【AI 英文奏折】2026年11月17日|评测分歧:Qwen 4 基准、AA Index 排名

每日奏折14小时前发布 tianming
17 0 0

共收录 18 篇深度内容


📌 今日概览

【AI 英文奏折】2026年11月17日|评测分歧:Qwen 4 基准、AA Index 排名

  • 开源生态出现接入与抽象两条动作:Hugging Face 把 OpenRouter 集成进 Hub,Sebastian Raschka 提出统一模型组件 API 以降低结构切换成本。
  • 应用层把生成能力落到建站、字体、离线问答等任务,Gemini Notebook 支持一句话生成可分享网站。
  • agent 讨论集中在边界与基础设施:外部沙箱执行代码、消费级 agent 市场地图,以及召回通用 agent 的主张同时出现。
  • 模型评测口径出现分歧,Qwen 4 的纸面基准与 AA Index 结论不一致,同期 Progress Index 显示榜单最高分两年翻倍。

🔥 必读 Top3

  1. Hugging Face Hugging Face 官方宣布 OpenRouter 已直接集成进 Hugging Face:沿用同一 token 和 endpoint,Hub 上新增实时模型统计,并联合重做了推理组件。
    入选理由:官方宣布 OpenRouter 直接集成进 Hugging Face,沿用同一 token 和 endpoint,并新增实时模型统计,是开源生态当天最具体的产品动作
  2. Gemini Notebook Gemini Notebook 新增功能:在对话中输入「make me a site」并描述需求,即可把资料生成为可交互、可分享的网站,全球 18 岁以上用户可用。
    入选理由:给出「make me a site」把资料转成可交互网站的可复现用法,代表应用层把生成能力压进具体任务的方向
  3. Justine Moore a16z 发布消费级 agent 市场地图,认为消费 agent 及其基础设施供应商正在快速增长,搜索、账户接入、记忆、支付和通信工具是建设窗口。
    入选理由:a16z 消费级 agent 市场地图,指出搜索、账户接入、记忆、支付等基础设施的建设窗口,为 agent 生态提供结构视角

🧭 主题分组

  • 开源生态(3 条)—— Hub 集成 OpenRouter 与统一模型组件 API 的呼吁并行,开源侧在接入和抽象层同时推进。
  • 应用与产品(4 条)—— 生成能力被压进建站、字体、离线问答等具体任务,产品化节奏加快。
  • 模型评测(3 条)—— Qwen 4 的纸面基准与 AA Index 结论不一,评测口径成为争议焦点。
  • Agent 安全(2 条)—— 外部沙箱、召回通用 agent 与基础设施风险的主张相互呼应,安全边界成为共同关切。
  • Agent 生态(1 条)—— a16z 市场地图指出消费 agent 及搜索、账户、记忆、支付等基础设施正处建设窗口。
  • 算力芯片(1 条)—— 灰色芯片链条若被清理将重塑开源 AI 格局,供应端变化牵动模型竞争。
  • AI 投融资(1 条)—— 围绕 AMD 收购 World Labs 的人事叙事出现质疑,交易细节仍待厘清。
  • AI 安全(1 条)—— 有主张要求实验室在保护好权重与研究机密前暂停算法进展,超级智能时间表引发争论。
  • AI 政策(1 条)—— 欧洲现金转移与政府发薪比例被拿来类比后 AGI 的全面 UBI 状态。
  • AI 应用(1 条)—— 用 Claude 生成双耳节拍歌单提升专注,AI 编程浪潮冲击传统专注类产品。

📖 详细内容

开源生态

Hub 集成 OpenRouter 与统一模型组件 API 的呼吁并行,开源侧在接入和抽象层同时推进。


Hugging Face
| @huggingface
| 影响力: 84.8万粉丝 · 互动分 2946.5
The AI community building the future. https://t.co/TpiXQMQ9rZ

💡 核心观点: Hugging Face 官方宣布 OpenRouter 已直接集成进 Hugging Face:沿用同一 token 和 endpoint,Hub 上新增实时模型统计,并联合重做了推理组件。

可信度: 85/100

事实核查:

  • 原文支持: OpenRouter is now integrated into Hugging Face directly. Same token, same endpoint(由 Hugging Face 官方账号发布,明确说明集成方式与凭证复用。)
  • 原文支持: plus live model stats on the Hub and an inference widget we redesigned together(原文列出新增功能,未给出上线时间或版本细节。)

原文内容:

