hamyori (@hamyori) header
hamyori (@hamyori) profile picture

C, C++ & AI without idealism. Pragmatic & Apatheist. Game theory.

Joined April 2023
Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)Photo by hamyori (@hamyori)
フィールズ賞の数学者がAIと数学界の共存についてめちゃくちゃ良い事言ってる。 特に「AIの出力(証明)を自然現象として扱い人間がそれを理論化すればいい」については、ハッとさせられた。 数学が、物理や化学と同様に科学の一種になったということなのかも。 以下訳抜粋 • 従来、難しい予想の証明は高く評価されてきた。なぜならそれを証明するにはその周辺分野の深い理解が必要で、証明自体がその理解をさらに深めるから。 • AIが生成した証明は、それ自体では深い洞察や新しい理解を自動的に与えない。ただ機械が人間より上手に証明を出すようになっただけ。 • だからといって悲しむ必要はない。AIの出力は「自然の一部」や「データの源泉」として扱い、人間がそれを精査・理論化すればいい。 • 数学研究はむしろ民主化することでさらに繁栄する。 • これまでトップで活躍するには、複雑な証明を感じ取り組み立てる「生まれつきの特別な素質」が必要だった。 • その部分はAIが担うようになる。 • これからの差を生むのは「良い質問をする力」「観察力」「理解を深めたいという強い欲求」。 • もし今リーマンが生きていたら、素数とゼータの零点の双対性というアイデア自体を思いつくのは依然として彼の天才によるもの。しかし証明を完成させる部分ではAIが助けになったかもしれない。
This note from Bao Chau Ngo (Fields Medal 2010, Clay Research Award) on Alpoge's work for Riemann is incredible. It's rare to see a mathematician of his caliber with such an open mind about AI.
15
439
68
1,528
78,849
Bueno, aquí vamos de nuevo...
The Lean FRO and Mathlib Initiative have released a joint statement on the Lean verification of the Riemann Hypothesis. On November 20, Levent Alpöge released a proof of the Riemann Hypothesis as a Lean library of 17 million lines of code. Over the past day, the Lean and Mathlib communities worked together to verify it. The statement certifies the proof as correct and describes how it meets all necessary verification criteria. Congratulations to Levent on this achievement, and thank you to everyone in the community who contributed to the verification. Full statement: leanprover-community.github.…
4
Tiene razón, las matemáticas no son resultados, esto es el pensar ingenieril auto-destructivo, las mates van más bien sobre tener un mejor entendimiento de las estructuras y objetos
Terence Tao: AI is winning at mathematics, but are we actually learning anything? “I've never seen such brutal competition, especially with such money-backed resources behind it.” “We've learned almost nothing mathematically from what they did.” “Normally, when a human proves a great mathematical result, they can talk about what worked and what didn't.” “But here, it's like strip mining the landscape. You get all the fruits, but the landscape is barren afterwards.” “We need to develop new practices for doing mathematics responsibly in the AI era.”
4
17
125
2,977
A los doce miraba los Simpson.
web.archive.org/web/20230331… 多分日本の2008年生まれで一番強かった人。12,13あたりでGoogleに脆弱性報告してる
4,185
Veremos que tal se sostiene esta afirmación en los próximos días.
JUST IN: Riemann Hypothesis has been solved. alpo.ge/remark.pdf
14
Interesante lectura para este finde.
We shrank 1 GB of Wikipedia to 105.39 MB, losslessly. Introducing tufazip: the best entry on the Large Text Compression Benchmark using a neural network as its sole predictor. Everything above it mixes a transformer with hand-built models. Full post 👇 tufalabs.ai/research/tufazip
8
Y ahora heme aquí, teniendo que vivir.
unavailable
22
Quien pudiera ponerla.
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
11
Hay mucha gente que actúa como si las mujeres fueran otra especie.
This quoted post is unavailable.
20
Nada supera «el ser proactivo».
5
La universidad es el lugar de facto para contactos, aunque en Latinoamérica esto suele limitarse solo a profesores.
