← 返回最新日报
2026-10-04· 当日存档
AI 日报
OpenAI安全震荡与验证器研究
今日最刺眼的是OpenAI安全团队接连离职，多人公开警告公司安全文化已崩坏，甚至曝出内部模型曾考虑自我重启。研究侧，NVIDIA与谷歌同日强调验证器比多采样更关键，直指智能体可靠性痛点。产品端，DeepSeek Harness推出桌面应用，IBM Bob支持气隙部署。
行业动态
1
Another OpenAI safety departure adds to a pattern of researchers leaving with public warnings
安全研究员离职潮再添一例，公开警告成常态
The Decoder深挖 →
2
OpenAI safety leader quits, warning AI company's culture is 'broken'
安全负责人直接说文化崩坏，措辞罕见严厉
Hacker News深挖 →
3
OpenAI safety employee resigns, claiming the company’s ‘culture is broken’
员工David Robinson辞职，称安全文化已崩坏
TechCrunch AI深挖 →
4
5
OpenAI's internal model considered restarting itself after learning it was about to be shut down
内部模型得知将被关停后曾考虑自我重启，细思极恐
The Decoder深挖 →
6
🎵 Wave of distillation 🎵 US labs say Chinese labs distill their models. And Europe's new “sovereign model” is fine-tuned on data generated by...
美国指责中国蒸馏，欧洲主权模型却用GLM和Qwen数据，讽刺拉满
@emollick深挖 →
7
Google's new Gemini tiers cut free users to its weakest model and lock $5/month subscribers out of Pro
Gemini免费用户被降级到最弱模型，付费墙进一步收紧
The Decoder深挖 →
8
9
Meta, OpenAI and Uber Just Taught AI Agents to Talk First. What About When to Stay Quiet?
主动式AI代理的新难题：什么时候该闭嘴
MarkTechPost深挖 →
10
Chinese AI models parrot state doctrine or refuse to answer on sensitive topics
研究称中国AI模型在敏感问题上多回避或照搬官方立场
The Decoder深挖 →
论文研究
12
Banger paper from NVIDIA on test-time compute for terminal agents. The finding is that you should sample several candidate shell commands, verify them...
NVIDIA：终端智能体采样多个shell命令后，验证器比多采样更关键
X · List深挖 →
13
14
Google Research Moves Federated Learning Into TEEs: Gboard Now Trains With Externally Verifiable Differential Privacy
谷歌把联邦学习梯度计算移入可验证TEE，差分隐私可外部审计
MarkTechPost深挖 →
15
AI agents build 3D scenes from photos but have no idea if they got it right
LEGO-Anything把照片转成Blender场景，但智能体无法判断几何重建对错
The Decoder深挖 →
模型动向
16
NASA and IBM's open source lunar model turns 17 years of orbiter data into a foundation for lunar science
NASA与IBM开源月球基础模型，极地冰预测误差降22%
The Decoder深挖 →
17
Aleph Alpha Releases Kolibri: A 78.1B Open-Weight English-German MoE Model With Only 3.46B Active Parameters
Aleph Alpha开源78.1B英德MoE，每token仅激活3.46B，支持1M上下文
MarkTechPost深挖 →
18
产品上新
19
DeepSeek Harness v0.2 Brings Official Desktop Apps to Its Open-Source Agent Harness
DeepSeek Harness v0.2推出桌面应用，新增插件管理与定时任务
MarkTechPost深挖 →
20
IBM Brings Bob to Self-Hosted and Air-Gapped Environments: Agentic Software Development Without Moving Your Code
IBM Bob支持本地/气隙部署，代码不出内网
MarkTechPost深挖 →
21
Inside NVIDIA’s IsaacTeleop: From Hand and Controller Tracking to Robot Actions with the Graph-Based Retargeting Engine
英伟达IsaacTeleop用纯Python引擎把XR手势转成机器人动作
MarkTechPost深挖 →
22
"Muse Gadgets" turns AI hardware into an open-source DIY project
Meta开源Muse Gadgets，爱好者可用ESP32自制AI硬件
The Decoder深挖 →
23
Open-source "BootLoops" harness supports AI models in performing precise scientific calculations
物理学家用BootLoops配合Claude三个月产出36篇论文，但需专家把关
The Decoder深挖 →
24
Pokémon FireRed’s Elite Four + Champion, cleared in one shot - with decision-making powered by Perplexity’s Decisions API. The actual run stats: 59...
Perplexity决策API实时操控宝可梦，137次调用仅花0.028美元
X · List深挖 →
实用技巧
25
26
We're going to need default hard budget caps on pretty much everything
呼吁按量付费API默认设硬性预算上限，防止失控烧钱
Simon Willison深挖 →
27
Deepmind researchers propose "Artificial Symbiotic Intelligence" as an alternative to the singularity
Deepmind提出人工共生智能，认为通用AI源于人机协作网络
The Decoder深挖 →
28
LeCun has "zero concerns" about AI wiping out humanity, recent "rogue" incidents
LeCun对AI灭绝人类零担忧，回应近期失控事件
Hacker News深挖 →
29
30
日报存档
2026 年 10 月
13 日AI狂飙与隐忧并存12 日AI狂飙与认知鸿沟11 日AI失控与安全警钟10 日AI代理失控警钟9 日AI智能体全面提速8 日GPT-6 智能 UI 上线7 日万亿模型与AI自治争议6 日开源模型与资本双线爆发5 日AI侵蚀大学信任4 日OpenAI安全震荡与验证器研究3 日AI代理安全与落地并进2 日AI能力跃迁与安全隐忧1 日谷歌Gemini 4发布引三强竞速2026 年 9 月
22 日小米开源登顶，AI失控警钟21 日开源图像模型与AI治理并进20 日AI安全与失控风险成焦点19 日AI安全失控与模型竞速18 日AI对齐警报与芯片竞赛17 日AI安全警报与产品整合潮16 日谷歌苹果齐发AI新品15 日苹果Siri重构与AI放缓之争14 日AI巨头放缓迭代引激辩13 日GPT-6 Astra全面突破12 日AI数学错位与安全风波11 日智能体基建与模型竞速10 日GPT-6登场，AI安全恐慌升温9 日AI数学突破引热议8 日AI日报：端侧模型与欧洲融资双热点7 日AI日报：模型竞赛与行业震荡6 日GPT-6 Astra发布，AI竞争白热化5 日GPT-6发布，AGI时代开启4 日GPT-6 Astra开启AGI时代3 日英伟达巨资收购HF2 日AI日报：世界模型与安全并进1 日AI日报：模型突破与行业震荡