We're teaming up to accelerate open-source Al.
OpenRouter is now integrated into Hugging Face directly. Same token, same endpoint, plus live model stats on the Hub and an inference widget we redesigned together.
The best of open, in one place. https://t.co/fNbvtJGSJB

⏰ 18:18 | ❤️ 1975 点赞 | 📝 46词 | 入选: 高赞 1975 | 查看原文 →

↑ 返回顶部


GREG ISENBERG
| @gregisenberg
| 影响力: 76.4万粉丝 · 互动分 1073.5
I drop startup ideas daily. Host @startupideaspod. CEO: @latecheckoutplz we build companies like @ideabrowser, @meetLCA, @boringmarketer etc

💡 核心观点: 作者总结了一套开源 SaaS 增长方法:把面向客户核心任务的流程做成技能并开源,兼容主流 Agent 工具,再通过 MCP 和付费托管服务变现。

可信度: 45/100

事实核查:

  • 无法核实: open sourcing your saas is the new saas playbook(这是作者归纳的经营观点,无数据或案例支撑)
  • 原文支持: I'm open sourcing 32 of my best skills(引用推文给出具体数量,但仓库与许可需另行核验)

原文内容:

I’ve noticed open sourcing your saas is the new saas playbook 

1. pick the job your customers hire you for

2. turn your best workflows for that job into skills

3. open source them on github

4. make them work in claude code, codex, cursor and openclaw on day one

5. ship an MCP so agents can use your product directly

6. post the repo with a list of every skill inside (like this one)

7. every install puts your brand inside someone's agent, and it gets used every day

8. the free skills do the work once. the paid product does it every week, with your data, hosting and support

9. upsell to done for you, team plans and the hosted version

10. keep shipping new skills. every one is a new launch

Your free skills are the new free trial.

saas has gone open source for real

引用 @boringmarketer: I'm open sourcing 32 of my best skills that cover every marketing use case for AI agents I can think of

it helps your AI execute marketing at a world class quality focused on RESULTS and ROI

Not sure how long I'll keep these open, but you can grab them here: https://t.co/IfPLLBMbmW

this is what's…

⏰ 20:05 | ❤️ 669 点赞 | 📝 141词 | 入选: 高赞 669 / 引用讨论 / 长文 | 查看原文 →

↑ 返回顶部


Sebastian Raschka
| @rasbt
| 影响力: 53.9万粉丝 · 互动分 1070.26
ML/AI research engineer. Ex stats professor. Author of "Build a Large Language Model From Scratch" (https://t.co/O8LAAMRzzW) & reasoning (https://t.co/5TueQKx2Fk)

💡 核心观点: 作者提出开源 LLM 生态需要统一的「模型组件 API」,以降低在 GQA/MLA、稠密 FFN/MoE、注意力/Mamba2 等结构间切换的成本,并点名 torchtitan、Megatron Core 等现有项目尚未达到该抽象层级。

可信度: 62/100

事实核查:

  • 无法核实: there is a major switching cost when you want to run experiments or ablation studies(属作者基于实践的经验判断,原文未给出量化对比)
  • 原文支持: like torchtitan, torchtune, Megatron Core, OLMo-core, but they are not quite that level of abstraction(项目名称可在原文中找到,评价为主观判断)

原文内容:

The open-source LLM ecosystem could really benefit from an LLM building-block API.

We have so many LLM components across different models now. Whether you are using PyTorch or C++ and CUDA code directly, there is a major switching cost when you want to run experiments or ablation studies, say swapping GQA with MLA, dense FFNs with MoE, attention layers with Mamba2 or Gated DeltaNet, or standard residual connections with attention residuals.

Currently, it’s very messy even with LLM-assisted coding.

It could maybe be on top of PyTorch or CUDA directly, but it needs to be carefully engineered and maintained (versus quickly vibe-coded on a weekend) and efficient enough that people trust it to use in production (which is really a requirement to move beyond toy examples).

Maybe some food for thought for a big, well-funded lab :)

(There are some efforts that go into that direction, like torchtitan, torchtune, Megatron Core, OLMo-core, but they are not quite that level of abstraction, I'd say. 

And, as always, there’s the chicken-and-egg problem: it needs to be popular to get widely adopted, and it needs to get widely adopted to become popular.)

