In the AI era, transparency is a topic of public discussion. To help users understand the basic principles and reasoning processes of AI, Google recently launched the new Gemini 2.0 Flash Thinking Experimental model. Google states that this model is suitable for multi-step reasoning, breaking down its output answers before providing conclusions and explaining how and why it arrived at those conclusions, as well as which thinking steps it followed. Similarly, DeepMind, Google’s AI research subsidiary, is also attempting to use tools such as SynthID to increase transparency for AI-generated content.
Google Photos will embed invisible watermarks on photos edited with its AI “Reimagine” feature for identification purposes.
DeepMind launched the SynthID watermarking tool as early as August 2023. To the human eye, these watermarks are quite imperceptible, but they can be easily detected using specialized AI recognition algorithms. So far, SynthID for images has been available via Vertex AI for text-to-image models such as Imagen 3 and Imagen 2, and it can also output images on ImageFX. Now the tool will begin adding image watermarks to a broader range of services.
Google announcesGoogle Photos will soon start using SynthID to add watermarks to images edited with the Magic Edit tool. Regardless, Google has already done this to some extent: photos you edit with tools like Magic Eraser, Magic Edit, and Zoom Enhance already contain AI markers in the metadata and info panels. Starting this week, Google Photos will add digital watermarks to images edited with Reimagine, a tool that can significantly alter image content.
Reimagine, launched alongside the Pixel 9, essentially lets you reimagine your photos. You can edit backgrounds with text prompts, add or remove elements, move objects, apply special effects, and more. This feature is part of Magic Edit. Photos modified using Reimagine will include a note in the About this image section.
Source: KOCPC Chinese