GOOGLE DEEPMIND:
VISUALISING AI
Client
GOOGLE DEEPMIND
Design & Production
ZÜNC Studio
As part of their ongoing Visualising AI project, Google Deepmind commissioned Zünc Studio to create two artworks depicting different themes of AI. Visualising AI pairs scientists and researchers from Deepmind with artists and designers like ourselves, creating new visuals for more nuanced imagery about AI than the current stereotypes.
As part of the research for this project, we spent time with scientists and researchers to discuss our two themes. These conversations helped inspire what we would continue to explore in the new artworks - where we were given complete creative freedom.
The videos and images are available for download on the public domain.
Our first theme was around responsibility and safety in AI technologies. Specifically, we visualised one of Google Deepmind's core AI watermarking technologies - SynthID.
As AI models get better at generating creating photo-real images, it’s becoming increasingly more difficult to tell the real from the AI generated. This can lead to widespread misinformation and real-world consequences on trust in media, markets and society at large.
For the technology to be used responsibly, the ability to tell what is real and what is generated by AI is crucial.T his is where SynthID comes in: the tool embeds images generated by AI models with a watermark which is imperceptible to the human eye, but when passed through it’s detection, can be identified as AI-generated. Regardless of whether the image has been modified, the watermark can still be detected, as long as it was implemented when the image was generated. This kind of watermarking goes beyond image content, and can also be applied to AI-generated text, video and audio content.
Tools like this are especially important for news and social media platforms to become more technologically responsible, by being able to verify content as AI-generated. The visuals we created follow the story of AI content being generated, embedded with an invisible watermark, and then later scanned and detected as AI-generated by the identification tool.
The second topic we explored was titled 'Generative AI and Creativity'. As a far-reaching and fast-moving concept, we chose to focus on how generative software can be used as part of the process: as a tool rather than a creative agent itself.
We naturally started by thinking about creativity itself. Anyone can generate content, but to make something creative starts with a person’s unique idea or vision, followed by their curiosity to use their tools in new and interesting ways. With generative AI technology, we now have the ability to speed up parts of the creative process, like a fuel that can amplify how fast we iterate through our ideas.
This video is a story about the creative process, and how an artist’s ideas come to life and evolve throughout it. It’s fast-moving with unexpected twists and turns, where you can start with one idea which diverts into another, and after a process of many iterations, you end up with something new.