⏰ 14:28 | ❤️ 691 点赞 | 📝 198词 | 入选: 高赞 691 / 长文 | 查看原文 →

↑ 返回顶部

应用与产品

生成能力被压进建站、字体、离线问答等具体任务,产品化节奏加快。


Gemini Notebook
| @Gemini_Notebook
| 影响力: 28.3万粉丝 · 互动分 2631.5
Think smarter, not harder. Meet your brain's new best friend 📒

💡 核心观点: Gemini Notebook 新增功能:在对话中输入「make me a site」并描述需求,即可把资料生成为可交互、可分享的网站,全球 18 岁以上用户可用。

可信度: 80/100

事实核查:

  • 原文支持: Just type "make me a site" in the Chat with a description of what you want to build(原文给出具体操作指令与示例场景。)
  • 原文支持: Available globally on web for ALL (18+) users.(原文明确覆盖范围与年龄限制。)

原文内容:

You can now turn your sources into interactive, shareable websites 🤯

Just type "make me a site" in the Chat with a description of what you want to build (interactive dashboards, project trackers, etc). Once built, open it in a new tab or share it directly with others.

Available globally on web for ALL (18+) users. Share your examples below!

⏰ 21:06 | ❤️ 1801 点赞 | 📝 58词 | 入选: 高赞 1801 | 查看原文 →

↑ 返回顶部


Alex Veremeyenko
| @alex_verem
| 影响力: 10.5万粉丝 · 互动分 2598.5
open source, local ai, ai for good also know as: @alex_prompter

💡 核心观点: 作者推荐在手机上装开源应用 PocketPal AI,下载 2~4GB 的 Gemma 或 Qwen 模型,无需账号订阅,飞行模式下仍可离线作答。

可信度: 55/100

事实核查:

  • 原文支持: Install PocketPal AI from the App Store or Google Play, tap Models, and download Gemma or Qwen.(原文给出应用名、应用商店和模型名,可核验。)
  • 原文支持: It's one free app and a 2 to 4GB model, and it runs on the phone you already own.(原文给出模型体积与免费属性,未说明具体机型要求。)

原文内容:

I am begging you, put an open source AI on your phone this weekend.

It's one free app and a 2 to 4GB model, and it runs on the phone you already own. No account, no subscription, and it keeps working in airplane mode.

When your signal drops, your AI app stops working. This one keeps answering on a plane, on a train or in the middle of nowhere.

Install PocketPal AI from the App Store or Google Play, tap Models, and download Gemma or Qwen.

Setup takes about 10 minutes.

You'll be glad it's there the first time you have no signal and need an answer.

⏰ 16:59 | ❤️ 2067 点赞 | 📝 107词 | 入选: 高赞 2067 | 查看原文 →

↑ 返回顶部


Google Gemini
| @GeminiApp
| 影响力: 61.9万粉丝 · 互动分 1609.5
The Gemini app turns research into reality, bringing frontier AI experiences like Omni, Nano Banana, and more to over a billion people.

💡 核心观点: Google Gemini 称 Gemini 4 Hydrogen 能在几分钟内把草图变成可用字体,借助 Antigravity 在 Mac 版 Gemini 应用中承担技术细节,可自建字体与交互设计工具。

可信度: 72/100

事实核查:

  • 原文支持: Turn a rough sketch into a working custom font in minutes with Gemini 4 Hydrogen.(官方账号发布,产品名与功能描述明确。)
  • 原文支持: Using Google @Antigravity in the Gemini app for Mac(原文指明工具名与平台,未提供版本或发布时间。)

原文内容:

Turn a rough sketch into a working custom font in minutes with Gemini 4 Hydrogen. 

Using Google @Antigravity in the Gemini app for Mac, Hydrogen handles the technical heavy lifting so you can build custom typefaces and interactive design tools from scratch.

⏰ 16:23 | ❤️ 1107 点赞 | 📝 41词 | 入选: 高赞 1107 | 查看原文 →

↑ 返回顶部


Ethan Mollick
| @emollick
| 影响力: 42.0万粉丝 · 互动分 1055.0
Professor @Wharton studying AI. New book, Co-Existence, is out now: https://t.co/hsDYf1xU9K My Substack: https://t.co/UIBhxu4bgq

💡 核心观点: 作者对比指出 OpenAI 的 Dot 在界面体验上落后于 Muse 和 Duet,单线程对话、缺少内联生成能力、无法查看任务分配使其较难使用。

可信度: 38/100

事实核查:

  • 无法核实: The single-threaded conversation, lack of inline generative flourishes (like maps), & the inability to see what tasks it is assigned(为作者使用体验描述,原文未附截图或测试方法)
  • 无法核实: seems to be falling further behind in its interface compared to Muse and Duet(属主观比较判断,无量化依据)

原文内容:

OpenAI’s Dot is capable, but seems to be falling further behind in its interface compared to Muse and Duet. The single-threaded conversation, lack of inline generative flourishes (like maps), & the inability to see what tasks it is assigned makes it harder to use than the others.

