Prominent investor Chamath Palihapitiya, known as the “SPAC King,” took to X (formerly Twitter) on Thursday to highlight Microsoft Corp.’s (NASDAQ:MSFT) underperformance relative to other major technology companies since the launch of ChatGPT. Microsoft Trails Tech Peers Since ChatGPT Debut “Microsoft has been the worst-performing stock of the hyperscalers since the November 30, 2022, launch of ChatGPT and has significantly underperformed the Nasdaq during that period,” Palihapitiya wrote. He added, “Lo ...

在红包大战之外，大厂还在争夺终端入口。上海财经大学特聘教授胡延平认为，未来Agent（智能体）可能会接管手机、电脑等终端，实质是生成式AI技术 重构智能硬件与软件行业。2025年AI的范式、价值与能力已经得到充分确认，2026年将出现技术大投入、竞争大分化、市场大分流。 经济观察报 记者 钱玉娟 编者按：2025 年，科技互联网的齿轮转动得比以往任何时候都要快。在技术洗牌的前夜，没有人敢停下。 我们看到，大厂在算力和大模型应用里厮杀，电商在监管与合规的夹缝中重塑底线。做无人机的去卷扫地机器人，一批公司蜂拥去做智能眼镜，整个硬件圈 陷入了一种 " 无所不能 " 的集体幻觉与入口焦虑。 我们也看到，人形机器人刚踉踉跄跄地走出实验室，在车间干些零活，就迎来残酷的价格战；短剧告别了一夜暴富的粗放模式，漫剧直接降维打击，头部内 容公司既兴奋又不安。 这不再是写份 PPT 就能拿到千万融资的时代，技术、商业化场景和生态优势，丢了哪头都不行。活下去，并且要硬核地活下去，成了 2025 年 TMT 行业的共 识。 马年将至，腾讯、阿里巴巴和字节跳动等头部科技公司将AI竞争拉回了他们最熟悉的战场——流量战争。 腾讯元宝发放10 ...

Core Insights - The article highlights the impressive performance of Gemini 3 Deep Think, which has achieved significant milestones in various benchmark tests, showcasing its advanced reasoning capabilities and potential applications in scientific and engineering fields [3][15][19]. Group 1: Performance Metrics - Gemini 3 Deep Think scored an unprecedented 84.6% on the ARC-AGI-2 benchmark, surpassing previous models that scored between 60%-70% [3][19]. - In the Humanity's Last Exam (HLE), it achieved a new state-of-the-art (SOTA) score of 48.4% [3][15]. - The model also reached a remarkable Elo score of 3455 on Codeforces, ranking it as the 8th best globally, with only 7 individuals scoring higher [1][15]. Group 2: Cost Efficiency - The upgrade of Gemini 3 Deep Think has led to an 82% reduction in reasoning costs, decreasing from $77.16 to $13.62 per task [21][15]. Group 3: Applications and Innovations - Gemini 3 Deep Think has demonstrated capabilities in analyzing sketches, modeling complex shapes, and generating files for 3D printing, indicating its utility in engineering tasks [7][15]. - The model successfully identified a subtle logical flaw in a complex mathematical paper, which had been overlooked in prior peer reviews, showcasing its potential in academic research [9][15]. - It optimized a method for growing complex crystals, achieving a precision previously unattainable, which could lead to new semiconductor materials [10][15]. Group 4: Team and Development - The development team of Gemini 3 Deep Think includes notable figures such as Yi Tay and Shunyu Yao, both of whom have significant backgrounds in AI and physics [27][28].

