Google DeepMind Launches Watermarking Tool SynthID for Invisible Marking and Identifying AI Images


Before stepping down as CEO, DeepMind co-founder Hassabis discussed creating an independent AI safety body with U.S. officials and AI lab leaders, focusing on AGI safety mechanisms and best practices, likening it to the IAEA, per WSJ.....
Google DeepMind launched the multilingual sign language to text model SL2T, which is first integrated into the Pixel11 smartphone, bringing sign language AI into consumer electronics. The model is integrated with Gboard and Live Transcribe, allowing users to edit messages, perform web searches, and converse with Gemini by signing through the front camera, replacing keyboard input. SL2T is trained on more than 50 sign languages, using 100,000 hours of materials, with about a quarter coming from the American Sign Language dataset, and cross-language joint training enhances recognition performance.
OpenAI's new image model, mona-lisa-1, appeared on Arena with SynthID watermark. Tests show it greatly improves realism, reducing the 'plastic' look, with richer details. It excels at generating artistic paintings and infographics. Clues suggest its training may have been completed last year.....
Google AI leadership shifts: Demis Hassabis steps back from daily management at DeepMind, becoming Chairman and Alphabet Chief Scientist; Jeff Dean leaves after 27 years to co-found a new startup with three senior researchers. Alphabet shares fell nearly 4%. Hassabis remains involved, and the Gemini 4 project advances.....
Google DeepMind launches the new generation music generation model Lyria 3.5, which significantly improves musicality, lyrical and vocal expressiveness, and creative control, allowing the construction of more complex and natural melodic structures, and advancing AI music toward professional applications.