OpenAI has tackled one of ChatGPT’s most notable quirks: its frequent use of em dashes in writing. For nearly two years, users worldwide expressed frustration at the chatbot’s inability to avoid this punctuation, even with specific instructions. Writers found it distracting, editors deemed it exhausting, and it became a running joke online. This week, the company announced that the issue has been fixed. OpenAI CEO Sam Altman shared the news on X, calling it a “small-but-happy win” for those who depend on the model to adhere to precise stylistic guidelines. He stated, “Small-but-happy win: If you tell ChatGPT not to use em dashes in your custom instructions, it finally does what it’s supposed to do!” Initial tests support his assertion.
When prompted to write without the punctuation, ChatGPT reportedly replied, “Got it! I’ll avoid using em dashes in all responses.” This behavior persists across conversations, aided by improved custom instruction memory. This change alleviates a long-term annoyance for content creators and indicates that OpenAI is taking style control more seriously. However, the fix has not been universally applied; some users responding to Altman noted that they still encounter the occasional em dash. Others remarked that while the punctuation change is beneficial, larger issues persist, such as the AI’s tendency to present incorrect information. One user humorously noted that the system still claims Joe Biden is president, highlighting the model’s known propensity for inaccuracies.
In addition to addressing the em dash issue, OpenAI has launched a significant new feature: Group Chats for ChatGPT. This enhancement brings a social aspect to the platform, enabling users to create shared chat spaces akin to those on WhatsApp or Slack. Within these groups, ChatGPT can assist all participants, providing ideas, answering queries, and facilitating collaborative discussions. Up to twenty individuals can join a single group via an invite link. Currently, the feature is available only in select regions, including Japan, New Zealand, South Korea, and Taiwan. OpenAI indicates that this limited rollout will allow for early feedback before a broader global expansion. The new functionality is part of OpenAI’s larger strategy to enhance ChatGPT beyond individual text exchanges.
Recent additions include Sora, an AI-driven short-video creation app, and Atlas, a Mac browser designed to compete with established options like Chrome and Brave. From minor punctuation adjustments to significant new features, OpenAI is committed to refining both the finer points and the overarching vision for its leading AI, demonstrating that every improvement, no matter how small, is worthwhile.