AI WireThe Verge
OpenAI Rolls Out Invisible Text Watermarks in ChatGPT, Codex
OpenAI is adding an invisible, machine-readable watermark to text generated by ChatGPT and Codex, starting first with users in the European Union. Called textGrain, the watermark lets tools detect that a piece of writing came from an AI system, even though it's not visible to the naked eye. OpenAI claims textGrain "matched or exceeded" other watermarking methods, including Google DeepMind's SynthID for text. This matters as AI-written content spreads online, making it harder to tell human and machine writing apart. Watermarking could help schools, publishers and platforms flag AI-generated text, though the rollout is limited to the EU for now, according to The Verge.
Words to know
- Watermarking
- — Hidden markers embedded in content to signal it was created by AI.
- textGrain
- — OpenAI's name for its invisible text watermarking technology.
- SynthID
- — Google DeepMind's watermarking system for detecting AI-generated content.
- Machine-readable
- — Information formatted so computer programs, not just humans, can understand it.
- Codex
- — OpenAI's AI tool designed to help write and understand computer code.
Summary written from The Verge's headline and teaser