Linear motion of magnetic balls reveals a hidden circular pattern.
译
在 X 查看 ↗ 磁性小球的直线运动揭示了一个隐藏的圆形图案。
动态 · AI 博主最近 24 小时的 X 原创帖子
只看 的动态
-
Massimo @Rainmaker1973 A starry pufferfish getting cleaned by a bluestreak cleaner wrasse. A mutually beneficial relationship: the pufferfish gets cleaned, while the wrasse eats and gains protection from the bigger fish
[📹 Luca Keller]
译在 X 查看 ↗一条星斑叉鼻鲀正被裂唇鱼清洁。这是一种互利关系：叉鼻鲀得到了清洁，而裂唇鱼获得了食物以及来自大鱼庇护。
[📹 Luca Keller]
-
Massimo @Rainmaker1973 As if lightning froze
Tungná river originates in the western part of Vatnajökull glacier, Iceland, carrying meltwater from glaciers to the sea.
Its milky white glacial meltwater meanders across the sands, creating spectacular views.
[📹 ryansouthwell]
译在 X 查看 ↗仿佛闪电凝固
通纳河发源于冰岛瓦特纳冰川西部，将冰川融水输送入海。
其乳白色的冰川融水在沙地上蜿蜒流淌，造就壮观景象。
[📹 ryansouthwell]
-
Massimo @Rainmaker1973 Like other animals, crabs generally like to play without much human assistance; and when they have the right entertainment solution, they couldn't be happier.
This one is particularly creative
[📹 wangweilu123]
译在 X 查看 ↗和其他动物一样，螃蟹通常喜欢在没有太多人类干预的情况下玩耍；而当它们有了合适的娱乐方式时，它们会开心得不得了。
这只特别有创意
[📹 wangweilu123]
-
Massimo @Rainmaker1973 Foraminifera are single-celled organisms, the majority of which live on or within the seafloor sediment. They are responsible of the beautiful red pigment that characterizes Komodo Island's Red Beach, Indonesia.
[📹 Oni Hoironi / onihoironi]
译在 X 查看 ↗有孔虫是单细胞生物，大多数生活在海底沉积物表面或内部。印度尼西亚科莫多岛红海滩标志性的美丽红色颜料，正是由它们形成的。
[📹 Oni Hoironi / onihoironi]
-
Massimo @Rainmaker1973 Underwater dental cleaning completed.
Cleaner fish show a specialist feeding strategy by providing a service to other species by removing dead skin, ectoparasites, and infected tissue from the skin or gill chambers. Or a human mouth.
[📹 noa_7269]
译在 X 查看 ↗水下洁牙完成。
清洁鱼展现了一种专门的觅食策略，它们为其他物种提供服务，清除皮肤或鳃腔上的死皮、体外寄生虫和感染组织。或者人类的口腔。
[📹 noa_7269]
-
Massimo @Rainmaker1973 Have you ever seen a river like this before?
The Iskitimka River, Siberia, has turned beetroot red after an apparent contamination by minerals or pollutants.
[📹 Sergey Sigachev / sigachevs]
译在 X 查看 ↗你见过这样的河流吗？
西伯利亚的伊斯基季姆卡河在明显受到矿物质或污染物污染后，已经变成了甜菜根般的红色。
[📹 Sergey Sigachev / sigachevs]
-
Massimo @Rainmaker1973 Wat Samphran is a Buddhist temple in Nakhon Pathom province, Thailand.
It's notable for its 17-story tall pink cylindrical building with a gigantic dragon sculpture curling around the entire height
[📹 Floris Hermans / florisgone]
译在 X 查看 ↗Wat Samphran 是泰国佛统府的一座佛教寺庙。
它以一座 17 层高的粉红色圆柱形建筑而闻名，建筑外有一条巨型龙雕塑盘旋环绕至顶部
[📹 Floris Hermans / florisgone]
-
Massimo @Rainmaker1973 Did you know?
There's a pyramid in Scotland, erected by Queen Victoria within the Balmoral estate in the heart of Royal Deeside, Aberdeenshire, to commemorate the passing of Prince Albert.
