Watermark coming for ChatGPT
According to reports, the relevant artificial intelligence watermark has actually been ready for a year, but has not been brought to light until now. OpenAI is “This is something we continue to consider as our teams explore alternatives. developed text watermarking method“ said. The company states that they are working on many solutions, including classifiers and metadata, and with the method they developed. very accurate results (99.9 percent) states that they achieved.
However, this performance point cannot be reached always and in every situation. Because using translation systems on the text, rephrasing it with another generative model, or asking the model to add a special character between each word and then delete this character, etc. Performance decreases as a result of certain tampers. OpenAI also explained why it hasn’t made the watermark tool available until now while trying to improve performance here. Company, text watermark as it may disproportionately affect some groups stated that they were concerned.
According to the blog post, OpenAI is assessing these risks. The company also says it will focus on authentication tools for audiovisual content. Once the development for the text watermark is completed, the tool can be put into service. However, a timetable for this has not yet been determined. The company is assessing the risk of “disproportionately impacting some groups” as well as possibly losing revenue.
Source
https://www.engadget.com/openai-confirms-its-looking-into-text-watermarking-for-chatgpt-that-could-expose-cheating-students-223920531.html
https://openai.com/index/understanding-the-source-of-what-we-see-and-hear-online/
This news our mobile application Download using
You can read it whenever you want (even offline):
Source link: https://www.donanimhaber.com/odevleri-chatgpt-ye-yaptiran-ogrenciler-icin-kotu-haber–180325