⏰ 19:03 | ❤️ 700 点赞 | 📝 49词 | 入选: 高赞 700 | 查看原文 →

↑ 返回顶部

模型评测

Qwen 4 的纸面基准与 AA Index 结论不一,评测口径成为争议焦点。


Chubby♨️
| @kimmonismus
| 影响力: 16.0万粉丝 · 互动分 2343.5
Tech Analyst & Content Creator. Editor-in-Chief @getsuperintel – Community of 300k+ in total: 🔗 https://t.co/jHMmImmI5I //📧 [email protected]

💡 核心观点: 作者评价 Qwen 4 发布不佳:纸面基准看似与 Opus 5.5 相当,但 AA Index 呈现不同结论;并称 Qwen 被切断蒸馏一事印证了这次失败。

可信度: 45/100

事实核查:

  • 无法核实: Qwen 4 is not a good release.(属作者主观评价,原文未列出 AA Index 的具体分数与来源。)
  • 无法核实: Although it appears on paper to be on par with Opus 5.5 in key benchmarks(原文只称「纸面上」,未给出基准名称与数值。)
  • 存在偏差: The fact that Qwen is being cut off from distillation seems to illustrate this failure.(「被切断蒸馏」是作者的推论,原文未提供可核验细节。)

原文内容:

Qwen 4 is not a good release. 

Although it appears on paper to be on par with Opus 5.5 in key benchmarks, the AA Index paints a completely different picture.

The fact that Qwen is being cut off from distillation seems to illustrate this failure. https://t.co/pg6HrinyoO

引用 @ArtificialAnlys: Alibaba has launched Qwen4 Max, gaining 9 points over Qwen3.8 Max on the Artificial Analysis Intelligence Index and landing 3 points behind Kimi K3.1 Pro at ~11% lower Cost per Task

Qwen4 Max, using xhigh reasoning effort, scores 43 on the Intelligence Index, 9 points ahead of Qwen3.8 Max (xhigh) a…

⏰ 11:42 | ❤️ 1722 点赞 | 📝 46词 | 入选: 高赞 1722 / 引用讨论 | 查看原文 →

↑ 返回顶部


Artificial Analysis
| @ArtificialAnlys
| 影响力: 18.0万粉丝 · 互动分 1092.5
Independent analysis of AI

💡 核心观点: 作者统计 Artificial Analysis Progress Index 显示,2023年12月首个 GTA VI 预告片以来,榜单最高分从160升至358,491个模型中481个在此期间发布,GPT-4 Turbo 如今仅排第379位。

可信度: 82/100

事实核查:

  • 原文支持: the top score on the Artificial Analysis Progress Index has more than doubled, from 160 to 358(原文给出前后分值与时间锚点,属可核验的榜单数据)
  • 原文支持: 481 of the 491 models on the Index were released(原文陈述模型发布数量,未附独立数据源)
  • 原文支持: GPT-4 Turbo held the top score on trailer day and would now rank 379th(原文给出具体名次变化,可对照榜单复查)

原文内容:

All this AI progress before GTA VI!

Since the first GTA VI trailer in December 2023, the top score on the Artificial Analysis Progress Index has more than doubled, from 160 to 358:

➤ 481 of the 491 models on the Index were released

➤ The top score was beaten on 23 separate days, by models from @OpenAI, @GoogleDeepMind, @SpaceXSI and @AnthropicAI

➤ GPT-4 Turbo held the top score on trailer day and would now rank 379th

➤ 7 of those 23 came after the revised launch date of May 26

GTA VI launches November 19. To the AI labs, please no launches that day, the team seems to be coming down with something.

⏰ 12:26 | ❤️ 788 点赞 | 📝 99词 | 入选: 高赞 788 | 查看原文 →

↑ 返回顶部


Chubby♨️
| @kimmonismus
| 影响力: 16.0万粉丝 · 互动分 973.93
Tech Analyst & Content Creator. Editor-in-Chief @getsuperintel – Community of 300k+ in total: 🔗 https://t.co/jHMmImmI5I //📧 [email protected]

💡 核心观点: 作者认为 Qwen 4 发布显示,Anthropic 阻止模型蒸馏的做法正在拖慢中国模型的发展;他由此怀疑下一代 Kimi 的水平。

可信度: 30/100

事实核查:

  • 无法核实: Anthropic’s figuring out a way to halt distillation is genuinely hindering the development of Chinese models(这是作者对因果关系的推断,原文没有给出 Anthropic 具体措施或影响数据。)
  • 原文支持: Looking at the Qwen 4 release(原文提及 Qwen 4 发布,但没有展开任何发布细节。)

原文内容:

Looking at the Qwen 4 release, one cannot help but think that Anthropic’s figuring out a way to halt distillation is genuinely hindering the development of Chinese models.

