Any%

株 機械学習 SF メモ

[Noe1017 @Noe1017_Monad]

所谓优雅的数学结构，本质上是人类为了迁就自己那点可怜的脑容量，被迫发明出来的拐杖。

一个顶尖人类数学家的工作内存，撑死也就同时追踪五六个变量的交互。

当你在算一个偏微分方程或者组合不等式时，一旦展开到三阶以上，几十个交叉项涌出来，人脑就直接宕机了。

所以人类数学家极其厌恶展开高阶项，一旦看到高阶项发散，人类的本能反应是此路不通，赶紧退回来找对称性、找不变量、搞同调代数或者几何直观，试图不展开式子就把问题绕过去。

但 AI 没有这个生理缺陷。

对于上万个并发的 Agent 来说，同时追踪一万个五阶交叉项的系数变化，跟我们算 1+1 一样轻松。

在它的视野里，根本不需要什么几何直观，它直接站在高维的代数展开式里往下看，人类觉得像乱麻一样的百项高阶余项，在它眼里就是一张明明白白的线性方程组。

既然把式子硬展开到第七阶，再凑几个待定系数就能让所有讨厌的发散项精准归零，它为什么要费劲去帮你发明一个人类能看懂的新概念？

这 700 个结果背后最无奈的黑色幽默：我们终于拿到了上帝账本里的答案，却发现上帝算账根本不用微积分，他老人家是直接拿计算器硬按出来的。

[Dream Labs @dreamlabs_ai]

Today, Dream Labs is coming out of stealth with Dream-1, a robot foundation model that controls bimanual robots it has never seen, on tasks it was never shown, in places it has never been, from $3/hour.

Dream API is live for everyone, today. 🧵

[Arian van Putten @ProgrammerDude]

“How did you keep the development of this secret for so long?”

“I didnt. I simply told everyone and nobody believes us”

Lmao

[ChatGPT @ChatGPT]

A few big upgrades for Chrome users today! We’re bringing the power of computer use to anyone who uses Chrome.

First up: ChatGPT can now use your tabs in Chrome to make answers more useful and get things done for you.

If you’re in Chat, it can look across your open tabs and gather info.

If you’re in Work, ChatGPT can control your Chrome tabs to get things, like sending messages, filling out forms, or downloading files. It can even help you organize and group that pile of tabs you have open.

[Any% @ouchideikiteru]

何兆円か出してGPU買ってデータセンターつくって、あらゆる手段使ってトップAIラボから人材引き抜いて最先端に食らいつく努力をするのが科学と資本主義に則った正しいAI開発であって、しょぼい資源としょぼい人材でしょぼい国産モデル作ってるのはただの遊びだろという気がしないでもない

[Pushmeet Kohli @pushmeet]

Very happy to announce that our team @GoogleDeepmind has pushed the boundaries of generative biology, achieving the successful synthesis of AI-designed proteins that are both functional and watermarked.

This proof-of-concept watermarking of the building blocks of life is enabled by SynthID Bio, our new protein watermarking method. It is designed to safeguard the new era of AI-powered generative biology and strengthen global biosecurity.

You can read my thoughts here on why watermarking AI-designed proteins is an important research breakthrough: https://t.co/dfoSdh8ynz

[Google DeepMind @GoogleDeepMind]

Introducing Gemini 4 Argon – our new frontier model.

It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.

[imjustnewatai @imjustnewatai]

Yeah this was the most disappointing OpenAI dev day, OpenAI lost the plot and I am SOOOO disappointed in them, Anthropic will be seeing my 200 dollars instead

[Eric Daugherty @EricLDaugh]

🚨 JUST IN: By stroke of pen and flanked by Cabinet members, President Trump just signed an executive order directing every federal agency in the United States to IMMEDIATELY begin combining into America .gov — a Super Intelligence-powered page letting Americans complete every interaction with the federal government in ONE SEARCH, ONE FORM

NO MORE going to multiple agency sites and looking at DOZENS of different pages to accomplish one thing.

THIS IS HUGE!

Many Americans are about to have HOURS freed up 🇺🇸

[Eric Daugherty @EricLDaugh]

🚨 NOW: The Trump administration is launching a Super Intelligence site that will let Americans change their last name after marriage in ONE SINGLE FORM — which usually requires HOURS or DAYS to fix every government document

"You'll be able to change your name right on the platform."

"This would've taken weeks, multiple filings, different agency websites, if you even knew where to start."

