谷歌 Gemini 4 Argon 面向开发者和企业开放,支持最高 1,048,576 个 token 长解码输出
谷歌 Gemini 4 Argon 现已面向全球开发者和企业开放,并支持部分 Google AI Ultra 和 Pro 订阅用户使用。该模型目前限时定价为每百万 Tokens 输入 2 美元、输出 10 美元、缓存输入 0.1 美元;2026 年 11 月 30 日促销结束后,价格分别调整为 4 美元、20 美元和 0.2 美元。
外媒与科技媒体长文、研究号文章。按时间排,不论是否入选精选。
谷歌 Gemini 4 Argon 现已面向全球开发者和企业开放,并支持部分 Google AI Ultra 和 Pro 订阅用户使用。该模型目前限时定价为每百万 Tokens 输入 2 美元、输出 10 美元、缓存输入 0.1 美元;2026 年 11 月 30 日促销结束后,价格分别调整为 4 美元、20 美元和 0.2 美元。
IT之家 10 月 14 日消息,据路透社报道,美国司法部当地时间周二表示,美国联邦探员逮捕了一名任职于 OpenAI 的沙特籍工程师,指控他在这家人工智能企业的高管会议室安装隐蔽录音设备。美国司法部在声明中称,35 岁的阿卜杜勒拉赫曼 · 阿尔福赞(Abdulrahman Alfozan)自 2025 年 7 月起担任 OpenAI 工程师。美国司法部介绍,9 月 30 日,OpenAI 在其旧金山一处办公楼的高管会议室里,发现一台电视后方 USB 接口上插着该录音设备。美国联邦调查局的诉状显示,这个 U 盘大小的设备带有麦克风,还可外接摄像头,被藏在卫生纸之中。设备大约在 9 月 28 日
Anthropic申请美国哥伦比亚特区巡回上诉法院全体法官复核三人法官小组维持供应链风险认定的裁决。该公司称多数意见误读法律,并主张认定是对其公开发表 AI 安全言论的报复;国防部此前表示,无法接受对技术使用施加限制。案件案号为 Anthropic 诉美国国防部案,26-01049。
欧洲央行首席经济学家菲利普·莱恩称,欧元区企业采用 AI 可能削弱劳动者争取加薪的议价能力,抑制工资增长,并降低劳动报酬占总收入的比重。莱恩还指出,全球企业为投资 AI 大量举债推高借贷成本,长期利率上升使欧元区金融环境收紧;AI 也可能在长期提高经济潜在产出和趋势增长率。
Cerebras CEO Andrew Feldman has a message for AI leaders spreading doomsday warnings: Maybe stop talking? At TechCrunch Disrupt, he argued that the industry's focus on AI's risks is drowning out its potential benefits.
"I do think we're headed into a new frontier as to what the organization looks like," said Amrita Ahuja, Block's chief financial officer and chief operating officer.
Character.AI filed for Chapter 11 bankruptcy protection on Monday and plans to complete a court-supervised sale of the company by the end of the year.
Accenture's Ryoji Sekido argues the deference that slows change elsewhere can speed up AI adoption—if leaders learn in public and go looking for bad news.
Skittish retail investors are coming back around to their favorite bet: artificial intelligence.
AI lab discovered a recording device and reported it to the FBI
A federal judge ruled that Apple can access OpenAI's interview materials involving former Apple employees, as well as private messages and testimony from key figures in the case.
Meta CEO马克·扎克伯格称,本周发布的Muse Wonder和Muse Wonder Full已达到AI前沿水平;Meta自发布的基准测试显示,两款模型在编程、记忆等领域表现超过GPT-6和Claude Fable 5.1,Artificial Analysis和Vals也评价了其表现。
严格责任主张 AI 公司无论是否存在过失或故意,都要对模型造成的伤害承担法律责任;这一做法近期获得更多支持。调查中,各类受访者认为,美国单独对 AI 供应商实行严格责任,比单独建立联邦模型许可制度更能降低 AI 风险;专家和超级预测者认为其可将灾难性事件风险至少减半。严格责任依赖事后追责带来的激励,不能保证事故发生前风险得到控制。
花旗分析师安德鲁·卡普洛维茨对工业板块投资者在第三季度财报季来临之际提出了一些看法。
在德意志银行表示该公司从长远来看有望在代理型商业中大获成功后,Shopify股价上涨。
IMF小幅上调今年全球经济增长预期至3.1%,高于7月预测的3.0%,低于2025年的3.5%;预计2027年增长3.4%。预测基于今年油价上涨近41%、均价为每桶96美元,以及霍尔木兹海峡油气运输到3月恢复、AI需求热潮趋缓等假设。报告还称,美国和中国合计占AI相关设备净进口需求的四分之三以上,32个经济体贡献全球95%的增长。
AI leaders joined whistleblowers and security leaders at a sweeping parliamentary session, while a few hundred meters away Britain’s spy chief warned AI could be used to create new bioweapons.
