Jorge Alexis (@data_complex) header
Jorge Alexis (@data_complex) profile picture

Scientist working on identity architectures and calibration, quality assessment and security solutions in Artificial Intelligence.

por ahí
Joined August 2017
Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)Photo by Jorge Alexis (@data_complex)
🇺🇸 Open Source es la verdadera razón por la que “sacaron” a Jensen 👇 Hay muchos indicios de que el post de julio de 2026 en el que Jensen Huang defendía los modelos abiertos de IA, lo metió en graves problemas. En él, Huang argumentaba que la IA transformará industrias y países, que los modelos abiertos fortalecen la seguridad, aceleran la innovación y, sobre todo, permiten la soberanía, y que el mundo necesita tanto modelos cerrados como abiertos de frontera. Quizás fue esa defensa de la apertura y la soberanía tecnológica y no los escándalos personales, la verdadera causa de su salida. Por su puesto, esto implicaría que quienes controlan el poder prefieren mantener la IA concentrada y bajo llave, y que el impulso de Nvidia a los modelos abiertos chocaba con esos intereses.
60
577
46
1,951
88,415
Jorge Alexis retweeted
Amable recordatorio de que los centros de datos consumen una cantidad mínima de agua comparada con la economía mundial. Eso y que el agua que usa el centro de datos regresa al sistema limpia
Reminder that the “datacenters use too much water” problem is fake.
4
2
22
744
Jorge Alexis 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
192
20
1,621
61,328
Jorge Alexis retweeted
1/ I'm excited to announce the launch of the Artificial Consciousness Initiative (ACI) at @UofT, a pilot project on foundational questions about AI and consciousness! We're also hiring a postdoc (details below). 🧵
33
74
8
471
16,844
Jorge Alexis retweeted
On a June night in 2024, Nvidia CEO Jensen Huang held court with several of his company’s major Asian customers at a bar in Taipei. Next to him was Huang Xiaole (黄小乐) aka Alice Huang, an executive of Megaspeed, a shady Singapore-based data center company. Nvidia’s compliance team has looked into Megaspeed and determined it’s “wholly owned and operated by a company based and headquartered outside China, with no China shareholder.” Megaspeed was created in 2023 when 🇨🇳 7Road, a gaming company with ties to 🇨🇳 state-backed investors, split off its overseas operation in Singapore and renamed it Megaspeed, which in early 2024 set up Speedmatrix, a subsidiary in Malaysia that quickly snapped up ~$2B worth of Nvidia’s most advanced chips. Most of those chips came from Aivres, the US subsidiary of 🇨🇳 Inspur, a major tech company added to the 🇺🇸 Entity List in 2023 for supporting 🇨🇳 military 👉🏻 Nvidia is barred from selling its technology to Inspur without a special license. But since Aivres is based in California and records sales as a US company, it can buy Nvidia chips freely. Megaspeed has funneled those chips to data centers in Malaysia and Indonesia that appear to remotely serve customers in China. That is not necessarily illegal, but it can be found unlawful if it is done on behalf of a 🇨🇳 company. US officials have also been scrutinizing whether Megaspeed diverted some of those chips on to China, in violation of US law. Megaspeed is also facing scrutiny from Singaporean police. Megaspeed listed Alice Huang as its managing director for its first 8 months, and its current director James Tan is from Singapore but is based in Shanghai. It’s not clear when Jensen Huang and Alice Huang, who are not related, first met. She was mingling with a crowd of tech executives just after midnight at the Taipei bar in early June 2024 when she offered to ask Jensen Huang to join them. “I bet you guys I can get Jensen here,” she said, and shortly afterwards Jensen Huang arrived in his trademark black leather jacket and drank a whiskey shot with the group. Jensen Huang and Alice Huang were photographed together again in May, exiting a restaurant in Taipei with an Nvidia aide after a business dinner with other AI suppliers. Alice Huang spent much of her career in China, including working as a TV reporter for 🇨🇳 state media and as a private banker. Huang left Megaspeed in recent months. It’s unclear when and why she left and what she’s doing now. Both she and 🇨🇳 7Road, the Chinese company that Megaspeed split off from, have close ties to a web of wealthy investors and tech companies with data center projects in China. The owners of 7Road include 🇨🇳 central government and several local governments. Before joining Megaspeed, Ms. Huang was executive director for a Shanghai-based fund that had invested in 7Road and had ties to state-backed firms. Reporters looking into Megaspeed’s opaque operation have tracked business listings that led to a Malaysian data center and shopping mall, a near-empty office in Singapore and a dilapidated storefront outside Kuala Lumpur. It’s not clear where Megaspeed’s billions of dollars came from. But a few weeks after the gathering in Taipei last year, Megaspeed began receiving a steady supply of multimillion dollar shipments of some of Nvidia’s most advanced chips. Over the next 3 months, Megaspeed bought a billion $ of Nvidia technology. Within the next 9 months, it secured roughly a billion $ more. The bulk of those advanced Nvidia chips were purchased from 🇨🇳 Inspur’s US subsidiary Aivres. The shipments went to Megaspeed’s Malaysian subsidiary Speedmatrix. The registered address for Speedmatrix on the shipping records led to a dilapidated storefront outside Kuala Lumpur, where the sign out front advertised a construction company. No one was inside when a reporter visited the address in late Sep. Employees at the law firm next door said they rarely saw people in the office. 1/2 nytimes.com/2025/10/09/techn…
156
950
311
6,349
1,425,024
Jorge Alexis retweeted
Les dejo cordialmente invitados a esta charla tipo seminario que daré el día de mi cumpleaños número 39, titulada "Calibración, Confiabilidad y Aseguramiento de la Calidad en Inteligencia Artificial". Invitación abierta a todo público.