This is perfect!

Saving Americans' time is INVALUABLE

[ScottieFox @ScottieFoxTTV]

Opensource. Lite & Full version. 💦 @threejs
Use as you wish. Github:
https://t.co/bCh71K77s9

[MAX @maxxmalist]

The race for AGI

Script: Sherpa by Pocket FM
Video: Seedance 2.5

[Dr. Eli David @DrEliDavid]

SpaceX IPO: $1.75 trillion
- Reusable giant spaceships
- Largest satellite network
- Owns huge AI data center
- Fontier LLM
- Owns X

Anthropic IPO target: $2 trillion
- Frontier LLM
- Performance matched by Chinese open source
- Rapidly losing market share

Make it make sense.

[Kagurazaka Shinkai @kk_shinkai]

一次 Codex Reset 的价值远远比想象中低。

这里以 20x Pro、$200/月的基准讨论，只计算额度的名义价值，不讨论 OpenAI 的实际成本。

一周的额度名义上约值 $46，但这并不意味着每次 reset 都等于白送 $46。原因很简单：reset 会同时延后下一次正常重置的时间。

只有一种极端情况，你才能真正拿到完整的 $46：额度刚刚重置，你在一瞬间用光 100%，然后 Codex 又恰好立刻 reset。
假如你分别需要 1、3、5 天才能消耗完 100% 的额度，并且恰好在额度耗尽后发生 reset，那么这次 reset 带来的额外价值分别约为 $39、$26、$13。这可以近似理解为一次 banked reset 的价值。

对于非 banked reset，假设它在一周内随机出现，那么期望价值正好约为对应 banked reset 的一半：如果你通常需要 1、3、5 天烧完额度，那么一次随机 reset 的期望价值分别约为 $20、$13、$7。

如果 reset 提前有预兆，用户可以刻意多消耗一些剩余额度，因此价值会有所提高，但通常仍然远低于 $46。

更重要的是，reset 甚至可能是负收益。
如果你的工作负载有明显的波峰和波谷，极端情况下，假设前六天几乎没有使用额度，原本准备在第七天集中用掉 100%，结果第六天突然 reset，那么原本积累下来的六天重置周期被直接清零，相当于损失了接近 $40 的名义额度价值。

再假设两种用户：一种人在前半周使用 30%，后半周使用 70%；另一种恰好相反。那么随机发生一次 reset，前者平均损失约 $5，后者平均获得约 $5。假设二者数量相等，混合之后 reset 的期望价值接近 0。

当然，这些都只是理想化模型。真实用户既不会完美控制自己的额度消耗，也未必按照均匀计划使用。

多账号用户因为更容易产生闲置额度，reset 带来的收益可能稍高；另一方面，还有大量用户本来就烧不完额度，或者习惯性保留额度。对于这些没有遭遇额度限制的人来说，reset 本身几乎没有实际价值。

上面的数字没有考虑真实用户的行为。有些假设会高估 reset 的价值，有些则会低估。

但综合来看，我猜对于 20x Pro 订阅者，一次随机 reset 给单个用户带来的平均额外价值很可能低于 $5，甚至远远更低。
相对于 $200 的月费，这基本上可以忽略。

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[深夜漫遊放送 @shinya_manyu]

Re：ゼロから始める異世界学園生活・序（同人アニメ）

-#リゼロ #Reゼロ #菜月昴 #ベアトリス #Reゼロファンアニメ

[ろてじん @rotejin]

GPT‑6 Astra × Marvelous Designerでトマリの衣装を作ってみた。カーディガンのマテリアルも自分で画像生成して作ってた。Marvelous Designer APIと公式チュートリアルのURLを渡せばそれに沿って作れるみたい。全く手直し無し（リボンとかズレてるけど）凄い時代だ。

[The A. I. Generalist @TheAIGeneralist]

@sama So the public gets screwed and the superior tech stays behind closed doors and available to a select few who control the world. You also get to charge whatever you want because nobody is allowed to sell a better product. This is bs anti-trust manipulation disguised as ethics.

[Antoine Tilloy @AntoineTilloy]

Je trouve la lettre des 25 médaillés Fields assez triste, elle vieillira mal. Les meilleurs problem solvers de notre génération nous expliquent qu'en fait, résoudre des problèmes n'était pas si important, et que ce qui comptait, c'était les amis qu'on se faisait sur le chemin.