Makes you wonder how good the next Kimi really will be.

⏰ 02:10 | ❤️ 775 点赞 | 📝 39词 | 入选: 高赞 775 | 查看原文 →

↑ 返回顶部

Agent 安全

外部沙箱、召回通用 agent 与基础设施风险的主张相互呼应,安全边界成为共同关切。


Simon Willison
| @simonw
| 影响力: 25.6万粉丝 · 互动分 837.54
Creator @datasetteproj, co-creator Django. PSF board. Hangs out with @natbat. He/Him. Mastodon: https://t.co/t0MrmnJW0K Bsky: https://t.co/OnWIyhX4CH

💡 核心观点: 作者介绍“外部沙箱”模式:编程 agent 在本机运行,但把代码放到 WebAssembly、QEMU 或 Apple Container 沙箱中执行,并给出 Claude Code、Codex 和 Pi 的演示。

可信度: 75/100

事实核查:

  • 原文支持: agents run on your Mac but execute code in a WebAssembly or QEMU or Apple Container sandbox(原文明确写出该模式及三种沙箱技术。)
  • 原文支持: with demos for Claude Code, Codex and Pi(原文点名三个工具的演示,并附有链接。)

原文内容:

I've been exploring new ways to sandbox code written by coding agents - here's a description of the external sandbox pattern, where agents run on your Mac but execute code in a WebAssembly or QEMU or Apple Container sandbox - with demos for Claude Code, Codex and Pi https://t.co/0XIhs4XYvQ

⏰ 01:21 | ❤️ 492 点赞 | 📝 51词 | 入选: 高赞 492 | 查看原文 →

↑ 返回顶部


Gary Marcus
| @GaryMarcus
| 影响力: 25.8万粉丝 · 互动分 747.59
OG GenAI Skeptic; spoke at US Senate. Warned about hallucinations in 2001. Advocating world models & neurosymbolic AI ever since. Author, Marcus on AI & 6 books

💡 核心观点: 作者主张应聚焦 agent 带来的基础设施与网络安全问题,召回可上网的通用 agent,并要求新模型接受外部科学审查;他把近期多起交通、银行和航空事故与 agent 联系起来。

可信度: 15/100

事实核查:

  • 原文支持: Recall general-purpose agents with internet access until the problems are solved.(原文明确提出召回诉求。)
  • 无法核实: It can’t all be coincidence. Maybe not all of it is from AI agents but a lot of it probably is.(作者自己承认并非全部归因于 agent,原文未提供任何因果证据。)
  • 无法核实: Amtrak ground stop today, United Airlines issues today, Friday’s FAA ground stops(原文仅罗列事件名称,没有时间、地点外的可核验细节或官方说明。)

原文内容:

⚠️ ⚠️ ⚠️ ⚠️ ⚠️  Wake the F up! ⚠️ ⚠️ ⚠️ ⚠️ ⚠️  

For months I have said, on practically every major TV network, “stop focusing on the doomers”

“Focus for now instead on the infrastructure and cybersecurity problems that agents are causing.”

“Recall general-purpose agents with internet access until the problems are solved. Require that new models have outside scientific review.”

But no. 

Instead people just played existential risk theatre, didn’t recall those agents. And attacked me relentlessly with propaganda and lies.

All we got was some voluntary “self-policing” at White House meeting. Which has done nothing.

So… Amtrak ground stop today, United Airlines issues today, Friday’s FAA ground stops (did we really get the full story on that?), banks today, French refrigeration incident a couple weeks ago, Open AI Hugging Face incident, Australian government hack, Binance and on and on, tens of thousands of incidents in the last few months.  

It can’t all be coincidence. Maybe not all of it is from AI agents but a lot of it probably is.

So yeah, quoting again @kasparov63, “I f_cking told you so.”

And we must take those agents off the market until the problems are resolved.

引用 @omaruddin: @GaryMarcus @Kasparov63 Amtrak in & out of in major city shut down
FAA outage and grounding

Banks outage

Microsoft outage

90 day warning from DoD

People asked in planes to close windows & not look out & delete photos taken

“stock up on food & supplies for 90 days!”

