Google Unveils New Tool To Detect AI-Generated Images via @sejournal, @MattGSouthern

As artificial intelligence (AI) systems create increasingly realistic synthetic imagery, Google has developed a new tool called SynthID to help identify computer-generated photos and artworks.

While initially available to select Google Cloud customers, this technology represents a step toward identifying AI-generated content.

SynthID embeds imperceptible digital watermarks into AI-generated images, allowing them to be detected even after modifications like cropping or color changes.

Using Imagen, a new text-to-image model, Google is testing SynthID with select Google Cloud customers.

SynthID: How It Works

Unlike visible watermarks commonly used today, SynthID’s digital watermark is woven directly into the pixel data.

The tool uses two AI models trained together – one for adding the imperceptible watermarks and another for identifying them.

Google optimized these models to embed watermarks that align with the original image content, maintaining visual quality while enabling detection.

In internal testing, SynthID accurately identified AI-generated images after heavy editing. It provides three confidence levels to indicate the likelihood an image contains the SynthID watermark.

While Google doesn’t promise infallibility against extreme image manipulations, SynthID provides a technical approach to utilizing AI-generated content responsibly.

The tool is expected to evolve alongside other AI models, extending its capabilities beyond image identification to audio, video, and text.

Implications For Businesses & Digital Marketers

Though this tool is in its infancy, the advancement of AI-generated image detection has several implications for businesses and marketers.

  • More accountability: Embedding imperceptible watermarks into AI-generated images provides a way to identify the AI system used to create it.
  • New considerations: Marketers must consider whether and how to identify AI-generated images they use for campaigns. Proper attribution may build trust with consumers.
  • New opportunities: Watermarking technology opens up creative possibilities for generating AI imagery at scale while maintaining attribution. Savvy marketers may find novel applications.

In short, SynthID could reshape the conversation around responsible AI use. It will be an intriguing technology to follow as the AI race heats up.


Featured Image: IB Photography/Shutterstock

Leave a Reply

Your email address will not be published. Required fields are marked *