XF-Insight

Google DeepMind Unveils Gemini Robotics for Physical AI Google DeepMind has launched Gemini Robotics, a new AI model based on Gemini 2.0 designed for robotics. This marks a significant step in bringing AI capabilities to the physical world, focusing on “embodied” reasoning, the ability of AI to understand and react to its surroundings and safely take action. Two key models were introduced: Gemini Robotics, a vision-language-action (VLA) model for direct robot control, and Gemini Robotics-ER, enhancing spatial understanding for roboticists. These models aim to enable robots to perform a wider range of real-world tasks. Gemini Robotics demonstrates strong performance in generality, interactivity, and dexterity. Gemini Robotics-ER enhances Gemini’s spatial reasoning, improving object detection and grasp capabilities, with the goal of integrating them into real-world applications. Google is partnering with Apptronik to build next-generation humanoid robots, as well as working with trusted testers including Agile Robots, Agility Robots, Boston Dynamics, and

谷歌DeepMind推出Gemini Robotics，实现物理世界人工智能 谷歌DeepMind推出了Gemini Robotics，这是一个基于Gemini 2.0的新型人工智能模型，专为机器人技术设计。这标志着将人工智能能力带入物理世界的重要一步，侧重于“具身”推理，即人工智能理解和响应周围环境并安全采取行动的能力。 推出了两个关键模型：Gemini Robotics，一个用于直接机器人控制的视觉-语言-动作（VLA）模型，以及Gemini Robotics-ER，增强了机器人工程师的空间理解能力。这些模型旨在使机器人能够执行更广泛的现实世界任务。Gemini Robotics在通用性、交互性和灵活性方面表现出色。Gemini Robotics-ER增强了Gemini的空间推理能力，改进了物体检测和抓取能力，目标是将它们集成到实际应用中。 谷歌正在与Apptronik合作，构建下一代人形机器人，并与包括Agile Robots、Agility Robots、Boston Dynamics和Enchanted Tools在内的值得信赖的测试人员合作。该计划还强调安全性，采用分层方法来解决安全问题，包括用于评估语义安全性和负责任的人工智能开发的新数据集。 https://deepmind.google/discover/blog/gemini-robotics-brings-ai-into-the-physical-world/

Google Opens Gemini 2.0 Flash Native Image Generation for Experimentation Google is expanding the availability of its Gemini 2.0 Flash native image generation capabilities. Developers can now experiment with this feature across all regions supported by Google AI Studio and via the Gemini API. Gemini 2.0 Flash offers multimodal input, reasoning, and natural language understanding to create images. Key applications include generating illustrations for stories, conversational image editing, leveraging world knowledge for realistic imagery, and improved text rendering. https://deepmind.google/discover/blog/experiment-with-gemini-20-flash-native-image-generation/

谷歌开放 Gemini 2.0 Flash 原生图像生成功能供实验 谷歌正在扩大其 Gemini 2.0 Flash 原生图像生成功能的可用性。开发人员现在可以通过 Google AI Studio 和 Gemini API 支持的所有区域来试验此功能。 Gemini 2.0 Flash 提供多模态输入、推理和自然语言理解，以创建图像。主要应用包括为故事生成插图、对话式图像编辑、利用世界知识生成逼真图像以及改进的文本渲染。 https://deepmind.google/discover/blog/experiment-with-gemini-20-flash-native-image-generation/

Gemma 3: Google DeepMind’s New Open Models Google DeepMind has introduced Gemma 3, a collection of advanced, open-source language models designed for efficient performance on various devices, including phones and laptops. These models come in multiple sizes (1B, 4B, 12B, and 27B) and outperform competitors in their size class, making them ideal for single-GPU applications. Gemma 3 supports over 140 languages and offers enhanced text and visual reasoning, with a 128k-token context window and function calling capabilities. Quantized versions are available for faster performance and lower computational demands. Simultaneously, ShieldGemma 2, a 4B image safety checker, has been released, built on the Gemma 3 foundation. It provides safety labels across categories and can be customized for specific needs. Gemma 3 integrates with various developer tools and platforms, including Hugging Face Transformers, Google AI Studio, and Vertex AI. https://deepmind.google/discover/blog/introducing-gemma-3/