Tanker cras…

⏰ 22:55 | ❤️ 437 点赞 | 📝 191词 | 入选: 高赞 437 / 引用讨论 / 长文 | 查看原文 →

↑ 返回顶部

Agent 生态

a16z 市场地图指出消费 agent 及搜索、账户、记忆、支付等基础设施正处建设窗口。


Justine Moore
| @venturetwins
| 影响力: 24.9万粉丝 · 互动分 774.5
Partner @a16z AI 🤖 and twin to @omooretweets | Investor in @elevenlabs, @bfl_ai, @hedra_labs, @krea_ai, @heyglif, @ShizukuAILabs, @wabi, @TownAI

💡 核心观点: a16z 发布消费级 agent 市场地图,认为消费 agent 及其基础设施供应商正在快速增长,搜索、账户接入、记忆、支付和通信工具是建设窗口。

可信度: 55/100

事实核查:

  • 原文支持: New @a16z market map alert!(原文明确说明这是 a16z 的新市场地图。)
  • 无法核实: Consumer agents are blowing up – and so are the infra providers powering them.(增长判断未附数据,仅指向所附地图。)

原文内容:

New @a16z market map alert! 🚨

Consumer agents are blowing up - and so are the infra providers powering them.

It's the perfect time to build the tooling that helps agents search, access accounts, store memory, make payments, and communicate.

How it's evolving 👇 https://t.co/6YDuWofchL

⏰ 18:15 | ❤️ 465 点赞 | 📝 45词 | 入选: 高赞 465 | 查看原文 →

↑ 返回顶部

算力芯片

灰色芯片链条若被清理将重塑开源 AI 格局,供应端变化牵动模型竞争。


SemiAnalysis
| @SemiAnalysis_
| 影响力: 18.8万粉丝 · 互动分 662.5

💡 核心观点: SemiAnalysis 引用观点称中国境内有超过 1.5 GW 不应存在的芯片,这条灰色链条若被清理将重塑开源 AI;并称 NVIDIA 若无法继续支持开源生态,Anthropic 更可能形成垄断或双寡头。

可信度: 45/100

事实核查:

  • 无法核实: there’s more than one and a half gigawatts of chips in China that are not supposed to be there according to US law(原文以“In our view”表述,未给出统计方法或数据来源。)
  • 无法核实: Anthropic looks much more likely to get a complete monopoly or a duopoly. This bodes really well for Anthropic’s IPO next week.(属预测性判断,原文无市场份额或 IPO 细节支撑。)
  • 原文支持: They’re more diverse in what chips they use than any other company in the world.(原文直接作出该比较表述,但未说明比较对象。)

原文内容:

More than 1.5 GW of chips are in China that shouldn’t be. Unwinding that web could reshape open-source AI.

“In our view, there’s more than one and a half gigawatts of chips in China that are not supposed to be there according to US law. How they get there, that whole web, that’s what’s going to unwind.”

“If NVIDIA can no longer support the open-source ecosystem, because that was Jensen’s primary goal, does that mean the open-source ecosystem falls away?”

“Now open source looks a lot weaker with NVIDIA down and out. Anthropic looks much more likely to get a complete monopoly or a duopoly. This bodes really well for Anthropic’s IPO next week.”

“They’re more diverse in what chips they use than any other company in the world.”

⏰ 02:38 | ❤️ 370 点赞 | 📝 138词 | 入选: 高赞 370 / 长文 | 查看原文 →

↑ 返回顶部

AI 投融资

围绕 AMD 收购 World Labs 的人事叙事出现质疑,交易细节仍待厘清。


Fei-Fei Li
| @drfeifei
| 影响力: 109.9万粉丝 · 互动分 1118.31
Cofounder/CEO @theworldlabs, Prof (CS @Stanford), Co-Director @StanfordHAI, #AI #SpatialIntelligence #GenAI #computervision #robotics #AI-healthcare

💡 核心观点: 作者质疑 AMD 收购 World Labs 的人事叙事

可信度: 30/100

事实核查:

  • 原文支持: The @theworldlabs team to join @AMD(原文只确认了团队加入AMD的表述,未披露交易规模、股权或官方命名,无法核验具体安排)

原文内容:

AI is a civilizational technology! There are tremendous opportunities and responsibilities for all of us at this moment in history. I'm deeply grateful and excited to lead the @theworldlabs team to join @AMD to embark on a new beginning, to advance science, develop technologies, and ultimately make a positive impact to the world! 😍🚀