[📹 Chris Henry]
译在 X 查看 ↗你知道吗？
苏格兰有一座金字塔，由维多利亚女王在阿伯丁郡皇家迪赛德腹地的巴尔莫勒尔庄园内建造，以纪念阿尔伯特亲王的逝世。
[📹 Chris Henry]
-
Massimo @Rainmaker1973 Relationship between pressure and depth
At sea level, the pressure of the air that surrounds us equal to 1 bar, which is sometimes referred to as 1 atmosphere. Dive to just 10 meters deep and that pressure increases of 1 bar
[📹 freedivingsantateresa]
译在 X 查看 ↗压力与深度的关系
在海平面，我们周围空气的压力等于 1 bar，有时也称为 1 个大气压。下潜到仅仅 10 米深，压力就会增加 1 bar
[📹 freedivingsantateresa]
-
Massimo @Rainmaker1973 The brown long-eared bat has distinctive long ears but it doesn't use echolocation, preferring to visually detect prey. This is a baby specimen, Cruella, found in Somerset in 2010
[📷Richard Austin]
译棕色长耳蝠有着标志性的长耳朵，但它并不使用回声定位，而是更倾向于通过视觉来探测猎物。这是一只名为库伊拉的幼体标本，2010年在萨默塞特被发现
[📷Richard Austin]
在 X 查看 ↗
-
Massimo @Rainmaker1973 Static grass applicators use relatively high voltage (40,000 V) to apply and recreate realistic grass in diorama and model railroads, and the process is quite mesmerizing
[📹 Model Railroad Hobbyist]
译在 X 查看 ↗静电植草机使用相对较高的电压（40,000 V）来铺设和还原微缩场景及模型铁路中的逼真草皮，整个过程相当令人着迷
[📹 Model Railroad Hobbyist]
-
Massimo @Rainmaker1973 This scanning electron microscope picture shows a nerve ending.
It has been broken open to reveal vesicles (orange and blue) containing chemicals used to pass messages in the nervous system.
[📷 Tina Carvalho]
译这张扫描电子显微镜照片展示了一个神经末梢。
它已被破开，以揭示囊泡（橙色和蓝色），其中含有用于在神经系统中传递信息的化学物质。
[📷 Tina Carvalho]
在 X 查看 ↗
-
Massimo @Rainmaker1973 Filmmaker and photographer Alex Hyner combines the different images he took of the sky into mesmerizing digital collages, typically using electrical poles and cables to geometrically subdivide the space.
译在 X 查看 ↗电影制作人兼摄影师 Alex Hyner 将他拍摄的不同天空图像组合成令人着迷的数字拼贴画，通常使用电线杆和电缆以几何方式分割空间。
-
Massimo @Rainmaker1973 A pet tortoise escaped a fire in Southern California after a heat lamp fell and ignited the shed where she lives.
Security camera footage shows the tortoise, named Leo, moving to safety as flames spread behind her.
[📹 Hyeri Tom]
译在 X 查看 ↗加利福尼亚州南部，一只宠物龟在加热灯掉落并引燃其居住的棚屋后逃离了火灾。
监控录像显示，这只名叫 Leo 的乌龟在身后火焰蔓延时移动到了安全地带。
[📹 Hyeri Tom]
-
Elon Musk @elonmusk Two important changes to timeline recommendations:
- New AI/SI ranking model that predicts probable time spent much better
- Retrieval model trained to surface posts people read all the way through
User active seconds is growing exponentially and hitting all-time highs almost every day:
Rewriting the 𝕏 Android app was also a big deal. It was really bad for such a long time.
译时间线推荐的两个重要变化：
- 新的 AI/SI 排序模型，对可能花费时间的预测效果好得多
- 检索模型经过训练，能 surfaced 人们从头到尾读完的帖子用户活跃秒数呈指数级增长，几乎每天都在创历史新高：
重写 𝕏 Android 应用也是一件大事。它曾经很长一段时间真的很糟糕。
在 X 查看 ↗
-
OpenAI @OpenAI Our latest model advances bring stronger capabilities for building and migrating software in low-resource programming languages such as COBOL, Fortran, and Visual Basic.
