ChatGPT's new models are adding watermarks in generated text: Report
SMRTR summary
OpenAI's new GPT-3 and GPT-4 mini models are embedding unique character watermarks in generated text, discovered by Rumi Technologies. The watermarks use special Unicode characters, mainly Narrow No-Break Spaces, and appear only in longer outputs. This feature could help identify AI-generated content, though it can be easily removed once users are aware of it.
SMRTR provides this summary for quick context. The original article belongs to NewsBytes.
Read the original article