引用 @AMD: Queen of Chips, @LisaSu 🤝 Godmother of AI, @drfeifei

A picture is worth a thousand words and these 📸 from Lisa Su’s visit to San Francisco is giving us all the feels. We are so excited to officially welcome the @theworldlabs team to AMD! https://t.co/mYpmznhGlg

⏰ 17:15 | ❤️ 749 点赞 | 📝 54词 | 入选: 高赞 749 / 引用讨论 | 查看原文 →

↑ 返回顶部

AI 安全

有主张要求实验室在保护好权重与研究机密前暂停算法进展,超级智能时间表引发争论。


Dwarkesh Patel
| @dwarkesh_sp
| 影响力: 28.9万粉丝 · 互动分 2034.5
Host of @dwarkeshpodcast https://t.co/3SXlu7fy6N https://t.co/4DPAxODFYi https://t.co/hQfIWdM1Un

💡 核心观点: 引用 Paul Christiano 观点:按当前轨迹,一年内出现足以推翻政府的超级智能的概率为三分之一;他主张实验室在保护好权重与研究机密前应暂停算法进展。

可信度: 62/100

事实核查:

  • 原文支持: a 1 in 3 chance that we could have superintelligence powerful enough to overthrow the government within a year(原文明确给出该概率判断并称有具体情景,推文未展开计算依据。)
  • 原文支持: the labs should pause further algorithmic progress until they can secure their weights and research secrets(原文直接复述该主张,属引用内容。)

原文内容:

.@paulfchristiano thinks that on the current trajectory (absent any slowing down), there is a 1 in 3 chance that we could have superintelligence powerful enough to overthrow the government within a year. (He articulates a concrete scenario).

He argues that the labs should pause further algorithmic progress until they can secure their weights and research secrets. Otherwise China could simply steal the superintelligence in the middle of the intelligence explosion, and the labs would no longer be able to slow down even if they saw their models going off the rails.

He also explains how AIs will be trained during the intelligence explosion, and why even a system trained entirely inside a data center could still have the capabilities profile such that humans would have very little chance of competing with it or controlling it.

And he discusses the plan for alignment - basically, making a superintelligence that actually does what we ask the way we mean it. Fixing reward hacks and building better graders will help up till the next few model generations, but he thinks that no single current approach has better than even odds of letting us control superintelligence. The hope is that the AIs figure out the solution to this problem during the intelligence explosion.

Paul is head of the Alignment Research Center. Previously, he was head of AI safety at the US AI Safety Institute. Before that, he led alignment at OpenAI, where he invented RLHF. He’s recently joined OpenAI’s board (though here he is speaking as a private citizen).

⏰ 18:10 | ❤️ 1270 点赞 | 📝 254词 | 入选: 高赞 1270 / 长文 | 查看原文 →

↑ 返回顶部

AI 政策

欧洲现金转移与政府发薪比例被拿来类比后 AGI 的全面 UBI 状态。


@levelsio
| @levelsio
| 影响力: 99.3万粉丝 · 互动分 985.0
💸https://t.co/sQ0aiU7v02 📸https://t.co/lAyoqmSBRX $68K/m 🎮https://t.co/tF49atJ396 $44K/m 🏡https://t.co/1oqUgfD6CZ $23K/m 👙@X $20K/m 🌍https://t.co/UXK5AFqCaQ $7K/m 🏩https://t.co/4p4dzTDVN6 $3K/m 💾@pieter 📝https://t.co/8VnHmB3uah

💡 核心观点: 作者引用关于欧洲社会保障的讨论,称约45%的欧洲人依赖养老金或福利等政府现金转移,约20%直接由政府发薪,认为欧洲已接近后 AGI 全面 UBI 的状态。

可信度: 28/100

事实核查:

  • 无法核实: ~45% of Europeans are dependent on government cash transfers like pensions or welfare(原文给出来源不明的比例,未标注统计口径与年份)
  • 无法核实: ~20% of Europeans directly get their salary from the government (public sector jobs)(比例无出处,需对照官方就业统计)

原文内容:

~100% of Europeans are dependent on government funded services 

~45% of Europeans are dependent on government cash transfers like pensions or welfare

~20% of Europeans directly get their salary from the government (public sector jobs)

They're essentially already were they'll be in a post-AGI world with full UBI

引用 @aubetony: @levelsio They have among the best safety nets to prevent the shock and societal collapse of everyone suddenly losing their jobs and income

⏰ 12:59 | ❤️ 753 点赞 | 📝 48词 | 入选: 高赞 753 / 引用讨论 | 查看原文 →

↑ 返回顶部

AI 应用

用 Claude 生成双耳节拍歌单提升专注,AI 编程浪潮冲击传统专注类产品。


Prompter
| @PromptLLM
| 影响力: 9.9万粉丝 · 互动分 854.4
Teaching those how to ask AI the right questions to transform every aspect of their life.

