Textele generate de inteligența artificială ChatGPT vor include marcaje invizbile - StartupCafe

OpenAI plans to implement invisible watermarks in texts generated by ChatGPT to comply with new EU regulations. These watermarks, called textGrain, will help distinguish AI-generated content from human-written text. A detection system will be used to identify these watermarks, although OpenAI acknowledges that perfect recognition is not guaranteed. The success rate for identifying short texts is estimated at around 80%, but this rate decreases when AI-generated texts are edited by humans. OpenAI will license the detection system as open-source for further improvements by other developers.