We launched AI Overviews last year at I/O, and since then there’s been a profound shift in how people are using Google Search. People are coming to Google to ask more of their questions, including more complex, longer and multimodal questions.
AI in Search is making it easier to ask Google anything and get a helpful response, with links to the web. That’s why AI Overviews is one of the most successful launches in Search in the past decade. As people use AI Overviews, we see they’re happier with their results, and they search more often. In our biggest markets like the U.S. and India, AI Overviews is driving over 10% increase in usage of Google for the types of queries that show AI Overviews 1 . This means that once people use AI Overviews, they’re coming to do more of these types of queries, and what’s particularly exciting is how this growth increases over time. And we’re delivering this at the speed people expect of Google Search — AI Overviews delivers the fastest AI responses in the industry.
We’re continuing to advance Search with AI, and today at I/O, we showed the latest in how we’re building the future of Search, as we go beyond information to intelligence. Here’s a look at everything we announced. — Read More
Daily Archives: May 21, 2025
Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures
The rapid scaling of large language models (LLMs) has unveiled critical limitations in current hardware architectures, including constraints in memory capacity, computational efficiency, and interconnection bandwidth. DeepSeek-V3, trained on 2,048 NVIDIA H800 GPUs, demonstrates how hardware-aware model co-design can effectively address these challenges, enabling cost-efficient training and inference at scale. This paper presents an in-depth analysis of the DeepSeek-V3/R1 model architecture and its AI infrastructure, highlighting key innovations such as Multi-head Latent Attention (MLA) for enhanced memory efficiency, Mixture of Experts (MoE) architectures for optimized computation-communication trade-offs, FP8 mixed-precision training to unlock the full potential of hardware capabilities, and a Multi-Plane Network Topology to minimize cluster-level network overhead. Building on the hardware bottlenecks encountered during DeepSeek-V3’s development, we engage in a broader discussion with academic and industry peers on potential future hardware directions, including precise low-precision computation units, scale-up and scale-out convergence, and innovations in low-latency communication fabrics. These insights underscore the critical role of hardware and model co-design in meeting the escalating demands of AI workloads, offering a practical blueprint for innovation in next-generation AI systems. — Read More
Rick's Cafe AI 