OpenAI’s ChatGPT, a renowned AI model, has recently made a notable advancement in its text generation capabilities by adhering to custom instructions regarding the use of em dashes. Em dashes, often considered a distinctive trait of AI-generated content, have been a subject of scrutiny in AI chatbot outputs, including ChatGPT. Users have frequently associated excessive em dash usage with AI writing, although humans can exhibit the same tendency.
Sam Altman, CEO of OpenAI, announced this achievement on X, stating, ‘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!’ This development, following the recent launch of OpenAI’s GPT-5.1 AI model, elicited a mixed response from users who have grappled with ensuring specific formatting preferences are met by the chatbot.
Altman’s public discourse often touches on the concepts of artificial general intelligence (AGI) and superintelligence, while fundraising for OpenAI. However, the recent milestone with ChatGPT highlights the ongoing challenges in achieving precise control over AI models, indicating that the road to AGI may be longer than anticipated within the industry.
Despite the ambitious goals set for AI advancement and the aspiration for advanced intelligence capabilities, the persistence of struggles in basic instruction-following underscores the complexity of developing reliable artificial intelligence technology.
Source: Ars Technica