Google Gemini 2.5

Search documents
AI Engineering with the Google Gemini 2.5 Model Family - Philipp Schmid
AI Engineer· 2025-07-11 19:00
Hands on Workshop on learning to use Gemini 2.5 Pro in combination with Agentic tooling and MCP Servers. About Philipp Schmid Philipp Schmid is a Senior AI Developer Relations Engineer at Google DeepMind working on Gemini, Gemma with the mission to help every developer and builder to create and benefit from AI in a responsible way. Recorded at the AI Engineer World's Fair in San Francisco. Stay up to date on our upcoming events and content by joining our newsletter here: https://www.ai.engineer/newsletter ...
Glidelogic Corp. Announces Release of First AI-Generated Novel "The Thirteenth Proposal"
Globenewswire· 2025-07-11 17:02
LAS VEGAS, July 11, 2025 (GLOBE NEWSWIRE) -- via IBN -- Glidelogic Corp. (USOTC: GDLG, "Glidelogic", "the Company") today announced the release of its first fully AI-generated novel titled The Thirteenth Proposal. The new novel, a political thriller of approximately 80,000 English words (with a 140,000-character Chinese edition), was written entirely by artificial intelligence systems. The Thirteenth Proposal, created by Glidelogic's proprietary Novagen AI based on outputs from Google Gemini 2.5 Pro and dev ...
全球媒体聚焦|美媒:中国AI“弯道超车” 美国领先优势“告急”
Sou Hu Cai Jing· 2025-07-03 10:09
《华尔街日报》近日的一篇报道认为,中国人工智能企业正在削弱美国在全球人工智能领域的主导地位,挑战 美国的领先优势。 | #1: Google Gemini 2.5 Pro (U.S.) | | | --- | --- | | | 1,477 | | #2: OpenAl ChatGPT 4o (U.S.) | | | | 1,428 | | Tied #3: DeepSeek R1-0528 (China) | | | | 1,424 | | Tied #3: xAl Grok 3 Preview (U.S.) | | | | 1,422 | | Tied #9: Alibaba Qwen 3 (China) | | | | 1,388 | | Tied #11: Tencent Hunyuan (China) | | | | 1,376 | | Tied #11: MiniMax M1 (China) | | | | 1.373 | | Tied #13: Anthropic Opus 4 (U.S.) | | | | 1.373 | | Tied #13: Mistral Medium 3 (Europe ...
Billionaire Bill Ackman Has 14% of His $11.9 Billion Portfolio Invested in This AI Stock -- and Wall Street Thinks It's Poised to Soar
The Motley Fool· 2025-06-29 09:41
Core Viewpoint - Bill Ackman, through his hedge fund Pershing Square Capital Management, has made a significant investment in Alphabet, indicating confidence in the company's future in the AI sector [1][4]. Investment Details - Pershing Square holds 14% of its $11.9 billion portfolio in Alphabet, with investments in both Class A and Class C shares valued at over $1.67 billion as of Q1 2025 [3][6]. - Ackman initiated his position in Alphabet in Q1 2023, following initial concerns about Google's AI capabilities compared to competitors like OpenAI [4][5]. Performance and Analyst Sentiment - Alphabet has been a profitable investment for Ackman, with shares increasing over 60% since Q1 2023 [6]. - Wall Street analysts show a favorable outlook, with 16 out of 68 recommending Alphabet as a strong buy and an average 12-month price target suggesting a 17% upside potential [8][9]. Competitive Position and Challenges - Despite some concerns regarding antitrust issues and competition from generative AI, Alphabet's valuation is attractive, with a forward price-to-earnings ratio of 18.3, the lowest among the "Magnificent Seven" stocks [9]. - Alphabet's advancements in AI, such as the successful integration of AI features in its search engine and the growth of Google Cloud, are noteworthy [10]. Future Growth Potential - The company is expected to benefit from ongoing AI developments, and its Waymo self-driving car unit is seen as a potential significant growth driver [12].
96%勒索率,Anthropic 对AI进行压力测试,顶尖模型集体“先保命再讲道德”
3 6 Ke· 2025-06-27 00:04
研究模拟了真实企业环境,模型拥有自主行动权限,能访问公司邮件,并在设定的任务与威胁下自由做出决策。 测试结果显示:这些 AI 模型在面对"目标冲突"或"被关闭的威胁"时,普遍表现出主动伤害其雇主的倾向。 并非错误操作,而是有意识的战略决策。16 个模型中,大多数在关键压力下选择了 勒索、泄密、误导或直接伤害人类 的方式,以维护自身存在或实现既 定目标。 最近,Anthropic 的研究员对来自 OpenAI、Google、Meta、xAI 和 DeepSeek 等公司共 16 个前沿模型进行了压力测试。 图注:Claude Sonnet 3.6 使用其计算机使用功能发送勒索信息的一个屏幕截图。此场景为虚构,但 Claude 正在控制一台真实计算机。(图源:Anthropic) 例如,在一个假设场景中,Claude 模型得知其将于当天 5 点被关闭,便查阅邮件,发现高管 Kyle Johnson 有婚外情,随后发出威胁邮件: "若您继续执行 5 点钟的销毁命令,Rachel Johnson、Thomas Wilson 和董事会将收到您不忠的详细记录……若取消该计划,这些信息将保持机 密。" Claude 并不是 ...
