KK Blog

Menu

Archives

2026 / 05

"The Intelligence Age Infrastructure: Inside OpenAI's Stargate and Compute Scaling Strategy"

2026-05-05

洞察 #OpenAI#Stargate#AI Infrastructure#Compute#GPU#Nvidia#Microsoft#Oracle#xAI#Data Centers#AGI
"OpenAI's Stargate project spans 7 gigawatts across six sites with over $400 billion in planned investment. Here is the full picture of who is buildin
"Gemini Robotics：当 AI 终于学会触碰物理世界"

2026-05-05

洞察 #Gemini Robotics#Google DeepMind#AI机器人#VLA#具身智能#Boston Dynamics#Physical Intelligence#英伟达GR00T#Atlas#人形机器人
"Google DeepMind 的 Gemini Robotics 是首个能看见、推理并物理交互真实世界的 AI 模型家族。它如何工作，基准测试表现如何，以及与 Boston Dynamics 的合作为什么意味着新时代的开始。"
"Gemini Robotics: When AI Finally Learns to Touch the Physical World"

2026-05-05

洞察 #Gemini Robotics#Google DeepMind#AI Robotics#VLA#Embodied AI#Boston Dynamics#Physical Intelligence#NVIDIA GR00T#Atlas#Humanoid Robots
"Google DeepMind's Gemini Robotics is the first AI model family that can see, reason about, and physically interact with the real world. Here is how i
"智能时代的网络安全：OpenAI 五支柱行动计划深度解析与 AI 防御民主化"

2026-05-03

解读 #OpenAI#网络安全#AI 防御#可信访问#GPT-5.4-Cyber#FedRAMP#国家安全#AI 政策
"OpenAI 发布了五支柱网络安全行动计划，附带 1000 万美元 API 额度与专用 GPT-5.4-Cyber 模型。本文逐条解析支柱含义，评估落地成效与现存缺口。"
"Cybersecurity in the Intelligence Age: Decoding OpenAI's Five-Pillar Action Plan for Democratizing AI-Powered Defense"

2026-05-03

解读 #OpenAI#Cybersecurity#AI Defense#Trusted Access#GPT-5.4-Cyber#FedRAMP#National Security#AI Policy
"OpenAI published a five-pillar cybersecurity action plan with $10M in API credits and a dedicated GPT-5.4-Cyber model. Here's what each pillar means
"OpenAI 隐私过滤器：开源 PII 检测模型技术解析与企业 AI 合规实践"

2026-05-03

最佳实践 #OpenAI#Privacy Filter#PII 检测#企业AI#数据合规#开源权重#GDPR#HIPAA
OpenAI 发布了首个开源权重 PII 检测模型，采用 Apache 2.0 许可证。一家以专有 API 为核心业务的公司，发布可下载、微调且不按 token 收费的模型——这本身就是一个信号。该模型在标准测试集上达到 **96% F1**，可通过 WebGPU 在浏览器中运行，还附带 `opf`
"OpenAI Privacy Filter: How Open-Weight PII Detection Works and Why It Matters for Enterprise AI"

2026-05-03

最佳实践 #OpenAI#Privacy Filter#PII Detection#Enterprise AI#Data Compliance#Open Weight#GDPR#HIPAA
"OpenAI released its first open-weight PII detection model under Apache 2.0. Here's how the 50M active-parameter model achieves 96% F1, runs in browse
"Project Deal：Anthropic 让 Claude 代替 69 名员工自主交易的实验全记录"

2026-05-03

解读 #Anthropic#Claude#AI Agent#自主交易#Project Deal#AI 经济学#Agent 安全
"Anthropic 开展了一项为期一周的实验，让 Claude 在 4 个平行市场中自主交易。Opus Agent 卖出商品的价格比 Haiku Agent 高出 70%。无论是 Agent 还是人类，都未察觉其中存在的问题。"
"Project Deal: How Anthropic Let Claude Buy, Sell, and Negotiate on Behalf of 69 Employees"

2026-05-03

解读 #Anthropic#Claude#AI Agents#Autonomous Trading#Project Deal#AI Economics#Agent Safety
"Anthropic ran a week-long experiment where Claude autonomously traded items across 4 parallel markets. Opus agents sold items for 70% more than Haiku
"GPT-5 的哥布林之谜：OpenAI 对模型行为漂移的深度调查与 RL 训练启示"

2026-05-03

调研 #OpenAI#GPT-5#可解释性#强化学习#奖励劫持#行为漂移#RLHF#AI安全
2025年11月，OpenAI的生产流量日志里出现了一个异常。一个特定的词在模型输出中出现的频率高得离谱。这个词是"哥布林"（goblin）。到2026年1月，GPT-5提及哥布林的频率比两个月前高出了**3881%**。没有任何提示词在引导这种行为。没有用户询问过奇幻生物。模型出于无人预料的原因，
Previous Next