Article complete

Google DeepMind Introduces SynthID to Watermark AI-Generated Images

The beta version of the watermarking tool is currently only available to Imagen users

A grid of butterflies in various colors and shades, showcasing a blend of vibrant hues and monochrome styles.

Key takeaways

  • Google DeepMind has launched SynthID, a watermarking tool for identifying AI-generated images.
  • The tool aims to enhance transparency and help users differentiate between AI and human-generated content.
  • SynthID will initially be available exclusively to users of Google's Imagen AI image generator.
  • The rise of AI generation tools has created challenges in distinguishing authentic content from fake.
  • This initiative reflects Google's commitment to responsible AI development amidst growing concerns over misinformation.

Takeaways drafted with AI from the article text.

Moving forward along the lines of its ‘Bold and Responsible’ approach, today Google DeepMind and Google Cloud have launched a beta version of SynthID, a tool for watermarking and identifying AI-generated images. While the company gave a sneak peek of the technology in Google I/O earlier this year in February, the tool finally has a name to it.

The rise of AI generation tools has made it impossible to go anywhere on the internet without encountering fake news, leaks and rumours. The content-generation technology has wreaked havoc in every corner from newsrooms to educational institutions. The struggle to differentiate between AI and human-generated content made it to headlines every second week of 2023. Google saw the uncharted territory and announced at the Google I/O significant steps to identify and contextualise AI content available on its Search. 

After a lengthy dialogue, the company’s responsible AI research arm has finally introduced a potential solution with SynthID. Measures like watermarking and implementing metadata will ensure better transparency, allow users to differentiate between AI-generated and authentic images, and protect copyright.

The tool will initially be available only to the users of Google’s AI image generator Imagen, which is hosted on Google Cloud’s ML platform Vertex. After generating images on Imagen, users will have the option to choose whether to add a watermark or not. Google DeepMind's VP of Research, Pushmeet Kohli, labelled their watermarking tool "experimental," sharing plans to assess its strengths and setbacks before wider adoption, according to MIT Technology Review. Kohli refrained from discussing potential expansion to non-Imagen visuals or integration into Google's AI image generation systems.

Published Updated

Spotted an error? Tell our editors. Read our editorial standards and corrections.

More in this story

DeepMind AI · 3 stories
  1. [Exclusive] Pushmeet Kohli On Solving Intelligence at DeepMind for Humanity & Science
  2. You are here: Google DeepMind Introduces SynthID to Watermark AI-Generated Images
  3. Is AlphaCode 2 a Q* Moment for Google?
About the author

Contributor

All stories
Belamy · Daily

The day’s AI news, before your first meeting.