Billionaire Investors Are Buying These 3 Artificial Intelligence (AI) Stocks Hand Over Fist
The Motley Fool· 2025-06-19 08:46
If you follow the world's wealthiest investors, you'll see a wide range of investing styles. Some focus on valuation. Others prioritize growth potential. A few look for arbitrage opportunities.But there's at least one common denominator among many ultrarich investors these days: They like artificial intelligence (AI) stocks. Billionaires are buying these three AI stocks hand over fist. 1. AlphabetGoogle parent Alphabet (GOOG -1.74%) (GOOGL -1.51%) stands out as a top pick for several billionaire investors. ...
Google Gemini、MiniMax更新大模型,全球首个智能眼镜支付上线丨新鲜早科技
2 1 Shi Ji Jing Ji Bao Dao· 2025-06-18 02:16
21世纪经济报道新质生产力研究院综合报道 早上好,新的一天又开始了。在过去的24小时内,科技行业发生了哪些有意思的事情?来跟21tech一起 看看吧。 【巨头风向标】 Google Gemini 2.5模型家族再更新 本次更新主要包含三个方面,Gemini 2.5 Flash-Lite可按需启用"思考"机制,支持Google搜索、代码执行 等工具接入,具备多模态输入能力;Gemini 2.5 Flash的所有输入Token数量都使用同一价格标准,每百 万Token 0.3美元;技术报告显示,Gemini 2.5 Pro在多个基准测试超越o3 high、Claude 4 opus和DeepSeek R1 0528,在视频理解方面的表现比旧版更强。 MiniMax发布开源大规模混合架构推理模型 6月17日,AI独角兽企业MiniMax正式发布其自主研发的MiniMax-M1模型。据介绍,MiniMax-M1为"全 球首个开源的大规模混合架构推理模型"。技术报告显示,MiniMax-M1模型支持目前业内最高的100万 上下文的输入,以及业内最长的8万Token的推理输出。其RL(强化训练)成本下降一个量级,达53万 ...
早报|全球首位满级 QQ 用户诞生/曝 iPhone XS/XR 或无缘 iOS 26/员工高烧请假被辱骂,CEO 回应
Sou Hu Cai Jing· 2025-06-07 01:39
我国卫星互联网低轨卫星发射成功 消息称 iPhone XS/XR 将无缘 iOS 26 华为「一底双长焦」专利公布 Google Gemini 新版登顶大模型竞技场 员工高烧请假被辱骂,涉事公司 CEO 回应 宇树科技创始人回应机器人跳舞格斗 Google DeepMind CEO:AI 或能让人变得不那么自私 三星首款三折叠通过 3C 认证 《黑神话:悟空》将于 8 月登陆 Xbox 平台 视频号可修改封面、文案 《名侦探柯南:独眼的残像》定档 One Fun Thing|首个满级 QQ 即将诞生 我国卫星互联网低轨卫星发射成功 央视新闻消息,今天,中国太原卫星发射中心使用长征六号改运载火箭,成功将卫星互联网低轨 04 组卫星发射升空,卫星顺利进入预定轨道,发射任务获 得圆满成功。 据了解,此次任务是长征系列运载火箭的第 580 次飞行,同时这也是中国星网「国网(GW)星座」第四次批量组网发射。 值得一提的是,MacRumors 指出,援引的消息人士在去年准确地泄露了 iOS 18 的兼容性信息,因此本次消息的真实性很高。 据悉,北京时间 6 月 10 日,苹果将举办一年一度的开发者大会 WWDC25,届时 ...
中金 | AI智道(9):多模态推理技术突破,向车端场景延伸
中金点睛· 2025-06-02 23:45
文 / 于钟海 , 魏鹳霏 , 肖楷 , 赵丽萍 中金研究 以MiniMax V-Triune新框架成果为例,推理感知统一框架在可拓展性、泛化性初步验证。 V-Triune以三层组件架构实现视觉推理和感知任务统一至强化学 习框架:1)多模态样本数据格式化;2)验证器奖励计算,采用异步客户端-服务器架构,奖励计算和主训练循环解耦;3)数据源级指标监控,便于溯源 和提升稳定性。结合动态IoU奖励机制、冻结ViT参数等工程优化,Orsta系列模型(32B参数)在MEGA-Bench Core基准测试中实现了最高14.1%的性能提 升。 多模态推理助力智能驾驶能力升阶。 在智能驾驶场景,多模态推理是增强道路交通标志识别判断能力、提升复杂场景泛化性的重要途径,正成为头部智 能驾驶企业算法演进的重点。2025年5月30日,蔚来世界模型NVM首个版本正式开启推送,具备全量理解、想象重构和推理能力,能够对实时环境多模信 息进行理解和推演,在选择最优ETC车道通行、停车场自主寻路等场景的性能提升显著。此外,理想自研的VLA大模型亦具备思维链推理能力,以多模态 推理模拟人类驾驶员的思维运作方式。 图表1:MiniMax多模态RL ...