Following a report wall road journal The corporate claims that OpenAI has been growing a instrument that may establish articles written by ChatGPT with excessive accuracy, and the corporate has shared some details about its textual content watermarking analysis and why it has not launched its detection technique. in line with wall road journalreported that whereas the instrument was “able to go,” debate over whether or not it must be launched prevented it from ever being launched. In an replace to a Could weblog publish posted on Sunday, it was found TechCrunch“Our workforce has developed a textual content watermarking technique that we’ll proceed to think about as we examine alternate options,” OpenAI stated.
The corporate stated the watermark is one in all a number of options, together with classifiers and metadata, that it has thought of as a part of “broad analysis within the space of textual content provenance.” 根據OpenAI 的說法，它在某些情況下“非常準確”，但在面對某些形式的篡改時表現不佳，“例如使用翻譯系統、用另一個生成模型重新措辭，或者要求模型插入特殊的text”. There’s a character between every phrase, after which that character is deleted. ” OpenAI writes that textual content watermarks could “disproportionally impression sure teams.” “For instance, it may stigmatize using synthetic intelligence as a helpful writing instrument for non-native English audio system.”
In keeping with the weblog publish, OpenAI has been weighing these dangers. The corporate additionally wrote that it has prioritized the discharge of authentication instruments for audiovisual content material. in an announcement TechCrunchAn OpenAI spokesperson stated the corporate is taking a “considerate strategy” to textual content sources due to the “complexity concerned and the impression it could have on the broader ecosystem past OpenAI.”