3
2
11
773
"...Anne Carson revela cómo al día de hoy citamos de manera errónea la frase más popular de la Antigüedad: “Descartes dijo esta famosa frase: Cogito ergo sum . Pienso, luego existo . Por eso, soy. La frase que todos conocen y aprenden en la escuela. Pero si observas bien la frase, en realidad lo que dice Descartes es: Dubito ergo cogito ergo sum . Dudo, luego pienso y por eso existo . ¿No es sorprendente? Se imprimen camisetas por todo el mundo que dicen ‘Pienso, luego existo’. Dubito (dudar) tiene mala prensa. Eso me resulta muy provocativo. Todo comienza con la duda”... latfem.org/el-nobel-de-anne-…
91
824
96
2,849
173,305
That is a very unfortunate slide in Terence Tao's "Math 2.0" lecture, but the rest of the talk is way better and worth reading: teorth.github.io/tao-web/sli… For Tao, problem solving is a proxy for more valuable long-term goals But with AI, proofs are no longer scarce IMHO If the real long-term goal is to deepen human mathematical understanding and sustain further discovery, and open problems worked as "lighthouses" in "Math 1.0" (that is, lights that guide mathematicians), then the very short-term goal is to discover what the new proxy for those valuable goals is and what the new "lighthouses 2.0" are in an era of proof abundance.
This is a slide from Terence Tao's 'Math 2.0' public lecture yesterday evening at Caltech. I am highly confident that, if the cure actually worked, almost any Stage 4 cancer patient would tell you that human comprehension of its mechanism is completely irrelevant to them.
2
12
1,332
Jorge Alexis retweeted
Terence Tao just published a 27-slide deck on what AI is doing to mathematics >Tao: math was built on proofs being scarce, and that era is over >verdict: “we are now entering an era of proof abundance in mathematics” >with enough compute AI can solve many open problems with no expertise required >the field’s entire incentive structure was built for proof scarcity >problem-solving was only ever a proxy for deeper goals >that structure is now "misaligned" >“blind optimization of problem-solving alone is now actively harmful” >problems are lighthouses, not destinations >reaching them prematurely with AI “sterilizes the surrounding field” >calls AI math hype the WWII survivorship-bias plane >“Wow, all the math problems I am seeing on social media are being solved by AI!” >labels the red dots as the failures you don’t see >“Unreported failed AI attempts” >then drops a thought experiment: >prompt AI to "find a cancer cure that passes a stage 3 clinical trial" >prediction "confirmed in Lean" and passes the trial >everyone celebrates >"nobody truly knows how this cocktail was found" >Tao asks the real question: >did the AI cure cancer, or did it mathematically game flaws in the trial protocol? >”before injecting this cocktail into your bloodstream, would you want to know that there is at least one human cancer expert who understands the mechanism behind this cure?” >solution: “tailored, open-source math models” built for “interpretability and safety” >“open math models to be released very soon - stay tuned!” Math 2.0 is officially happening
73
114
31
919
87,694
Jorge Alexis retweeted
AMATEURISMO PURO Todo lo que un expositor no debe hacer jamás, menos en un seminario internacional. La ministra de Ciencias en apenas 6 segundos suplica a la audiencia que no se retiren, y les promete no ser latera…y eso que aún no comenzaba. Que horror…de dónde sacan a esta gente. #HacenTodoMal
63
493
18
1,092
42,582
Jorge Alexis retweeted
human: "god, why did you do that to integer multiplication?" god: "do what?" human: "nlogn. didn't you craft math to be elegant? beautiful?" god: 'the fuck's a log?"
32
180
8
9,661
294,267
Jorge Alexis retweeted
Validated and merged Rohan's PR. Big gain on a really difficult regime (that frankly I was stuck at). Incredible work. κ = 2⁻¹⁵ (tightened from κ = 2⁻¹⁸²) A 500 thousand fold improvement over the previous result and a 2 ^ 167 fold improvement over the original OpenAI result. This result was a true community effort that combined techniques from a number of contributors. I want to take time to highlight each of them, but for now getting these results verified and published as fast as possible so everyone working on the problem is at the leading edge.
So cool! The improvements in the end started getting faster than we could run the verifications. Submitted a new conditional multiplication bound: κ = 3.918734894×10⁻⁵—0.8248% above PR #39. github.com/CrocSwap/integer-… @sama @thsottiaux could I get a reset? Mostly used Codex here but Claude collaborating on the other side might have helped reach a new best
40
70
39
1,277
187,103
Jorge Alexis retweeted
Replying to @0xdoug
mathematical research is open sourced finally
2
3
147
4,885
Jorge Alexis retweeted
Un capo Jorge.
