DeepMind's CTO Explains Their Invisible "AI Watermark"
Google DeepMind's CTO discusses SynthID, a digital watermarking system embedded in all videos created or edited with their Omni tool. The watermark enables users to verify whether content is AI-generated through Gemini, Chrome, or Android's circle-to-search feature, addressing concerns about distinguishing real from synthetic media.
Summary
The speaker expresses concern about the increasing realism of AI-generated images and videos, noting that the inability to distinguish real content from AI-generated content could be damaging to society. In response, Koray Kavukcuoglu, CTO of Google DeepMind, explains that every video created or edited using Omni automatically includes SynthID, a digital watermarking system. This watermark is invisible but detectable through multiple user-friendly interfaces: users can verify AI-generated content by asking Gemini directly whether a video is AI-generated or not, or by using circle-to-search on Android phones and Chrome browsers. Kavukcuoglu emphasizes that transparency about content origin is crucial for maintaining trust and control. While he acknowledges that AI tools are helpful and beneficial, he stresses the importance of helping people understand the source of information they encounter. The approach reflects DeepMind's philosophy of building technology that makes people feel safe and informed about the media they consume.
About this episode
Did you know your phone can easily tell you which videos or images or AI? AI videos are getting so realistic to the point where it's bordering on scary. So I point blank asked the CTO of Google DeepMind at Google I/O this year: What is your company doing about this? His answer was pretty cutting edge: Deepmind build a system called SynthID, an invisible digital watermark baked right into AI-generated content. So if you’re ever watching a video and think, "Wait, is this real?", you can literally just use Circle to Search on your phone or drop the content into the Gemini app and ask, “Is this AI?” It will instantly scan the invisible watermark and tell you the truth. Send this to someone who NEEDS to know about this (or tag that one friend who always falls for fake AI videos...) #AI #Gemini #GoogleDeepMind #genAI #aiart
Key Insights
- Every video created or edited with Omni automatically includes SynthID digital watermarking, which is embedded invisibly but remains detectable for verification purposes
- Users can verify whether content is AI-generated by querying Gemini, using circle-to-search on Android or Chrome, eliminating the need for manual detection
- Kavukcuoglu argues that building technology where people feel safe and in control is as important as creating helpful AI tools, requiring transparency about information sources
Topics
Transcript
[0:00] So I actually like and use AI every day. But even I have to admit that AI images and videos are getting so realistic it's a little bit scary. Like I don't want to live in a world where I can never tell what's real or not anymore. I think that can actually be pretty damaging in a lot of ways. So I got to ask Koray Kavukcuoglu, the CTO of Google DeepMind, point-blank what they're doing about this massive issue. >> Every video that is that you create or edit with Omni has this SynthID system, this digital watermarking with that. If you are using an like Android phone or [0:31] Chrome browser, you can even do circle…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Matt Wolfe
AI News: Dots, GPT-6.1 Sol, Sonnet 5.5, Gemini 4, and everything you need to know
A comprehensive review of major AI announcements from the week, including OpenAI's new Dots agent, GPT-6.1 Soul model, Anthropic's Sonnet 5.5, and Google's Gemini 4 Argon, with analysis of pricing, capabilities, and competitive positioning across different models.
The Hands Down Best Coding Model Right Now
The speaker reviews Opus 5.5, claiming it's currently the best state-of-the-art coding model at a reasonable price point. They demonstrate its capabilities by running a 20-hour test where the AI built a nearly complete recreation of the Mega Bonk game, including accurate character tiers, sound design, and UI elements.
This AI Model Does’t Use Words?!
Typesafe AI released Jev, a new AI model that outputs structured decisions (choices, scores, or booleans) instead of generated text like traditional LLMs. This approach makes Jev dramatically cheaper and faster, with input costs at 4 cents per million tokens and free output tokens.
AI News: Opus 5.5, GPT-6 Sol, Jev, Muse and More!
This week in AI saw major announcements from Meta Connect (Muse agent, new AR/VR glasses), three new language models (Claude Opus 5.5, GPT-6 Soul, Grok 4.7), significant buzz around Typesafe AI's Jev decision model, and new features from YouTube, Microsoft, Google, and Spotify. Claude Opus 5.5 emerged as the new state-of-the-art model while Jev introduced a fundamentally different approach to AI outputs focused on decision-making rather than text generation.
5 Ways To Use ChatGPT's New Features
The transcript demonstrates five practical applications of ChatGPT's new features, including computer use for software automation, advanced image editing, cloud browser integration for account access, voice-powered live agents, and persistent session agents for long-term user interaction tracking.