OpenAI mulls watermarking ChatGPT generated text, but treads with caution

Posted under: AI technologies
Date: 2024-08-05
OpenAI mulls watermarking ChatGPT generated text, but treads with caution

OpenAI has developed a method to watermark text generated by ChatGPT, aiming to identify AI-produced content. This watermarking subtly alters word patterns to create detectable signatures without affecting quality. OpenAI is considering embedding metadata for more accurate detection. Concerns remain about potential stigma for AI use, especially for non-native English speakers, delaying the tool's release.