Google's SynthID watermark is hard to break, but it doesn't solve AI misinformation
Summary
Google has developed a technology called SynthID that invisibly marks images and videos created or edited by artificial intelligence (AI). This watermark is designed to stay intact even when images are resized, compressed, or edited, making it easier to identify AI-generated content. Several companies like OpenAI, Runway, and Nvidia plan to use SynthID to label AI media.Key Facts
- AI can create images and videos much faster than humans; for example, it took humanity 149 years to create 1.5 billion images, while AI did the same in 18 months.
- Google announced that its AI tools have been used to make over 100 billion AI images and videos in a few years.
- SynthID is an invisible digital watermark embedded in the pixels or audio of AI-created content.
- The watermark is designed to survive typical edits like cropping, filtering, resizing, and compression.
- Another method to label AI content is metadata schemas like C2PA, but metadata can be easily removed by simple editing; SynthID watermarks are harder to eliminate.
- Google has worked to make SynthID strong against attacks and alterations, but the company has not shared many technical details publicly.
- Testing showed that even after heavy degradation, such as repeated compression and resizing, SynthID remains detectable.
- OpenAI, Runway, Nvidia, and others are starting to apply SynthID in their AI-generated media.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
