Identifying AI-generated images made possible: Collaboration between Google's DeepMind and Google Cloud yields watermarking tool

In an exciting partnership, Google DeepMind, renowned for its AI research, has teamed up with Google Cloud to introduce a cutting-edge tool named SynthID in its latest blog post. This tool is designed to watermark and identify AI-generated images, but with a unique twist: it specifically targets images created by Google's image-generating model.

The launch of SynthID is a significant stride in enhancing transparency and responsible use of AI-generated content.

SynthID: Adding unseen marks to AI-created images

SynthID, now available in beta for selected users of Vertex AI, Google's platform for building AI applications and models, serves as an advanced watermarking tool. This innovation involves embedding a digital watermark directly into the image pixels, rendering it nearly invisible to the human eye but detectable by algorithms.

Unlike traditional watermarks that are placed visibly on images, SynthID's watermark is hidden within the image data itself.

We’re excited to launch 𝗦𝘆𝗻𝘁𝗵𝗜𝗗 today with @GoogleCloud: a digital tool to watermark and identify AI-generated images. 🖼️

It will be available on Imagen, one of @Google’s latest text-to-image models.

Here’s how it works. 🧵 https://t.co/p8s5XK5Ekr #GoogleCloudNext pic.twitter.com/0hixO5ehJo

— Google DeepMind (@GoogleDeepMind) August 29, 2023

Addressing the need for transparency in AI-generated content

The motivation behind SynthID's creation is the evolving landscape of AI-generated content. While AI has immense creative potential, it also brings forth new challenges, such as the potential to propagate misleading or false information.

SynthID addresses this concern by allowing people to identify AI-generated content reliably. This step is crucial for enabling users to distinguish between human-created and AI-generated media, thereby aiding in the fight against misinformation.

“SynthID isn’t foolproof against extreme image manipulations, but it does provide a promising technical approach for empowering people and organisations to work with AI-generated content responsibly,”

Google DeepMind in its blog post

DeepMind's development of SynthID, in collaboration with Google Research, utilises two AI models. One model is dedicated to watermarking images, while the other specialises in identifying watermarks. These models underwent joint training on a diverse range of images to enhance their accuracy.

The path forward for SynthID & responsible AI usage

While SynthID isn't foolproof against extreme image manipulations, it presents a promising approach to fostering responsible interaction with AI-generated content. The tool's potential isn't limited to images alone; it could extend to other media types like audio, video, and text.

The rising pressure on tech companies to demarcate AI-generated works is evident, with regulatory bodies in China and voices in the U.S. Senate advocating for transparency measures.

Although watermarking techniques for generative art exist, SynthID introduces a unique way of embedding watermarks that survive alterations like resising, cropping, and colour changes.

While currently limited to Google's Imagen model, DeepMind is contemplating making SynthID accessible to third parties in the future. However, the adoption of such technology by open-source AI generators remains uncertain, adding an intriguing dimension to the story's unfolding.

In the broader context of the AI landscape, SynthID joins the ranks of other watermarking initiatives, like Microsoft's cryptographic methods and OpenAI's watermark placement in AI-generated content.

While a standardised approach to watermarking is still a work in progress, SynthID stands out as a pioneering step toward ensuring transparency and accountability in the realm of AI-generated creations.