AI Briefing#google-research
The latest AI and developer news about #google-research, with the original source and a short summary.
The latest AI and developer news about #google-research, with the original source and a short summary.
The system uses Gemini 4 models to solve open mathematical problems, rewrite C++ code in Rust, and optimize TPU kernels with verified results.
The model resolves slow-channel glare and write sensitivity bottlenecks to train 2.7x faster than Splash Attention at 32K context.
The system identifies borderline cases to help users refine instructions, reducing effort and stress compared to manual labeling.
Sleep improved the AIME-24 score by 5.4%p through offline integration.
The framework uses MoE expansion and RL-based distillation to consolidate short-term context into long-term parameters, achieving 79.2% accuracy on AIME 24/25.
The third lesson in the AI Quests series, 'Polar Peaks,' teaches students ages 11 to 14 data cleaning and model testing through neuroscience, with over 400,000 quests completed to date.
In a recent study, the AI system securely gathered and summarized patient information, allowing physicians to shift from verifying baseline facts to discussing treatment options.
The open-source framework improves agent performance on benchmarks like WorkBuddy-Bench by up to 11.5 points without changing model weights.
The event showcased breakthroughs including the AMIE medical AI study, Earth AI for public health, and Project Suncatcher.
Deep Review offers two free weekly document reviews and expands beyond academic papers to legal and financial domains.
Google teams have published 90 peer-reviewed papers with over 90,000 citations and provided over $60 million in funding to health research organizations in 2026.
Juniors' skills stagnate due to AI reliance, while seniors benefit from enhanced judgment.
It proposes a multi-layered defense strategy at the system, model, and user levels based on contextual security theory.
A study involving 2,000 mothers showed AI can interpret blind-sweep ultrasounds as accurately as trained sonographers.
Gboard has already adopted the new system, achieving faster training speeds and stronger privacy guarantees for next-word prediction models.
Google Research and UIUC unveiled R4T, which distills RL-trained query fan-out behaviors into a 53.9M parameter diffusion model, reducing search latency by 12–20x.
Google is developing an AI-based early warning system using the FireSat satellite constellation to detect wildfires globally every 20 minutes.
Google and NASA JPL released MAPL-EMIT, a deep learning model trained on synthetic data, which discovered approximately 1.5 times more methane plumes than existing methods.
Google and NASA have released MAPL-EMIT, an AI model that detects global methane emission sources using satellite data.
Google researchers released 'Procedural Graph', which structures the procedural knowledge of LLM agents into a graph and self-improves through failure cases.
Google Research has released TimesFM-3, a 330-million-parameter time-series foundation model that predicts multivariate data and covariates in a single forward pass.
Google Research and HHMI Janelia researchers have released the first complete brain map, using AI to fully map the 166,000 neuronal connections in the male fruit fly brain.
Out of 136 AI adoption cases released in the last 20 days, 10 key examples were selected and published, including hospital drone logistics, autonomous trucks, and flight path optimization.
Google has released TimesFM-3, a foundation model for multivariate time series forecasting based on 330 billion parameters.