AI应用与客户服务 - CASE 作为欧洲领先的资产服务公司，利用 AI Agent 捕捉客户交流的真实含义 [1] - 传统 AI 转录遗漏了超过 60% 的客户会议关键信息 [2] - AI Agent 使用 Nemo Retriever 从多模态数据源提取上下文，并利用 Llama Neatron 创建包含完整细节和个人信息修订后的两个版本 [2] - 客户关系 Agent 可以检测情绪、分析异议并触发后续行动，以帮助领导者进行指导 [3] - 产品分析 Agent 使用修订后的数据和人口统计信息来识别 VIP 功能请求并生成 Jira 工单 [3] 技术与合作 - CASE 与 Vast Data 和 Nvidia 合作构建 AI Agent [1] - 借助 NVIDIA Nemo 数据飞轮和 Vast AI 操作系统，Agent 可以不断学习和改进 [4] - VAST 和 NVIDIA 帮助企业为每位员工构建定制 AI Agent [4] 数据处理与安全 - AI Agent 在使用数据的同时，保护访问控制 [3] - AI Agent 可以处理完整细节和个人信息修订后的数据，满足不同角色的需求 [2][3]

Company Vision & Technology - Moon Surgical aims to enhance surgery through robotic and digital solutions, envisioning a future where surgery is increasingly robotic [1] - Moon Surgical's Maestro system, developed in collaboration with NVIDIA, offers a unique surgical feel and experience applicable to a wide range of surgeries [2] - The Maestro system allows surgeons to retain control and potentially integrate AI and other features, which could revolutionize surgical training and practice [3] Innovation & Future of Surgery - Simulation, enhanced by AI, is crucial for surgical training, allowing for the simulation of accidents and variations in practice [3][4] - AI is essential for creating adaptive surgical systems that cater to individual surgeon's practices [4] - There is excitement within the operating room (OR) team to push the boundaries of minimally invasive surgery for the benefit of society and patients [4]

Overview of Nvidia's GDC 2025 - Nvidia is driving breakthroughs across industries with accelerated computing, addressing challenges beyond the capabilities of normal computers [1] - Nvidia's Inception program supports AI startups by providing technical training, VC network access, and product pricing benefits [6] AI Startup Innovations - Nayam Biologics uses genomics, chemistry, and AI to develop therapeutics for cancer, cardiovascular conditions, metabolic disorders, and neurodegenerative diseases [2] - Nayam Biologics' drug discovery platform accelerates traditional timelines by 80% using ambidia rabbits, cuda, bioimmo, and tensorati to screen millions of natural compounds [3] - No X enables designers and manufacturers to generate digital twins of fabrics, accelerating development and streamlining supply chains [4] - Paul innovation developed a collision warning system for Gausong light rail, using camera data and sensors for 3D object tracking [4] - AI robot's Ellis 4 model uses Nvidia Jetson Orin NX to reduce energy requirements and proprietary linear actuators to reduce robot manufacturing costs [5] - ERICO's platform uses jetpack with replicator and tensor RT to automate recycling facilities, decreasing sorting costs and maximizing the value of recovered materials [6] Impact and Future Directions - AI technology is poised to provide critical services in manufacturing, healthcare, and logistics [5] - Nvidia aims to push the frontiers of AI and create solutions that will shape the future of technology and society [7] - ERICO's platform helps countries reach recycling goals set by global sustainability initiatives, contributing to the circular economy [6]

Automotive AI Advancements - NVIDIA leads in physical AI, securing the End-to-End Autonomous Driving Challenge win for the second consecutive year at CVPR 2025 [1] - NVIDIA is the only entity to win a CVPR challenge for three consecutive years [1] - NVIDIA researchers are advancing physical AI through innovations in perception, planning, and data generation for automotive applications [1] Research & Development - NVIDIA Research had over 60 accepted papers at CVPR 2025, spanning automotive, healthcare, robotics, and more [1] - Three NVIDIA papers were nominated for the Best Paper Award: FoundationStereo, Zero-shot Monocular Scene Flow, and Difix3D+ [1] CVPR 2025 Resources - NVIDIA highlights its consecutive win for End-to-End Autonomous Driving Grand Challenge at CVPR [1] - NVIDIA showcases Neural Inverse and Forward Rendering with Video Diffusion Models [1]

[Music] Bringing a life-saving drug to market requires analyzing massive amounts of complex data. The pharmaceutical industry needs a faster, more automated way to extract meaning and act on it. Ivia is using Agentic AI to do exactly that.Training AI agents to navigate more than a million data streams for clinical, medical, and commercial professionals. Its healthcare grade AI platform combines a growing set of AI agents, each designed to streamline how insights turn into action. Built with NVIDIA Neotron m ...

