O Wall Street Journal reported that the OpenAI has a system for marking texts created by the ChatGPT with watermark and a tool to detect this watermark ready about a year ago. However, the company is divided internally over whether to launch it. On the one hand, it seems like the responsible thing to do; on the other, it could hurt your bottom line.
ADVERTISING
The watermark of OpenAI is described as an adjustment to the way the model predicts the most likely words and phrases that will follow those before, creating a detectable pattern. (This is a simplification, but you can check the more detailed explanation of Google over the text watermark Gemini for corn informações).
Offer any way to detect material written by artificial intelligence (AI) is a potential boon for teachers trying to dissuade students from turning in writing assignments to AI. O Wall Street Journal reports that the company found that the watermark did not affect the quality of the text produced by the chatbot. In a survey commissioned by the company, “people around the world supported the idea of an AI detection tool by a four-to-one margin,” the paper writes.
After the Journal published the story, the OpenAI confirmed that he has been working on marking texts with a watermark in a blog update, which was seen by TechCrunch. In it, the company says that its method is very accurate (“99,9% effective”, according to documents seen by the Journal) and resistant to “tampering, such as paraphrasing”. But he says that techniques such as rewriting with another model make “evasion by malicious actors trivial.” The company also says it is concerned about the stigmatization of the usefulness of AI tools for non-native speakers.
ADVERTISING
But it seems that the OpenAI is also concerned that the use of watermarking may drive searched users away from the ChatGPT, nearly 30 percent of whom apparently told the company that they usedarialess software if watermarking were implemented.
Despite this, some employees still feel that tagging is effective. In light of lingering user sentiment, however, the Journal says some have suggested trying methods that are “potentially less controversial among users but unproven.” In its blog update today, the company said it is “in the early stages” of exploring incorporating metadata. He says it is still “too early” to know how this will work, but that because it is cryptographically signed, there would be no false positives.
Read also
