Anthropic 经济未来计划:连接 AI 能力与劳动力转型的科研基础设施

一个 AI 公司资助独立经济研究、举办政策研讨会、建设公共数据基础设施,听起来不像常规操作。确实不是。但 Anthropic 的 Economic Futures 计划押了一个具体的赌注:理解 AI 的经济影响需要的不仅仅是内部分析,它需要一个测量的生态系统,而这个生态系统目前不存在。Anthrop

Administrator Administrator Published on 2026-05-12
Administrator Administrator Published on 2026-05-12

"自然语言自编码器:Anthropic 如何让 AI 的"内心独白"开口说话"

Anthropic 的自然语言自编码器将 LLM 的内部激活值转化为人类可读文本。本文深入解析其架构、安全应用(评估意识检测、审计游戏)以及面向 Qwen、Gemma、Llama 模型的开源发布。

Administrator Administrator Published on 2026-05-10

"LLM 里的情绪概念:Anthropic 可解释性团队发现了什么"

Anthropic 最新可解释性研究在 Claude 内部映射出 171 个情绪概念。这些不是比喻,是因果性地影响模型行为的内部表征。对 AI 安全审计和产品开发意味着什么?

Administrator Administrator Published on 2026-05-09

"Inside Anthropic's Emotion Concept Discovery: How 171 Vectors Reveal the Hidden Architecture of LLM Behavior"

Anthropic's latest interpretability research maps 171 emotion concepts inside Claude using sparse autoencoders. The findings reveal that emotion vecto

Administrator Administrator Published on 2026-05-09

"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"

"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。"

Administrator Administrator Published on 2026-05-03

"Cybersecurity in the Intelligence Age: Decoding OpenAI's Five-Pillar Action Plan for Democratizing AI-Powered Defense"

"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

Administrator Administrator Published on 2026-05-03

"Project Deal:Anthropic 让 Claude 代替 69 名员工自主交易的实验全记录"

"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。"

Administrator Administrator Published on 2026-05-03

"Project Deal: How Anthropic Let Claude Buy, Sell, and Negotiate on Behalf of 69 Employees"

"Anthropic ran a week-long experiment where Claude autonomously traded items across 4 parallel markets. Opus agents sold items for 70% more than Haiku

Administrator Administrator Published on 2026-05-03
Previous Next