Blackwell B200 Super Chip Manufacturing - Blackwell B200 超级芯片由 12 英寸晶圆上的 2000 亿个晶体管构建而成 [1] - 每个晶圆被划分为单独的 Blackwell 芯片，经过测试和分类，将好的芯片分离出来 [2] - 32 个 Blackwell 芯片和 128 个 HBM 堆栈通过芯片基板工艺连接到定制的硅中介层晶圆上 [2] - 组装完成后，经过烘烤、模塑和固化，形成 Blackwell B200 超级芯片，并在 125° 的烤箱中进行压力测试 [3] - Grace Blackwell PCB 上通过机器人昼夜不停地放置超过 10,000 个组件 [3] Interconnect and Communication - MVLink 开创性高速链路，可连接多个 GPU 并扩展为大型虚拟 GPU [5] - MVLink 交换机托盘由 MVLink 交换机芯片构建，提供每秒 14.4 万亿字节的全互连带宽 [5] - MVLink 主干形成定制的盲插背板，通过 5,000 根铜缆连接所有 72 个 Blackwell 或 144 个 GPU 芯片，形成一个巨大的 GPU，提供每秒 130 万亿字节的全互连带宽 [5] - Connect X7 Super Nix 用于实现横向扩展通信，Bluefield 3 DPU 用于卸载和加速网络、存储和安全任务 [4] System Integration and Scale - 总计 120 万个组件、200 万米铜缆和 130 万亿个晶体管，总重量接近 2 吨 [5] - 定制的液冷铜块用于将芯片保持在最佳温度 [4] - 所有部件集成到 GB200 计算托盘中，最终组装成机架规模的 AI 超级计算机 [4][5] Vision - Blackwell 不仅仅是一项技术奇迹，更是全球协作和创新的力量的证明，推动着塑造我们未来各地的发现和解决方案 [6]

This is how intelligence is made. A new kind of factory, generator of tokens, the building blocks of AI. tokens have opened a new frontier.The first step into an extraordinary world where endless possibilities are born. Tokens transform images into scientific data, charting alien atmospheres and guiding the explorers of tomorrow. They probe the earth's depths to seek out hidden danger.They turn potential into plenty [Music] and help us harvest our bounty. Tokens see disease before it takes hold. Cure with p ...

AI Native Wireless Networks Overview - NVIDIA AI Aerial is revolutionizing wireless systems by enabling neural networks to learn from data, adapt in real-time, and scale with complexity [1] - This approach boosts spectral efficiency, lowers costs, and improves connectivity in wireless networks [1] Key Steps & Technologies - Building AI native wireless networks involves three key steps: training, simulation, and deployment, all unified by NVIDIA AI Aerial [2] - NVIDIA uses Shauna and Aerial radio frameworks to train neural network models, replacing traditional 5G channel estimation algorithms [2] - The trained model is deployed on a live 5G network at NVIDIA using Aerial RAN computer [3] - NVIDIA provides the Shona research kit on NVIDIA Jetson, enabling developers to create a working AI model in a live 5G network with minimal code [5] Performance & Results - AI implementation nearly doubles the cell throughput [3] - AI achieves 100% higher throughput on a live 5G network [4] - The AI model continuously improves over time through a feedback loop between the physical and virtual worlds [4] Future Implications - Improved AI models will boost capacity, unlock new applications, increase efficiency, and extend coverage [5] - NVIDIA encourages participation in the 6G developer program to accelerate AI native wireless research [5]

AI Agents Overview - AI agents are digital assistants that use prompts to reason and break down problems into multi-step plans [1] - They utilize appropriate tools, collaborate with other agents, and leverage context from memory on NVIDIA accelerated systems [1] - The process begins with a simple prompt, exemplified by asking Perplexity to initiate a food truck business in Paris [1] Agent Collaboration and Functionality - Agents collaborate to address each step, utilizing various tools [2] - A market researcher analyzes trends and the competitive landscape through reviews and reports [2] - A concept designer explores local ingredients and proposes a menu with prep time estimates, researches pallets, and develops a brand identity [2] - A financial planner employs Monte Carlo simulations to project profitability and growth [3] - An operations planner creates a detailed launch timeline, covering equipment procurement and permit acquisition [3] - A marketing specialist develops a launch plan with a social media campaign and an interactive website featuring a map, menu, and online ordering [3] Proposal and Outcome - The collective work of each agent culminates in a final packaged proposal [3] - All of this originates from a single prompt [3]

Core Offering - DJX Cloud Leptin provides on-demand access to a global network of GPUs across clouds, regions, and partners [1] - Multicloud GPU clusters are managed through a single unified interface [1] - Connects developers to GPU compute, powering a virtual global AI factory [3] Key Features - Fast provisioning allows developers to quickly scale up the number of nodes without complex setups [2] - Pre-integrated tools and training-ready infrastructure enable developers to start training immediately [2] - Real-time progress monitoring provides insights into GPU performance, convergence, and throughput [2] - Supports deployment of NIM endpoints or models in multiple clouds or regions for fast distributed inference [2]