research.
51 writings found
Latest Archives
Science One Framework: AI Research That Actually Verifies Its Claims
Google researchers tackle AI hallucinations in autonomous science with Chain-of-Evidence framework. Zero phantom references, fully reproducible results.
Science One Framework: Making AI Research Actually Verifiable
Google researchers tackle AI hallucinations in autonomous research with Chain-of-Evidence framework and verifiable evidence chains that eliminate phantom citations.
Transferring GPU Expertise Across Hardware with AI
How K-Search and structured translation layers enable automatic kernel optimization across CUDA, MLX, and beyond without expert teams.
Science One Framework: Making AI Research Actually Verifiable
Google Cloud researchers introduce Chain-of-Evidence to eliminate hallucinations in AI-generated scientific papers. A critical shift toward trustworthy autonomous research systems.
Transferring GPU Expertise Across Hardware: K-Search Brings CUDA Knowledge to Apple Silicon
How evolutionary search and structured translation layers let AI automatically port decades of CUDA kernel optimizations to Apple Silicon, reaching near-expert performance without manual rewrites.
Gemini Robotics ER 2: The High-Level Brain Robots Actually Need
Google's new embodied reasoning model enables real-time spatial reasoning, multi-step task planning, and multi-robot collaboration. What this means for physical AI development.
Google DeepMind commits $40M to accelerate US scientific discovery with AI
Google invests in frontier AI for science, deploying tools across 17 DOE labs and achieving 8x speedups in autonomous materials research.
ABBEL: How LLMs Learn to Remember What Matters
ABBEL introduces belief states to fix the context summarization bottleneck in long-horizon AI tasks. Here's why this matters for coding assistants and beyond.
SymptomAI Beats Clinicians at Diagnosis: What This Means for AI in Healthcare
Google Research's SymptomAI outperforms real doctors in differential diagnosis. Here's what developers need to know about conversational AI at clinical scale.
Google DeepMind's $40M Genesis Mission Bet on AI for Science
Google commits $40M in AI tokens to accelerate scientific discovery at DOE labs. What this means for the future of AI-assisted research.
The Era of Free Intelligence Demands New Data Architecture
As AI inference costs collapse, data systems must evolve to handle agentic workloads. We're entering an age where intelligence is abundant—but infrastructure isn't ready.
Google's Flash Models Show How Efficiency Beats Raw Power in AI
Gemini 3.6 Flash and 3.5 Flash-Lite prove that token efficiency and lower latency are what production AI agents actually need. Here's what it means for developers.
AI's Role in Biosecurity: A Developer's Guide to Responsible Deployment
Google DeepMind and Isomorphic Labs outline how frontier AI models can prevent biosecurity threats while enabling rapid pandemic response and drug discovery.
Why Diffusion Models Aren't Just Memorizing Your Training Data
New research reveals diffusion model creativity stems from neural network regularization creating interpolation zones between training samples, not from memorization.
How Near-Free AI Intelligence Rewrites Data Systems
AI inference costs have fallen 50x per year. Here's what that means for how we build, query, and trust data systems going forward.