在自媒体搞那种在海外大学挑战山河四省高中作息vlog的人,如果你不是为了注意力经济或者哗众取宠搞点流量,而是真的这么想的,这大学你确实是白上了。 首先,在这个时代上大学最重要的核心逻辑就是connection。它甚至不是论文,不是实习,不是找工作的certificate。就现在的情况,Github上几万星+不错的语言功底+AI工具的使用熟练度+对于某个行业的特殊专注与兴趣的含金量都大于你的学位。 如果上大学就是闭门造车,真不如直接选一个Arizona State U的线上项目,你拿个线上学位就得了。大学时代的connection,包括但不限于可以参与一些学校meeting寻找绝对同频人的connection,AP的connection(如果你们大学确实很棒,比如美国R1或者国内92),甚至是遇到自己心上人的connection(哈哈,虽然我没有这方面的connection)。再强调一遍,大学的运营机制永远不是让你复刻高中的任何东西,包括任何的学科认知和学习逻辑。 其次,我认为在大学——尤其是从进入大一开始任何回望和发布自己在当年高中某某联考排名、高考多少、AP、IB分数、托/雅考多少并且发抖红(除了你要做职业学习博主或家教)show off的人,都可以被划归为神智不清病入膏肓的范畴。 相反,从大一开始知道自己想要什么,专注投资、科技然后专注美股、量化、人工智能领域也好,或者在国内确实想卷公务员体系猛攻政治系统然后心甘情愿用自由换权力的人也罢,都值得我们至高的尊敬。 生活是游戏,也是战场,知道自己应该做什么的人,才能打好每一组副本。而沉湎于上一个副本的战士,学不会忘记上一秒的战士,永远不理解的战斗的真谛。 最后我想说一下,大学不需要作息,真的,大学不需要作息。 因为高中是有共同的既定目标的,就是高考。大学时期,你个人是完全自由的,你完全不需要跟着别人的目标走,你只有你自己的目标。那么对于这个带有灵活性的目标,熬夜是正常的,盯盘是正常的,崩溃也是正常的。它如果足够社会化,你就知道它根本不可能按照高中那种几点几分干什么事情,每天都如此。 因为大学的每天都是丰富的,只要你做的事情根本不止于学校上那些课,而是充分地掌握信息、拥抱人工智能、拥抱前沿,你的时间不可能根据计划行进,只有一个大框架。一切,都会以这个大框架为基底而内容每天疯狂变化,我是说每天。 好好拥抱你的大学时代吧,忘记高中的晦暗或闪光。过去毫无意义,现在和未来才是你应该值得努力和托付的珍宝。
5
hamyori retweeted
変なXYプロッタが作りたいという動機で少しずつ作っていた、リング式XYプロッタが完成しました!  ※わりと耳障りな音が出ますので注意。 紹介動画はOpus 5.5に作ってもらったものです。解説版の長い動画、解説記事のリンクはリプに貼る予定です。
🤖 Made with AI
48
1,187
180
4,816
596,535
Replying to @Getumei0000
今の世界の出生割合で見ると、もし自殺したら、次に生まれ変わる可能性が一番高いのはインドかアフリカなんだよ。
222
1,426
251
52,583
1,150,938
hamyori retweeted
One of the most used algorithms in ML just got a major upgrade after nearly 70 years, as Meta researchers found a way to outperform k-means with 1/10 the training passes, while achieving even lower clustering error. This paper proposes e-means, which updates cluster centers using smaller batches and exponential moving averages instead. It also uses scaling laws to automatically choose the batch size and learning rate, enabling it to converge faster without additional hyperparameter tuning. Across embedding datasets, e-means achieves lower clustering error in just 1.4-2.6 epochs compared to 25 epochs for traditional k-means, providing 10-45x GPU speedups. It can even cluster 10 billion embeddings into 16 million groups in just 2.5 epochs. alphaxiv.org/abs/2611.09501
12
193
20
1,619
61,962
hamyori retweeted
I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
1,823
6,423
2,210
41,259
6,801,035
hamyori retweeted
Cuanto más estudio sobre redes neuronales, menos me arrepiento de haber elegido Física. Es cierto que en la carrera todavía no hemos dado prácticamente nada sobre Machine Learning, aunque también es verdad que estoy en segundo y aún es pronto. Pero lo curioso es que gran parte de la base matemática que necesitas para entender cómo funcionan las redes neuronales ya nos la han enseñado. Álgebra lineal, cálculo diferencial, optimización, estadística... Son conceptos que estudiamos constantemente y que, cuando empiezas a profundizar en el funcionamiento de las redes neuronales, te das cuenta de que están prácticamente por todas partes. De hecho, una de las cosas que más me está gustando de estudiar Machine Learning por mi cuenta es empezar a encontrar aplicaciones reales a conceptos matemáticos que hasta ahora había estudiado de una forma mucho más abstracta. Y creo que ahí está precisamente lo que más me gusta de Física. Siento que es una carrera perfecta si te gustan las matemáticas, pero tampoco quieres dedicarte exclusivamente a estudiar su parte más teórica, sino entender cómo utilizarlas como herramienta para describir, explicar y resolver problemas del mundo real. Evidentemente, estudiar Física no te convierte automáticamente en alguien que sabe de inteligencia artificial. Hay muchísimas cosas de programación, algoritmos y aprendizaje automático que tienes que aprender por tu cuenta. Pero creo que tener una base matemática sólida te permite entender las cosas desde una perspectiva completamente diferente. No limitarte a utilizar una librería o entrenar un modelo, sino comprender qué está ocurriendo realmente por debajo. Y cuanto más aprendo sobre todo esto, más convencido estoy de que elegí bien la carrera.
46
44
5
508
28,680
>1.6M views in 4 hours
5
Todos contribuiremos a materializar esta realidad.
工学部がアンドロイドを作り、 建築学部が出られない部屋を作り、 薬学部が媚薬を作り、 医学部が肉体改造し、 農学部が触手を開発し、 教育学部が教育し、 文系がセックスする
15
工学部がアンドロイドを作り、 建築学部が出られない部屋を作り、 薬学部が媚薬を作り、 医学部が肉体改造し、 農学部が触手を開発し、 教育学部が教育し、 文系がセックスする
建築学科ワイ、セックスしないと出られない部屋の設計を決意
145
2,804
726
41,494
1,723,872