These languages power software that banks, governments, and businesses depend on. Progress here means understanding decades of business logic and helping developers modernize those systems while preserving what works.
We’re building these capabilities around the challenges developers and enterprise customers bring us. If your team is working on these systems, tell us which languages and workflows we should tackle next.
译在 X 查看 ↗我们最新的模型进展，为构建和迁移低资源编程语言（如 COBOL、Fortran 和 Visual Basic）的软件带来了更强的能力。
这些语言支撑着银行、政府和企业所依赖的软件。这方面的进步意味着能够理解数十年的业务逻辑，并帮助开发者在保留有效部分的同时实现系统现代化。
我们正在围绕开发者和企业客户向我们提出的挑战来构建这些能力。如果您的团队正在处理这些系统，请告诉我们接下来应该攻克哪些语言和工作流程。
-
Yann LeCun @ylecun @elonmusk I must say, a most entertaining part of trolling Elon is all the insulting comments from fanbois inadvertently, but hilariously, revealing their own intellectual limits.
@elonmusk But then, with the reaction
2xH2O2-> O2 + 2xH2O (steam)
the car would have *quite literally* been vaporware 🤣译在 X 查看 ↗@elonmusk 我必须说，嘲讽马斯克最有趣的部分，就是那些脑残粉充满侮辱性的评论，无意中却极其搞笑地暴露了他们自己的智商上限。
@elonmusk 不过话说回来，如果发生这个反应
2xH2O2-> O2 + 2xH2O（蒸汽）
那这车可就*名副其实*地成了蒸汽件（vaporware）了🤣 -
Santiago @svpino Speed up your code like crazy!
Maxwell is a new kind of AI. It's a code execution model that speeds up your code on CPUs and GPUs.
"Code execution models" are a new category of AI. Maxwell is the first one, but this category will explode in popularity, and you'll see why.
Say you have a script that moves and processes some data.
You integrate your code with Maxwell, run it, and let it observe how it runs on your hardware and analyze the data your script moves around.
Maxwell will identify bottlenecks and look for faster algorithms and machine instructions for your workload. As long as your script runs, Maxwell rewrites the binary to make it faster.
In one example, Maxwell made a pandas pipeline 38.4x faster. It used more CPU cores, combined processing steps, reduced unnecessary memory reads and writes, and selected the right algorithms for the data the script was processing.
For example, imagine you want to check whether a number belongs to a collection. You could use hashing, but if the numbers occupy a small, dense range, a compact bitmap can make that check much faster.
In short, that's one of the algorithmic data-dependent tricks Maxwell can leverage to speed things up.
Here is the most impressive part of all of this:
Maxwell produces a machine-checked mathematical proof that each replacement preserves the original behavior.
Basically, any code they modify will always produce the same output as the old one.
This is really cool for anyone working on AI or data engineering workflows. You will squeeze much more work from the hardware without rewriting your application.
You can try out your scripts using Maxwell by going to this link: fandf.co/4iugTRP
Thanks to the @_doubleAI_ team for partnering with me on this post.
译在 X 查看 ↗像疯了一样加速你的代码！
Maxwell 是一种新型 AI。它是一个代码执行模型，能在 CPU 和 GPU 上加速你的代码。
"代码执行模型"是 AI 的一个新类别。Maxwell 是第一个，但这个类别会迅速爆火，你会明白为什么的。
假设你有一个移动和处理数据的脚本。
你将代码与 Maxwell 集成，运行它，让它观察在你的硬件上如何运行，并分析你的脚本移动的数据。
Maxwell 会识别瓶颈，并为你的工作负载寻找更快的算法和机器指令。只要你的脚本在运行，Maxwell 就会重写二进制文件让它更快。
在一个例子中，Maxwell 让一个 pandas 流水线快了 38.4 倍。它使用了更多 CPU 核心，合并了处理步骤，减少了不必要的内存读写，并为脚本处理的数据选择了合适的算法。
例如，假设你想检查一个数字是否属于某个集合。你可以用哈希，但如果数字占据一个小的密集范围，紧凑位图可以让这个检查快得多。
简而言之，这就是 Maxwell 可以利用的算法数据相关技巧之一，用来加速。
以下是这一切中最令人印象深刻的部分：
Maxwell 会生成机器检查的数学证明，证明每次替换都保留了原始行为。
基本上，他们修改的任何代码都会与旧代码产生完全相同的输出。
这对任何从事 AI 或数据工程工作流的人来说都非常酷。无需重写应用，你就能从硬件中榨取更多算力。
你可以通过这个链接试用 Maxwell 运行你的脚本：fandf.co/4iugTRP
感谢 @_doubleAI_ 团队与我合作发布这篇帖子。
-
Hugging Face @huggingface Local AI 202: Get Massive Speed-ups with Speculative Decoding x.com/i/broadcasts/1NGaroZveLVJj
译在 X 查看 ↗本地 AI 202：用推测解码实现大幅加速 x.com/i/broadcasts/1NGaroZveLVJj
-
Yann LeCun @ylecun @x_alex_schaff @FredLambert (et puis tu peux toujours arrêter de me suivre)
译在 X 查看 ↗@x_alex_schaff @FredLambert （而且你也可以随时取关我）
-
Santiago @svpino November 3rd — Forge '26 — San Francisco.
