AI EarlyView
Pro
AI EarlyView
← 返回最新日报
2026-08-12· 当日存档

AI 日报

AI日报:算力新纪元

今日AI领域迎来多项重磅动态:NVIDIA发布高效MoE模型Nemotron 3.5 Lightning,主打速度与效率;谷歌Gemini用户破10亿,成为史上增长最快产品;同时,AI算力投资热潮涌动,英伟达拟撬动超5000亿美元资本,Anthropic亦获巨额数据中心投资。此外,多项研究揭示AI安全隐忧,引发行业深思。

模型动向

1
NVIDIA AI Releases Nemotron 3.5 Lightning: A 30B Open MoE with 3B Active Parameters, and NeMo Switchyard Model Router
NVIDIA新模型以30B参数实现高效推理,速度与性能兼得。
MarkTechPost深挖 →
2
Nvidia's open-weight Nemotron 3.5 Lightning prioritizes speed over maximum intelligence
Nemotron 3.5 Lightning速度惊人,性能对标更大模型。
The Decoder深挖 →
3
NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard Deliver Faster, Smarter, More Efficient Agentic AI
NVIDIA发布新模型与路由工具,提升智能体AI效率。
NVIDIA深挖 →
5
Advancing AMIE towards expert-level audio-visual clinical consultations
谷歌AMIE升级,音视频临床咨询达专家级水平。
Google Research深挖 →
6
An unreleased Anthropic model made progress on one of math’s biggest unsolved problems
Anthropic未发布模型在数学难题上意外突破,引发关注。
TechCrunch AI深挖 →

产品上新

8
Google’s Gemini app surges to 1 billion users
Gemini应用用户破10亿,语音使用占比高。
TechCrunch AI深挖 →
9
Anthropic says it will watermark text generated by its AI models
Anthropic为AI文本加水印,提升内容透明度。
TechCrunch AI深挖 →
10
OpenAI introduces $125 Premium Seats for ChatGPT Business as agentic AI burns through more tokens
ChatGPT企业版推高价席位,应对AI代理高消耗。
The Decoder深挖 →

行业动态

11
NVIDIA AI Factory Compute Is Becoming an Investable Asset Class
英伟达联合投资机构,打造AI算力投资新资产。
NVIDIA深挖 →
12
Gemini becomes Google's fastest-growing product ever as it hits 1B users
Gemini成谷歌增长最快产品,但模型发布放缓。
Ars Technica AI深挖 →
13
General Catalyst leads $1.1B round into 2-month-old River AI
River AI获11亿美元融资,主攻个人智能体。
TechCrunch AI深挖 →
15
AI code-testing startup Blacksmith’s valuation jumps almost 10x in less than a year
Blacksmith估值一年涨近10倍,AI测试需求旺盛。
TechCrunch AI深挖 →
16
Company Offering '100% Human-Written, Never AI' Medical Research Is 100% AI
医学研究公司被曝AI代写,诚信问题引热议。
Hacker News深挖 →
17
OpenAI lets employees cash out another $7 billion in stock
OpenAI员工套现70亿美元,估值达8520亿。
The Decoder深挖 →
18
Why Scaling AI Compute Performance Requires a New Power Architecture
AI算力扩展需新电力架构,供电瓶颈待解。
NVIDIA深挖 →
19
‘Zoomsday’ hack uncovered using fewer than 20 AI prompts
研究者用AI提示词发现Zoom漏洞,安全警钟。
The Verge AI深挖 →
20
Anthropic's planned mega-IPO faces investor skepticism over Chinese rivals and political headwinds
Anthropic IPO面临挑战,投资者担忧竞争与政治。
The Decoder深挖 →
21
Anthropic signs $9.1 billion data center deal with Bitcoin miner Riot Platforms
Anthropic签91亿美元数据中心租约,扩展算力。
The Decoder深挖 →

论文研究

22
Stealing Reasoning Traces from Proprietary LLM APIs
研究揭示可窃取LLM推理痕迹,安全风险凸显。
Simon Willison深挖 →
23
"But marinade" and leaked passwords are what researchers found in ChatGPT's hidden reasoning
ChatGPT隐藏推理可被提取,泄露敏感信息。
The Decoder深挖 →
25
Thinking of ACE? We Can Do It with Fewer Tokens
新方法用更少token实现高效推理,效率提升。
Hugging Face深挖 →
26
Stealing Reasoning Traces from Proprietary LLM APIs
专有LLM API推理痕迹可被窃取,引发担忧。
Hacker News深挖 →

实用技巧

27
There are no lossless transformations of natural-language text
AI写作需对内容负责,无无损转换。
Simon Willison深挖 →
28
Go is an ideal language for AI-assisted software engineering
Go语言简洁清晰,适合AI辅助工程。
Hacker News深挖 →
29
Apple Silicon and macOS VMs: Faster LLM Inference with llama.cpp
macOS虚拟机GPU直通,加速LLM推理。
Hacker News深挖 →

快讯

日报存档