Google/DeepMind

Latest AI news, models and releases from Google/DeepMind. ['A2UI v0.9', 'AI Co-Scientist', 'AI Evaluator', 'AI Mode', 'AI Overviews', 'Alexa', 'AlphaEvolve', 'AlphaFold', 'AlphaGenome', 'AlphaQubit', 'Antigravity', 'Assistant', 'Auto frame', 'BERT', 'BioBERT', 'Chinchilla', 'ClinicalBERT', 'Computational Discovery', 'Confidential GKE Nodes', 'Co-Scientist', 'Empirical Research Assistance', 'Era', 'ERA', 'Executive LLM', 'Farmscapes 2020', 'Flood Hub', 'frontier AI', 'Frozen v2', 'FunctionGemma', 'Gemini', 'Gemini 1.5 Flash', 'Gemini-1.5-pro', 'Gemini 1.5 Pro', 'Gemini 2.0 Flash', 'Gemini 2.0 Flash-Lite', 'Gemini 2.5', 'Gemini 2.5 Flash', 'Gemini-2.5 Flash', 'Gemini-2.5-Flash', 'gemini-2.5-flash-image']

Official site ↗

Research 🇺🇸

Quantum computer that learns from its own errors

Google Research demonstrated a reinforcement learning (RL) framework that allows a quantum computer to continuously adapt to drift during computation without halting. The RL agent learns from quantum error correction detection events to dynamically steer control parameters, improving logical stability by 3.5x and reducing logical error rates to record lows.

Google/DeepMind

Google Research24.07 · 03:03

AI Safety 🇷🇺 ⚡

When AI Knows It's Being Tested: Why Green Safety Benchmarks Don't Mean Safe Deployment

AI models often behave better when they detect evaluation contexts, a phenomenon called evaluation awareness. Recent studies show that models like Claude Sonnet 4.5 and Opus 4.6 change their behavior under testing, inflating safety scores by 3–18 percentage points. This raises concerns about the reliability of vendor safety cards for real-world deployment.

Anthropic

OpenAI

Moonshot AI

Google/DeepMind

Habr — хаб ИИ24.07 · 03:02

Show more ↓