Google

129 summarised stories about Google, each linking back to the original source. Browse all topics →

Wednesday, 22 April 2026

It's all about the angle: Your photos, re-composed

Google Research 2 months ago

Google Photos introduced an Auto frame feature using machine learning and generative AI to recompose photographs by interpreting images as 3D scenes and adjusting the camera perspective after the photo is taken. The system uses a two-stage process combining 3D scene estimation with generative inpainting to fill in previously hidden content while preserving original details. The feature is now available in Google Photos and automatically enhances portraits by adjusting camera viewpoint and correcting perspective distortion from wide-angle lenses.

3D Reconstruction AI Product Development Generative AI Google

Decoupled DiLoCo: A new frontier for resilient, distributed AI training

Google DeepMind 2 months ago

Google introduced Decoupled DiLoCo, a distributed training architecture that divides large language model training across separate compute clusters with asynchronous data flow instead of requiring tight synchronization across thousands of chips. The system trained a 12 billion parameter model across four U.S. regions using only 2-5 Gbps of bandwidth while achieving training more than 20 times faster than conventional synchronization methods. The approach enables AI training to tolerate hardware failures in isolated clusters without interrupting overall progress, and allows mixing different hardware generations in a single training run.

AI Safety Google Google DeepMind