It has been reported that OpenAI has developed a text detection engine using ChatGPT. However, the company is deliberating whether to make it accessible to the public.
According to TechSpot, OpenAI has created an ‘invisible watermark’ tool capable of identifying text generated by ChatGPT. This tool works by establishing unique patterns in word choice, making it easily recognizable by another tool.
Although the tool was ready for release about a year ago, OpenAI is still assessing its potential impact before reaching a decision. One concern is the potential negative effects on certain user groups, such as the stigma for non-native English speakers who use AI for writing.
OpenAI has also recognized limitations in the tool’s ability to detect edited or translated text. Additionally, an internal survey revealed that 30% of ChatGPT users would reduce their usage if watermarking was implemented, which could potentially impact OpenAI’s revenue.
Nonetheless, the existence of this tool is positive news for educators and businesses seeking to prevent the misuse of ChatGPT for cheating or plagiarism. If eventually released, this tool could provide them with a means to easily identify AI-generated content and take appropriate action.
As of now, the final decision on the release of this tool is still pending. OpenAI has stated that they are exploring alternative options and will continue to carefully evaluate before making a decision.