OpenAI says it’s fixed ChatGPT’s em dash problem

OpenAI says ChatGPT will now ditch the em dashes in the event you inform it to. The telltale signal that supposedly indicators textual content written by AI has popped up in every single place in current months, together with at school papers, emails, feedback, customer support chats, LinkedIn posts, on-line boards, advert copy, and extra. The inclusion of the em sprint has led folks to criticize these writers for being lazy and turning to an AI chatbot to do their work.

In fact, many have additionally argued for the em sprint, saying it’s been part of their writing properly earlier than LLMs adopted the punctuation. Nevertheless, the truth that chatbots couldn’t appear to keep away from its use made the so-called “ChatGPT hyphen” a newly objectionable addition to any textual content, even when they weren’t a dependable sign of content material created by generative AI.

The issue had stumped OpenAI for a while, as ChatGPT customers have been unable to get the chatbot to cease utilizing the image, even after they particularly requested it to not.

Now, OpenAI CEO Sam Altman says the corporate has fastened the issue. In a publish on X, Altman writes, “In case you inform ChatGPT to not use em-dashes in your customized directions, it lastly does what it’s imagined to do,” calling the replace a “small-but-happy win.”

The corporate explains in a publish on Threads (the place it pressured ChatGPT to apologize for “ruining the em sprint”) that ChatGPT will probably be higher at not utilizing the em sprint in the event you instruct it to not by way of the customized directions in your personalization settings. Meaning it gained’t essentially get rid of the em sprint from its output by default, however you’ll no less than have extra management over the frequency of its look.