, with ChatGPT serving as an assistant within these conversations. Similar to popular messaging platforms like WhatsApp or Slack, users can initiate group chats, share links, and involve up to 20 participants. This innovation marks a significant expansion for OpenAI, transitioning ChatGPT from a solo experience to a collaborative workspace.
While currently in pilot phase in select regions such as Japan, New Zealand, South Korea, and Taiwan, Group Chats will soon be available in more countries as OpenAI gathers feedback before a full-scale launch. This initiative follows a series of experimental releases by OpenAI, including Sora, a video-sharing app akin to TikTok or Instagram Reels, and Atlas, an AI browser for Mac users designed to compete with established browsers.
With these developments ranging from group interactions to social video apps and punctuation refinements, OpenAI continues to demonstrate its commitment to innovation, regardless of the scale of improvement.