Thanks to the report from The Wall Street JournalMicrosoft-backed OpenAI has confirmed that it has a tool that can watermark ChatGPT-generated text. According to OpenAI, the system has been ready for over a year, but there has been debate over whether to release it.
OpenAI Announces New Voice AI for ChatGPT (No Scarlett Johansson Mode)
The admission came in the form of an update to OpenAI’s May blog post, “Understanding the source of what you see and hear online.” The Wall Street Journal The system reportedly works by modifying the way ChatGPT generates words, so the changes made by the system leave a pattern or watermark that can be detected by anti-fraud tools.
Obtained internal documents The Wall Street Journal It shows that the watermarking system is 99.9 percent effective “when enough new texts are created by ChatGPT.”
Mashable Lightspeed
The tool would be a godsend for teachers and educators across the country who are plagued by students using ChatGPT to cheat on assignments and essays, and despite the tool’s internal testing showing a high 99 percent effectiveness, OpenAI has been reluctant to release it.
The reason behind this is The Wall Street Journal According to a company spokesperson, OpenAI believes that “the tool may have a disproportionate impact on groups such as non-native English speakers.” But another major reason is that an April 2023 survey revealed that 30% of ChatGPT respondents said they would stop using the generative AI if watermarking was introduced.
Internal discussions about releasing the tool have been ongoing at OpenAI for two years, a company official said. The Wall Street Journal“Just press the button.”
