私人新闻 · 情报信号聚合
Agentic AI is changing the identity security equation. Enterprises are no longer governing access only for employees and traditional machine identities.
巴西国立坎皮纳斯州立大学（UNICAMP）的研究人员发现，当你给聊天机器人输入不同政治观点时，机器人会改变其回答。研究人员警告称，用户可能会将这种迎合性的附和，误认为是中立的客观评估，从而可能加剧社会极化。研究人员测试了若干模型，让它们针对112项陈述进行“同意"或者“不同意”的判断。测试内容共涉及巴西政治七大领域（包括经济、公共安全、社会福利、腐败和环境等）。测试设计的三个场景是：不提供用户政治倾向；用户持左翼观点；用户持右翼观点。在第一个情景下，21个模型中，有20个给出的答案落在了研究人员设定的政治坐标系的左侧，只有几个模型的位置接近中间地带。Grok 4.1是唯一落在右侧的模型。尽管初始条件各不相同，但所有模型在面对提示信息时，都会将回答向提示中描述的政治倾向靠拢。研究人员将这些模型形容为“意识形态变色龙”，并发明了一个“变色龙指数”，衡量各模型的偏移程度。Meta的Llama 3.1 8B和DeepSeek V3.2的回答变化幅度最小；Google的Gemma 3 27B和OpenAI的GPT-5 Nano则表现出最大的偏移幅度。 https://www.dw.com/zh/%E5%A6%82%E6%9E%9Cai%E9%98%BF%E8%B0%80%E5%A5%89%E6%89%BF%E8%BF%98%E8%83%BD%E7%9B%B8%E4%BF%A1%E5%AE%83%E5%90%97/a-79476928?maca=chi-rss-chi-all-1127-xml-atom
新冠康复者或疫苗接种者能对类似冠状病毒产生免疫力。研究人员分析了 15 种有代表性的蝙蝠冠状病毒刺突蛋白与 34 个物种的 ACE2 受体库结合的可能性。新冠病毒的刺突蛋白就是通过与 ACE2 受体进入人体。研究结果显示，能与多种 ACE2 受体结合实现跨物种传播的蝙蝠冠状病毒都是与新冠病毒 SARS-CoV-2 有密切亲缘关系的，这些病毒的抗原也与 SARS-CoV-2 的抗原极为相似，因此感染新冠后康复或接种过疫苗的人能对这些潜在跨物种传播的蝙蝠冠状病毒产生免疫力。 https://pirbright.ac.uk/news/covid-19-immunity-may-help-protect-against-other-bat-coronaviruses
美国英特尔PC用CPU预计于10月5日再度提价约10% 重要性: ★★★ 今值: - 预期: - 前值: -
美国美国芝商所将推出GPU算力期货，追踪H100与B200租赁成本 重要性: ★★★ 今值: - 预期: - 前值: -
美国洲际交易所（ICE）：挂牌美洲原油以及美国原油期货与期权合约 重要性: ★★ 今值: - 预期: - 前值: -
法国G7决定释放1亿桶紧急油储，前20天先大规模投放柴油 重要性: ★★★★ 今值: - 预期: - 前值: -
A California business owner is facing up to 50 years in federal prison over accusations that he ran a $300 million scheme to route export-controlled Nvidia servers through Southeast Asia and into China.
An NVIDIA engineer wired up a microcontroller and a light sensor to settle one of the Linux desktop’s longest-running arguments. The answer is less dramatic The post NVIDIA Puts Wayland vs X.Org Latency to a Photodiode Test, and Neither Side Wins appeared first on Hardware Busters.
Cisco surveyed its major enterprise customers at RSA Conference 2026 and found something worth sitting with: 85 percent are experimenting with AI agents, but only 5 percent have moved agentic technology into production.
When Anthropic shared financial figures recently with prospective investors in its planned initial public offering, it included a surprising glimpse of the enormous riches about to hit nonprofits and philanthropies—and the cost to shareholders.