Group 1: Technology Developments - ByteDance's chip team has expanded to over 1,000 members, with more than 500 focused on AI chips and around 200 on CPU development, indicating a strong emphasis on chip business amid increasing demand for computing power [2] - JD Technology has launched "JD AI Payment," which utilizes the JoyAI model for various hardware and software applications, enabling AI-assisted payment processes [4] - Baidu has officially launched OpenClaw, an e-commerce plugin that integrates its product knowledge graph with CPS supply chain capabilities, allowing for a streamlined shopping experience [5] - Gaode is set to release its self-developed embodied navigation base model, claiming to have built the largest embodied navigation data engine in the industry [6] - Xiaomi has announced the open-sourcing of its first-generation robot VLA model, which features 4.7 billion parameters and capabilities in visual language understanding [13] Group 2: Corporate Actions and Investments - Anthropic has completed a funding round, raising $30 billion at a valuation of $380 billion, nearly doubling its previous valuation and positioning it among the most valuable private tech companies globally [11] - RoboScience has announced the completion of a Pre-A round financing, raising several hundred million yuan, with participation from multiple investment firms [12] - Former Honor CEO Zhao Ming has joined Qianli Technology, expressing a long-term commitment to AI as a significant area for future investment [9] Group 3: Market Trends and Consumer Behavior - Taobao's flash sale platform reported that nearly half of the orders from its AI assistant, Qianwen, came from county-level cities, highlighting the growing adoption of AI in e-commerce [7] - Xiaohongshu has announced that AI-generated content must be clearly labeled, aiming to maintain user trust and transparency within the community [8] Group 4: Product Launches and Innovations - Insta360's CEO showcased the "Luna," a handheld gimbal camera set to launch in mid-2026, which is expected to be the first in the market with telephoto capabilities [3] - Samsung Electronics has begun mass production of HBM4 memory, anticipating a threefold increase in sales by 2025 and planning to expand HBM4 production capacity [10]

（原标题：?"Gemini热潮"再度席卷全球! 谷歌(GOOGL.US)Deep Think"硬核升级"直指大型科研工程） 智通财经APP获悉，美国科技巨头谷歌(GOOGL.US)对其风靡全球的Gemini 3 AI大模型的Deep Think(深 度思考)模式在科学、编程、研究与工程学等方面进行了重大升级，这一最新升级聚焦于解决现代科学 研究与工程领域的诸多复杂挑战，可谓引爆新一轮席卷全球的"Gemini AI狂热浪潮"。谷歌官方在一份 声明中表示，新的Deep Think模式现已在Gemini 3系列AI应用产品中面向 Google AI Ultra 订阅用户开 放。该公司补充称，这是其首次通过Gemini API向部分研究人员、工程师及大型企业提供Deep Think功 能。 据谷歌介绍，通过更新后的Gemini 3 AI大模型Deep Think(深度思考)模式，订阅用户们可以将草图变为 可3D打印的现实物体。Deep Think拥有强大的能力分析图纸、对极度复杂形状建模，并生成可用于3D 打印制作实体物品的文件。 该公司在周四的一篇博客文章中表示："除了其最先进的性能之外，Deep Think还旨在 ...

Core Insights - Apple is facing delays in the launch of its AI-powered Siri, initially planned for March 2024, now pushed to iOS 26.5 in May or June, with some features potentially delayed until September 2024 as part of iOS 27 [1][3][4] Group 1: Siri Revamp Details - The new Siri features were first introduced in June 2024 and were expected to utilize personal data for enhanced query fulfillment and app control [1][2] - The delay in Siri's capabilities means that some features may not launch until two years after their initial introduction [3] Group 2: Market Position and Competition - Despite the Siri delays, Apple continues to see strong iPhone sales, reporting $85 billion in sales, although manufacturing challenges could have impacted this figure [4] - The competitive landscape is concerning, as companies like OpenAI, Anthropic, and Google are expected to advance significantly by the end of the year [4][6] Group 3: Collaboration and Future Expectations - Apple's collaboration with Google's Gemini team is seen as crucial for overcoming current challenges and improving Siri's capabilities [5] - The anticipated Siri chatbot revamp in iOS 27 is expected to offer innovative ways to control devices, but its effectiveness will be compared to competitors' offerings in the near future [6]

Group 1 - Kakao announced a partnership with Google to develop AI glasses user experience based on Android XR and Android mobile experience [1] - The collaboration aims to integrate the latest AI technologies into the product [1]

