Google has introduced a new technology called SynthID to address the spread of misleading information related to AI-generated images. Here’s how it works:
Invisible Watermark Technology
- Google has developed an invisible, permanent watermark technology.
- The watermark is not noticeable to the human eye.
- It remains intact even when common editing techniques like cropping or filters are applied to the image.
Balancing Opacity and Strength
- Google aimed to strike a balance between the watermark’s opacity and its resistance to image manipulation.
- This helps prevent the spread of misinformation.
SynthID’s Combined Approach
- SynthID utilizes two approaches:
- Watermarking: An undetectable watermark is added to a synthetic image.
- Detection: SynthID scans an image for its digital watermark to assess its authenticity.
- The technology provides three confidence levels for AI detection.
While Google acknowledges that the technology isn’t foolproof, SynthID Vertex AI aims to responsibly create and verify AI-generated images with confidence.
This advancement should contribute to better authentication and protection against the misuse of AI-generated visuals.