ChatGPT removes 'traces' of AI

OpenAI claims to have fixed a bug where ChatGPT repeatedly used hyphens, opening up more customization options for users to control the writing style as they wish.

OpenAI says its ChatGPT tool can now reduce or remove hyphens in text when requested by users, an issue that has been debated for months, as hyphens are often a hallmark of AI-generated text.

This method of recognition is increasingly used in checking homework, emails, online comments, customer service exchanges, and forum posts.

Many critics say that hyphens make text look “artificial” and suggest that the writer is relying on a chatbot. On the other hand, some argue that hyphens were already widely used in English before large language models became popular.

However, the fact that chatbots almost always use this type of punctuation makes the phrase “ChatGPT dash” a bit of a pain, even though it is not a reliable indicator of AI-generated text.

For a long time, users have complained that ChatGPT continues to insert hyphens even when they have asked the system to stop, an issue that has put OpenAI under a lot of pressure, especially as debates about AI text recognition have grown.

“If you tell ChatGPT not to use hyphens in its custom instructions, it will do so,” OpenAI CEO Sam Altman wrote on X, describing the update as a “small but welcome victory.”

As explained by OpenAI, the model refinement helps ChatGPT better comply with hyphen-free requests if users explicitly set them in their personalization settings. This means the chatbot won't automatically remove hyphens entirely, but users can control how often they appear.

The new update is OpenAI's attempt to meet growing expectations for ChatGPT's customization capabilities, as many users want AI writing to be closer to natural human writing.

Source: https://znews.vn/chatgpt-xoa-dau-vet-ai-post1603036.html