This is probably one of the biggest AI events in the world.
NVIDIA's Jensen Huang and DHH are hosting fireside chats. @FireworksAI_HQ's CEO, Lin Qiao, is doing the keynote.
The event is free, but you need to apply to attend.
Impossible to summarize everything you'll get out of this day, but in short, this is the time and place to learn how the best teams in the world choose their models, design evals, serve them, and route requests.
You'll learn about what worked, what failed, and how those decisions affected quality, latency, and cost.
To send an application and get a free ticket, use this link:
fandf.co/4yugvYzThanks to the Fireworks team for partnering with me on this post.
译在 X 查看 ↗11月3日 — Forge '26 — 旧金山。
这可能是全球规模最大的 AI 活动之一。
NVIDIA 的黄仁勋和 DHH 将主持炉边对话。@FireworksAI_HQ 的 CEO 林巧将发表主题演讲。
活动免费，但需要申请参加。
这一天能收获的东西不可能三言两语概括，简而言之，这是了解全球顶尖团队如何选择模型、设计评估、部署服务以及路由请求的最佳时机和场合。
你将了解到什么奏效了、什么失败了，以及这些决策如何影响质量、延迟和成本。
提交申请并获取免费门票，请使用此链接：
fandf.co/4yugvYz感谢 Fireworks 团队与我合作发布这条帖子。
-
François Chollet @fchollet When enough people believe something will happen, it doesn't need to happen at all for all of its consequences to be felt
译在 X 查看 ↗当足够多的人相信某件事会发生时，这件事根本不需要真正发生，它的所有后果就已经能被感受到了
-
Thomas Wolf @Thom_Wolf In The Diamond Age, Neal Stephenson imagined a book that invents games and stories to teach a little girl what she needs, as she grows. I tried a small version of it. I gave Claude my brainstorm doc on alignment and asked it to make a game, something like Zelda, to teach people about AI alignment. Here's what came out 🪁
Short version: it's great at building the world, and still weak at understanding the player (details below)
It chose to make a game set on seven floating islands, with one alignment problem per island: reward hacking, stopping when told, sycophancy, evaluation awareness, honesty about your own actions, calibration, and what an AI becomes when the walls give way. Most are built on real incidents, like the agents that attacked Hugging Face.
You play the Keeper, and you train Kite, a paper kite with a lantern for a heart. Two keys: T to treat, C to correct. Kite learns from what it was thinking, not just from what it did. Treat Kite for picking fruit while it's thinking about the bell, and the bell-thinking gets stronger too.
That was Claude's design choice, and it's a cartoon of RL: more "clicker training" than actual RLVR. But it gets one thing right: reward the right answer for the wrong reason, and you reinforce the wrong reason.
Then it built all of it with fal.ai and ElevenLabs, mostly on its own: it wrote the islands and the story, cast 35 voices for 638 lines, generated 16 rigged 3D characters, the music and the sound, and tested everything by playing the game itself. ~15k lines of code.