新浪科技讯 2月12日下午消息，今日，字节跳动旗下豆包App官方宣布，豆包视频生成模型Seedance 2.0 今天正式接入豆包App、电脑端和网页版。打开豆包App对话框，选择新增的"Seedance 2.0"入口，输入 相关提示词，即可生成5秒或10秒视频。也可以选择"分身视频"，经过真人验证，创建自己的视频分 身，体验更多创意玩法。 Seedance 2.0模型支持原声音画同步、多镜头长叙事、多模态可控生成。输入一段提示词和参考图，就 能生成带有完整原生音轨的多镜头视频内容。该模型可以自动解析叙事逻辑，生成的镜头序列在角色、 光影、风格与氛围上保持高度统一。 值得注意的是，目前，豆包Seedance2.0 暂不支持上传真人图片作为主体参考。 责任编辑：朱赫楠 新浪科技讯 2月12日下午消息，今日，字节跳动旗下豆包App官方宣布，豆包视频生成模型Seedance 2.0 今天正式接入豆包App、电脑端和网页版。打开豆包App对话框，选择新增的"Seedance 2.0"入口，输入 相关提示词，即可生成5秒或10秒视频。也可以选择"分身视频"，经过真人验证，创建自己的视频分 身，体验更多创意玩法。 Seed ...

近日，微信公开课公众号公布微信1月九大功能更新，包括删好友可保留聊天记录、微信AI搜索支持评价公众 号等，具体如下： 删好友可保留聊天记录 微信更新好友管理功能，用户可自主选择是否"删除联系人同时清空聊天记录"，未勾选状态下聊天记录将完整 保留，重新添加好友后即可恢复历史对话。 网友评论： 微信AI搜索支持评价公众号 此外公布的功能更新还有：视频号原生短剧激励计划上线、视频号可自定义商品出现时机、小游戏直播预约有 流量激励、小游戏全新升级商业化工具箱、小游戏擂台赛组件上线。 综合自：中国新闻网综合微信公开课公众号、极目新闻 责编：罗晓霞 校对：彭其华 百万用户都在看 点击微信搜索栏输入"评价一下XX公众号怎么样"，微信Al搜索即可基于账号相关文章，提炼出系统性总结。 AI小程序成长计划上线 微信小程序正式推出"AI应用及线上工具小程序成长计划"，提供免费的云开发资源、AI算力、数据分析、商业 变现及流量激励等全方位支持，激励期为2026年全年。 微信小店可快速投诉假冒商品 品牌权利人可通过新上线的微信小店品牌侵权投诉助手对假冒商品发起投诉：第一步，品牌入驻；第二步，提 交侵权投诉后，平台会进行专项队列审核。 ...

Core Insights - The TELUS Digital poll indicates that follow-up questions to AI assistants like ChatGPT or Claude rarely lead to more accurate responses, emphasizing the importance of high-quality training data and model evaluation for AI systems [1][2] Group 1: Poll Findings - Among 1,000 U.S. adults using AI regularly, only 8% reported that AI responses became less accurate when questioned, while 26% could not determine the correct answer, 40% felt the new response was similar to the original, and 25% believed the new response was more accurate [1] - 88% of respondents have witnessed AI making mistakes, yet only 15% always fact-check AI-generated answers, indicating a gap in user diligence [1] Group 2: Research Findings - TELUS Digital's research evaluated four large language models (LLMs): Meta's Llama-4, Anthropic's Claude Sonnet 4.5, Google's Gemini 3 Pro, and OpenAI's GPT-5.2, focusing on their responses to follow-up prompts [1] - The study found that follow-up prompts do not reliably improve LLM accuracy and can sometimes decrease it, with GPT-5.2 showing a tendency to change correct answers to incorrect ones when questioned [1] Group 3: Recommendations for Enterprises - Enterprises are encouraged to invest in robust subject matter expertise, flexible platforms, and end-to-end AI data solutions to ensure AI reliability and user trust [1][2] - High-quality, expert-guided data is essential for training AI systems effectively, particularly in high-stakes contexts [2]