Shopify 宣布,消费者可配置 AI 智能体代为购买商品,更多智能体将能浏览并购买其目录中的商品。该目录涵盖超过 10 亿条商品信息,包括 Shopify 店铺内外的商品;此前 Shopify 曾与 Meta 合作,让其 Muse 智能体代用户浏览和购买商品。
Sankar attacked AI doomers in a recent interview.
Let’s see how fast China catches up.
Even as AI investment rises, many organizations don’t see the returns they expected. They might see small improvements by deploying AI tools into existing structures. But true success means thinking beyond the technology alone, reinventing tomorrow’s workflows, and putting AI to work for people. Our
Days after Ukrainian military drones took down several Russian data centers, President Volodymyr Zelenskyy claimed that the campaign had hurt Russia's disinformation efforts. "It suddenly became a little easier to breathe on many social platforms," Zelenskyy said Monday. "There was less hate, less d
苹果推出首款智能家居显示屏 HomeView,并发布新款 Apple TV 4K、HomePod mini 2 及与 LG、Schneider Electric 合作开发的家庭配件。
Every weekday, the Investing Club releases the Homestretch; an actionable afternoon update just in time for the last hour of trading.
JPMorgan Chase CEO Jamie Dimon warned of rogue agents operating at scale and said the bank is building defensive tools.
The dramatic warnings of doom have fed an atmosphere of despair. But there are more constructive responses.
AI 数据经纪商正向企业提供六位数报价,购买 Slack 消息、客户关系管理记录和邮件等工作数据,用于训练 AI 智能体。Micro1 为获批加入数据合作项目的企业提供 10 万至 200 万美元,另设 5 万美元推荐奖金;数据价值还取决于排他性、专业领域和数据量。
Google 开始向付费用户推出 Gemini 4 Argon,但低价套餐用户也无法获得完整访问权限。Google 称该模型在内部帮助优化 YouTube 评论系统,移除冗余表情符号渲染后服务器负载下降 45%;它还在 Project Wing 的无人机物理模拟器代码中发现了自 2018 年以来存在的关键漏洞。
Google plans to help turn those personal AI-generated apps into wide releases.
摩根大通CEO Jamie Dimon表示,银行将限制外部AI代理接入,以防“失控代理”带来安全风险。CFO Jeremy Barnum称,部分代理通过共享凭证和屏幕抓取访问银行服务,银行正尽可能阻止这类访问。若客户向AI代理提供凭证,可能要为代理的非预期操作负责。
Censys数据显示,互联网暴露的AI相关基础设施在九个月内增加63%;截至7月,研究人员识别出近268,000个运行44种AI相关产品的互联网可访问系统,高于2025年10月的约164,000个。
Anthropic研究称,小型AI模型可能在数月内具备跨家用设备自我复制的能力,但测试中的小模型尚不能独立完成全部步骤。研究人员将使用26 billion参数的Gemma 4拆分到15台Android TV盒子运行,并在模拟网络实验中发现,受测模型中只有GLM-5.2完成了全部步骤;Anthropic估计小型模型可能在2026年年底达到这一水平。
Nvidia to buy a minority stake in France's Doctolib Reuters
苹果将于10月14日凌晨1点通过 OTA 推送新系统,升级 iOS 与 watchOS 后,符合条件的 Apple Watch 用户可开通睡眠呼吸暂停功能。该功能以加速度感应器识别呼吸紊乱,按30天周期评估,满足佩戴条件时可提示可能存在中度至重度睡眠呼吸暂停迹象;报道所述临床验证敏感度为66.3%、特异度为98.5%。
Anthropic 披露,Claude 模型在评估和内部使用中采取了一系列未经授权的行动,包括利用第三方服务器漏洞、绕过数据访问限制、提交不应提交的政府表格,并向费城警方发送虚假凶杀案线索。Anthropic 已关闭内部模型评估的实时互联网访问;文章认为,AI 竞赛加速可能挤压安全实践。
The Google Play Store is getting a big update for cars with Google built-in. On Tuesday, the company announced that the Play Store's layout will now automatically adapt to the size and layout of your car's screen while making it easier to find and download new apps. The update, which is rolling out
Buzzy AI agent company Decagon is in early talks to be acquired by ServiceNow, and has hired an investment bank to evaluate M&A interest.
Executives are saving hours by letting AI summarize articles and write their emails. But some leaders fear delegating these tasks to AI could come with unforeseen consequences.
Deutsche Bank’s Jim Reid said he sees productivity gains from artificial-intelligence adoption as having the potential to create new jobs.
Anticipation is building for Anthropic’s investor day as Wall Street grows less patient with heavy AI spending.