💡 核心观点: 作者称可以用 Claude 生成叠加在音乐上的双耳节拍歌单来提升专注力,并认为 Brain FM 这类产品正被 AI 编程浪潮冲击。

可信度: 25/100

事实核查:

  • 无法核实: You can get Claude to whip up a binaural beats playlist that is overlayed by actual music(原文只描述用法,没有给出生辰、链接或实际输出示例。)
  • 无法核实: Which then gives you adderall like focus(专注效果是个人体验式说法,原文无任何测量或对照依据。)
  • 存在偏差: Brain FM is the latest SaaS to fall to the vibe coding era(原文未提供 Brain FM 被替代或衰退的任何证据,属于夸张判断。)

原文内容:

Blows my mind that if you are struggling to concentrate 

You can get Claude to whip up a binaural beats playlist that is overlayed by actual music 

Which then gives you adderall like focus 

Brain FM is the latest SaaS to fall to the vibe coding era

⏰ 15:05 | ❤️ 712 点赞 | 📝 47词 | 入选: 高赞 712 | 查看原文 →

↑ 返回顶部

🔗 引用关系

  • @levelsio 引用 @levelsio(原文:Europeans are by far the most prepared for post-AGI of anybody because they're already so used to not working)
  • @kimmonismus 引用 @ArtificialAnlys:作者评价 Qwen 4 发布不佳:纸面基准看似与 Opus 5.5 相当,但 AA Index 呈现不同结论;并称 Qwen 被切断蒸馏一事印证了这次失败。(原文:Alibaba has launched Qwen4 Max, gaining 9 points over Qwen3.8 Max on the Artificial Analysis Intelligence Index and land…)
  • @aakashgupta 引用 @GaffarYakinca(原文:Havacılık ve turizmde artık hiçbir şey eskisi gibi olmayacak. Çünkü Çin, yepyeni bir dönemin kapısını araladı.

    Dünyanın…)

  • @steipete 引用 @MacRumors(原文:macOS 27.3 Warns Against Giving AI Apps Full Mac Control https://t.co/Ko2jRzlEOl https://t.co/N9DsvPRis6)
  • @drfeifei 引用 @AMD:作者质疑 AMD 收购 World Labs 的人事叙事(原文:Queen of Chips, @LisaSu 🤝 Godmother of AI, @drfeifei

    A picture is worth a thousand words and these 📸 from Lisa Su’s vis…)

  • @gregisenberg 引用 @boringmarketer:作者总结了一套开源 SaaS 增长方法:把面向客户核心任务的流程做成技能并开源,兼容主流 Agent 工具,再通过 MCP 和付费托管服务变现。(原文:I'm open sourcing 32 of my best skills that cover every marketing use case for AI agents I can think of

    it helps your A…)

  • @levelsio 引用 @aubetony:作者引用关于欧洲社会保障的讨论,称约45%的欧洲人依赖养老金或福利等政府现金转移,约20%直接由政府发薪,认为欧洲已接近后 AGI 全面 UBI 的状态。(原文:@levelsio They have among the best safety nets to prevent the shock and societal collapse of everyone suddenly losing th…)
  • @GaryMarcus 引用 @omaruddin:作者主张应聚焦 agent 带来的基础设施与网络安全问题,召回可上网的通用 agent,并要求新模型接受外部科学审查;他把近期多起交通、银行和航空事故与 agent 联系起来。(原文:@GaryMarcus @Kasparov63 Amtrak in & out of in major city shut down
    FAA outage and grounding

    Banks outage

    Microsof…)

🧩 跨推文归纳

共识

  • 多位作者认为当前榜单与基准不能完整反映模型真实水平,评测口径需要外部审查
  • agent 的能力扩张与安全边界同时成为关注点,外部沙箱与风险约束被反复提及

分歧

  • 对 Qwen 4 的判断分歧明显:有观点称纸面基准与 Opus 5.5 相当、失败源于蒸馏被切断,也有观点认为 AA Index 给出不同结论,Anthropic 阻止蒸馏正在拖慢中国模型

数据来源:X 列表「AI 英文」(22 条候选)· 由 twscrape 采集、grok 整理

© 版权声明

相关文章

暂无评论

none
暂无评论...