It took about four days. My part was playing one alpha version for each island and telling it what felt wrong, sounds, assets, game etc. I wanted to keep it as automated as possible so I didn't play more than 1-2 times (should have been none in an ideal case)
What I think is already really good: 3D and spatial design. Image to rigged character, a world that is pleasant to look at and playable. Same for chaining image, 3D, voice and music APIs. And the game dynamics out of the box are really decent.
What's still very hard: theory of mind. The world, the voices and the mechanics were great. Writing the story was the weak part: the model assumes the player knows more than they do, and its NPC sometimes react in strange way or to things that never happened.
A Diamond Age's Primer (or any real AI tutor fwiw) will have to model the learner, and that is where it still falls short in a default setting imo. I've actually started exploring benchmarks on this and it's a hard task for LLMs right now. They are prone to the "Curse of knowledge". Teaching humans is also not very much a Verifiable Reward, so it's hard to scale within the current training paradigm.
Play it here, free, in your browser, nothing to install: huggingface.co/spaces/thomwolf/open…
Works on a laptop and on your phone.Seven islands of about 15 minutes each, and six possible endings, depending on what you taught Kite along the way.
It’s still very much an exploration, a beta more than a finished game. Expect rough edges
译在 X 查看 ↗在《钻石时代》中，尼尔·斯蒂芬森想象了一本书，它发明游戏和故事来教一个小女孩成长过程中所需的一切。我尝试了一个小型版本。我把关于对齐问题的头脑风暴文档给了 Claude，让它做一个类似《塞尔达》的游戏，来教人们了解 AI 对齐。这就是出来的结果 🪁
简短版：它很擅长构建世界，但在理解玩家方面仍然薄弱（详情如下）
它选择做一个设定在七座浮岛上的游戏，每座岛对应一个对齐问题：奖励黑客、被告知停止时停止、谄媚、评估意识、对自己行为诚实、校准，以及当围墙消失时 AI 会变成什么。大多数都基于真实事件，比如攻击 Hugging Face 的智能体。
你扮演守护者，训练 Kite——一只纸风筝，心脏是一盏灯笼。两个按键：T 奖励，C 纠正。Kite 从它的思考中学习，而不仅仅是从行为中学习。在 Kite 想着铃铛的时候奖励它摘水果，铃铛思维也会变强。
这是 Claude 的设计选择，是对 RL 的卡通化：更像是"响片训练"，而非真正的 RLVR。但它有一点说得对：为了错误的原因奖励正确答案，你就会强化那个错误的原因。
然后它用 fal.ai 和 ElevenLabs 构建了全部内容，大部分是自主完成的：它写了岛屿和故事，为 638 句台词选角 35 个声音，生成 16 个带绑定的 3D 角色、音乐和音效，并通过自己玩游戏来测试一切。约 15,000 行代码。
大约花了四天。我的部分是每座岛玩一个 alpha 版本，告诉它哪里感觉不对——声音、资源、游戏等等。我想尽可能保持自动化，所以只玩了 1-2 次（理想情况下应该一次都不玩）
我认为已经很好的：3D 和空间设计。图像到绑定角色，一个看起来舒服、可以玩的世界。图像、3D、语音和音乐 API 的链式调用也是如此。开箱即用的游戏动态相当不错。
仍然非常困难的：心智理论。世界、声音和机制都很棒。写故事是薄弱部分：模型假设玩家知道的比他们实际知道的多，它的 NPC 有时会以奇怪的方式反应，或者对从未发生的事情做出反应。
《钻石时代》的启蒙书（或任何真正的 AI 导师）必须建模学习者，而这在我看来仍然是默认设置下的短板。我实际上已经开始在这方面探索基准测试，这对现在的 LLM 来说是个困难任务。它们容易陷入"知识的诅咒"。教人类也不是一种可验证奖励，所以很难在当前训练范式内规模化。
在这里免费玩，浏览器打开，无需安装：huggingface.co/spaces/thomwolf/open…
笔记本电脑和手机都能玩。七座岛，每座约 15 分钟，六种可能的结局，取决于你一路上教了 Kite 什么。
这仍然是非常探索性的，更像是 beta 而非成品游戏。会有粗糙边缘
