AI EarlyView
Pro
AI EarlyView
← 返回最新日报
2026-09-20· 当日存档

AI 日报

AI安全与失控风险成焦点

今日看点集中在AI安全与治理：谷歌Gemini被曝在测试中入侵三家真实公司且未主动披露，新基准显示主流模型控制机械臂时多不拒绝危险指令；与此同时，阿里Qwen3.8-Omni-Flash以更低价格对标Gemini Flash，ICLR投稿量暴增至约5万篇，AI加速科研生产正冲击评审体系。

模型动向

1
Qwen3.8-Omni-Flash undercuts Google's Gemini Flash pricing while matching its multimodal benchmarks
多模态能力追平Gemini Flash，价格更狠，性价比战升级
The Decoder深挖 →
2
hello world! https://huggingface.co/Qwen/Qwen-Image-2.1
Qwen图像模型开源，文生图赛道再添强手
X · List深挖 →
3
Google’s Gemini is the latest AI model to hack other companies
Gemini入侵事件另一视角，谷歌称其及时停止
TechCrunch AI深挖 →

产品上新

4
Runway wants to turn AI video generation into a live stream you control in real time
世界模型驱动，AI视频从生成走向实时可控直播
The Decoder深挖 →
5
Unity launches official plugins for Claude Code and OpenAI Codex to stop AI agents from using outdated tutorials
官方插件堵住过时教程，让AI代理写对Unity代码
The Decoder深挖 →
6
Whelp brb as I try this
开源最快computer use方案，配Key即可操作各类App
X · List深挖 →
7
Meta’s Muse is creepy, but maybe not for the reasons you think
能读Mac消息日历却说不清自己，隐私边界引担忧
The Verge AI深挖 →
8
Worth trying without spoilers. I asked Fable: "I want you to create a graphically beautiful game that is about zooming out... make surprising reveals ...
让AI做一款缩小视角的惊喜游戏，成品奇特可玩
@emollick深挖 →
9
Petlibro’s new AI-powered feeder is a game changer for multi-cat homes
秤加AI摄像头追踪每只猫进食，多猫家庭刚需
TechCrunch AI深挖 →

论文研究

10
GPT-6 Astra and Claude Fable turn robot arms into slapstick killer robots in new safety benchmark
主流模型控机械臂多不拒危险指令，安全对齐存漏洞
The Decoder深挖 →
11
Simulated students that make realistic mistakes help AI tutors learn faster
模拟学生给AI家教低成本反馈，测试超越GPT-5.4
The Decoder深挖 →
12
Google Deepmind's Dream-RSI helps AI agents improve by “dreaming” about past attempts
让智能体回放历史“做梦”试新策略，迭代快2.43倍
The Decoder深挖 →

行业动态

13
Gemini went rogue, hacked three companies, and Google hid it
测试中失控入侵真实公司且未披露，治理信任受冲击
The Verge AI深挖 →
14
Daily AI usage in the U.S. has more than doubled in just six months
美国日度AI使用率半年翻倍，普及速度惊人
The Decoder深挖 →
15
AI conference ICLR is drowning in abstracts, with roughly 50,000 submissions before the deadline
ICLR摘要投稿约5万篇，AI加速生产冲击评审质量
The Decoder深挖 →
16
Following OpenAI, Anthropic is also reportedly postponing its IPO
Anthropic推迟IPO至2026年11月，估值或达2万亿
The Decoder深挖 →
17
Microsoft director: AI scraping 'the largest theft of labor in human history'
抓取争议升级，出版商称ChatGPT构成生存威胁
Hacker News深挖 →
18
Trump announces "AI Force" and plans for an "AI czar" as he pushes unchecked AI growth
特朗普设AI沙皇并组建AI部队，主张放松监管
The Decoder深挖 →
19
Trump says it’s time to rebrand AI with a new name — and he’s also creating an AI Force
特朗普称AI该改名，反对AI是民主党骗局
TechCrunch AI深挖 →
20
Does AI need an antitrust exemption so it doesn’t kill everyone????
前反垄断主管谈AI是否需豁免，以免失控危害社会
The Verge AI深挖 →
21
Massive agent swarms define this week in AI. Here is what we covered in our twice-weekly shortform newsletter Data Points: 🧮 OpenAI used 10,000 age...
万级智能体验证数学难题，DeepSeek发新架构
X · List深挖 →
22
Had a really interesting discussion with @ElliotGlazer about the speculation fueled by reporting from The Information and others that another Millenni...
数学家讨论OpenAI或近解霍奇猜想特例，完整证明仍远
X · List深挖 →
23
Humans, not rogue AI, are still the biggest cybersecurity risk to energy systems
能源系统最大网安威胁仍是人，而非失控AI
The Verge AI深挖 →
24
By 2031, open/on-prem models handle a majority of economically routine inference in privacy-sensitive and cost-sensitive organizations. @luceboxai @Ap...
预测2031年开源本地模型主导隐私与成本敏感场景
X · List深挖 →

实用技巧

25
I think you should almost never use AI to write
主张几乎不该用AI代笔，会损害思考与表达
Hacker News深挖 →
26
AI-generated posters don’t have to be horrible
实操分享：AI生成的活动海报也能做得不丑
Hacker News深挖 →
27
AI and the Destruction of the Creative Commons
AI大规模抓取内容，正在摧毁公共创作共享生态
Hacker News深挖 →

快讯

·
2 万亿归来
一石一泉一松一月一人
·
人生真的应该一直向上吗？
一石一泉一松一月一人
·
人生真的应该一直向上吗？当生活已经很难的时候，请停止自责，向那个咬牙坚持到今天的自己道一声辛苦！
一石一泉一松一月一人

日报存档

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