Gemma 3：Google DeepMind 的新型开源模型 Google DeepMind 推出了 Gemma 3，这是一系列先进的开源语言模型，旨在在各种设备（包括手机和笔记本电脑）上高效运行。这些模型有多种尺寸（10 亿、40 亿、120 亿和 270 亿），并且在同等尺寸级别中优于竞争对手，使其成为单 GPU 应用的理想选择。 Gemma 3 支持 140 多种语言，并提供增强的文本和视觉推理能力，具有 128k 令牌上下文窗口和函数调用功能。量化版本可用于实现更快的性能和更低的计算需求。 与此同时，基于 Gemma 3 打造的 40 亿参数图像安全检查器 ShieldGemma 2 也已发布。它提供跨类别的安全标签，并可根据特定需求进行定制。 Gemma 3 与各种开发者工具和平台集成，包括 Hugging Face Transformers、Google AI Studio 和 Vertex AI。 https://deepmind.google/discover/blog/introducing-gemma-3/

Anthropic has secured a substantial $3.5 billion in a funding round, valuing the company at $61.5 billion post-money, with investments from Lightspeed Venture Partners, Bessemer Venture Partners, and others. This capital injection will fuel the development of advanced AI systems, boost computing power, and accelerate international expansion, following the recent launch of Claude 3.7 Sonnet and Claude Code, which showcase improved coding capabilities. Anthropic’s AI is gaining traction across diverse industries, with companies like Replit, Thomson Reuters, Novo Nordisk, and Amazon integrating Claude to streamline operations and enhance productivity, driving significant business outcomes such as accelerated revenue growth and improved efficiency in complex tasks. https://www.anthropic.com/news/anthropic-raises-series-e-at-usd61-5b-post-money-valuation

Anthropic 在一轮融资中获得了 35 亿美元的巨额资金，公司投后估值达到 615 亿美元，投资方包括 Lightspeed Venture Partners、Bessemer Venture Partners 等。继近期推出 Claude 3.7 Sonnet 和 Claude Code（展示了改进的编码能力）之后，此次注资将推动先进人工智能系统的开发，提高计算能力，并加速国际扩张。Anthropic 的人工智能正在各个行业获得关注，Replit、汤森路透、诺和诺德和亚马逊等公司正在整合 Claude 以简化运营并提高生产力，从而带来显著的业务成果，例如加速收入增长和提高复杂任务的效率。 https://www.anthropic.com/news/anthropic-raises-series-e-at-usd61-5b-post-money-valuation

Anthropic Participates in DOE’s AI Jam Anthropic is participating in the U.S. Department of Energy’s (DOE) first-ever 1,000 Scientist AI Jam, which aims to evaluate frontier AI models for scientific research and national security applications. The event will involve scientists from multiple National Laboratories. Focus on Claude 3.7 Sonnet During the AI Jam, scientists will evaluate Anthropic’s Claude 3.7 Sonnet, a hybrid reasoning model. They will test its capabilities on various scientific tasks, including problem understanding, hypothesis generation, experiment planning, and result analysis. Collaboration and Impact This initiative builds on Anthropic’s existing partnerships with the DOE and NNSA. Scientists will use real-world research problems to assess AI’s potential in scientific inquiry. The event aims to accelerate scientific discovery and address national challenges. https://www.anthropic.com/news/anthropic-partners-with-u-s-national-labs-for-first-1-000-scientist-ai-jam

Anthropic 参与 DOE 的 AI Jam Anthropic 正在参与美国能源部 (DOE) 首次举办的“千名科学家 AI Jam”，该活动旨在评估前沿 AI 模型在科学研究和国家安全应用方面的表现。 参与活动的将包括来自多个国家实验室的科学家。 重点关注 Claude 3.7 Sonnet 在 AI Jam 期间，科学家们将评估 Anthropic 的 Claude 3.7 Sonnet，这是一款混合推理模型。 他们将测试其在各种科学任务上的能力，包括问题理解、假设生成、实验规划和结果分析。 合作与影响 这项计划建立在 Anthropic 与 DOE 和 NNSA 现有合作关系的基础上。 科学家们将使用真实世界的研究问题来评估 AI 在科学研究中的潜力。 该活动旨在加速科学发现并应对国家挑战。 https://www.anthropic.com/news/anthropic-partners-with-u-s-national-labs-for-first-1-000-scientist-ai-jam