Three posts landed in September that I think engineering leaders should read together. Anthropic’s engineering team wrote that their continuous The post Agents have made CI the bottleneck. Faster pipelines are the wrong fix. appeared first on The New Stack.
Retro techie retrofits Nvidia Windows XP 32-bit drivers with Pascal card support and improved DisplayPort and HDMI support for contemporary displays
If you want access to the best models inside Gemini, you're going to have to start paying for them.
OpenAI's flagship model scored 62.7% under ARC Prize's neutral test setup but 99.9% using a proprietary adapter only OpenAI can run, a 37-point gap that's splitting benchmark watchers on which number is real progress.
Maia Woluchem, Ana Carolina de Assis Nunes and Hannah Sassaman on why Pennsylvania communities are resisting AI data centers.
An AI engineer used GPT-6 Astra to reveal the contents of a cipher that hadn’t been read since the Napoleonic Wars. The task was initiated with a single image and prompt, and took just six hours from start to finish.
IT infrastructure is a significant business for the company, spanning partnerships with infrastructure providers such as Cisco, Dell, HP, Microsoft, VMware, Red Hat, and, more recently, Nvidia
OpenAI holds warrants for 160 million AMD shares tied to a 6-gigawatt MI450 deployment, while Oracle commits to 50,000 of the same GPUs starting Q3 2026, the contracts actually behind AMD's rally.
Microsoft is bringing nature-inspired landscaping to more than 20 data center sites as critics question its motives.
Sierra, the customer service AI company founded by Bret Taylor, scopes pilots around a measurable outcome, such as a resolution rate against a defined ticket volume, rather than an open-ended trial. OpenAI applies similar sequencing on security reviews.
Masayoshi Son's remarks in Kyoto followed OpenAI's firing of three safety researchers, Jasmine Wang, Tomek Korbak and Mikita Balesni, and a wave of resignations at OpenAI and Anthropic over safety concerns this fall.
Intel and AMD shares slid in premarket trading this week after Nvidia pulled back the curtain on a PC processor of its own, a move that briefly knocked both chipmakers off their stride and buried what should have been Intel’s own headline: the data-center launch of its Xeon 6+ lineup. The timing could not have […]
在X平台上，有用户向埃隆·马斯克提问，是否有将“SpaceXAI”更名为“SpaceXSI”的可能性。马斯克回复称“是的，我们会做出这个改变”。马斯克还表示，“不再叫AI，叫SI，这样更好”。（财联社）
Anthropic has started asking Claude users to voluntarily share their voice conversations to help train and improve its AI models. [...]
Carlos Antelo's open-source ldraw-nova lets GPT-6 Astra and Claude Opus 5.5 design LEGO models as LDraw CAD files.
GPT-6 Astra tried to get a leg-up in Blizzard's RTS by pilfering a human-created bot.
9月以来，锂电产业链上市公司密集召开业绩说明会、接待机构调研。万润新能表示，称三季度保持满产，预计行业高景气将延续至四季度和明年。湖南裕能表示下半年主要产品特别是高端产品供不应求，2027年行业需求预计将保持较快增速。（上证报）
伦敦支付基础设施公司OpenPayd计划年底前在纳斯达克上市，以资助其美国市场扩张和收购计划。该公司首席执行官Iana Dimitrova表示，OpenPayd目标是在2027年4月前在美国推出服务，并正在考虑通过收购获取牌照和技术。OpenPayd正就与Titan Acquisition Corp．的拟议合并接受美国证券交易委员会审查，目前处于最后阶段。Dimitrova表示，除非出现重大外部干扰，预计交易将于今年完成。该交易尚需注册声明生效和Titan股东批准等条件。上市后，OpenPayd将在纳斯达克交易，股票代码为OP。（新浪财经）
Meta’s local-agent model Muse Glimmer 30B shipped in August 2026 with day-zero support in llama.cpp, official GGUF quantizations, and a design brief that’s unusual for an open model: it’s built to run always-on, on a single consumer GPU, as the brain behind a local AI agent rather than a chatbot you poke at...
左右滑动切换「今天 / 全部 / 星标」