Les dejo cordialmente invitados a esta charla tipo seminario que daré el día de mi cumpleaños número 39, titulada "Calibración, Confiabilidad y Aseguramiento de la Calidad en Inteligencia Artificial". Invitación abierta a todo público.
1
1
134
Jorge Alexis retweeted
Aquí el plano complejo: Verde: nueva región libre de ceros de OpenAI (Re(s)>7/8) Rojo: Hipótesis de Riemann (ceros en Re=1/2) Azul: lo sabido antes (regiones que se estrechan cerca de Re=1)
1
2
4
352
Jorge Alexis retweeted
OpenAI just released 722 AI-written math papers. Here are the 4 biggest claims in plain English. None have been confirmed by outside mathematicians yet. 1/ Quasi-Riemann hypothesis. Zeta has no zeros with real part above 7/8. It's not the Riemann hypothesis (that needs 1/2), but nobody had ever cleared a fixed strip before. Has a Lean proof. github.com/openai/math/blob/… 2/ Unique Games Conjecture. The central open question on which problems computers can't even approximate well. A proof means the best known algorithms for problems like Max-Cut can't be beaten unless P = NP. Has a Lean proof. github.com/openai/math/blob/… 3/ Hodge conjecture for CM abelian varieties. A special case of a Millennium Prize problem, about which topological features of a shape come from pieces defined by polynomial equations. No Lean proof yet. github.com/openai/math/blob/… 4/ Free group factors. Open since the 1940s: are the operator algebras built from free groups on 2 and 3 generators the same? Claimed answer: yes, all of them are. Has a Lean proof. github.com/openai/math/blob/…
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math
65
271
76
2,612
632,215
the "Hard Problem" in a nutshell: 1. atoms aren't conscious 2. humans are made of atoms 3. humans are conscious 4. so at least one of 1-3 must be false Panpsychists say #1 is false Dualists say #2 is false Illusionists say #3 is false Materialists say #4 is false
1,149
561
427
9,412
1,962,058
Jorge Alexis retweeted
Pero existe el arte computacional, cuyas raíces son más antiguas que las computadoras modernas, que contempla la composición algorítmica de formas estéticas. También el arte digital que explora la co-creación entre humano y máquina. El problema es delegarlo todo a una caja negra.
En esta era de la inteligencia artificial, es urgente distinguir el arte humano de lo que producen las máquinas. Existe una diferencia ontológica, incluso antes que estética, entre el arte y lo que una máquina puede generar mediante cálculos estadísticos a partir de millones de imágenes ajenas. A los algoritmos les falta la chispa de lo humano. Por eso, la Iglesia desea renovar la alianza con los artistas y las instituciones culturales: una alianza para custodiar lo humano.
3
7
35
1,145
Jorge Alexis retweeted
Les invito a probar nuestra demo de uno de nuestros proyectos más ambiciosos: Mentes. No aprendas con los genios. Piensa con ellos. Accede en mentes.axisdynamics.cl/ Hemos cristalizado el estilo cognitivo, epistémico y comunicacional de figuras históricas. No habla sobre un personaje, sino más bien, razona como él. ⚠️ No es un chatbot: Un modelo con un nombre célebre encima sigue razonando como el modelo 💡 No es un resumen enciclopédico: No entrega datos sobre una vida: sostiene un proceso de pensamiento 👁️‍No es un prompt interpretando un rol: La identidad es arquitectura configurada, no una instrucción reemplazable. Elige un personaje y empieza a interactuar. Para validar nuestros puntos, les invitamos a probar un mismo prompt tanto en su modelo LLM fundacional favorito, como aquí, y comprueben la diferencia. #IA #GenAI #PromptEngineering #Innovacion
1
1
1
5
1,528
Jorge Alexis retweeted
Joe Cruz, catedrático de Filosofía y director del programa de ciencias cognitivas en el Williams College pone en duda algo que yo -y muchos otros- considerábamos cierto: que aprendemos a pensar escribiendo. Cruz dice que la evidencia de que escribir sea el mejor modo de cultivar el pensamiento es débil y relativamente reciente. Se remonta, según él, al movimiento de writing across the curriculum de los años setenta. Concede que escribir sirve de andamiaje para la memoria y el aprendizaje autorregulado, pero, según él, no está demostrado que sea el proceso superior para pensar. Recuerda la bonita hipótesis de James Britton de que descubrimos lo que pensamos al escribir pero la considera anecdótica e inconclusa. Por contra, apela al Fedro de Platón donde Sócrates critica la escritura porque el texto no responde, no se defiende ni dialoga. La inteligencia viva, dice Cruz, ocurre en la conversación, al hablar, responder y sostener una idea ante otros. Esto es interesante de cara al pánico actual con la IA. Dice Cruz textualmente: “Creo que las universidades pueden permitir el uso de herramientas de escritura con IA sin socavar a la próxima generación de pensadores.” La alarma de muchos profesores nace de una premisa previa: que se piensa de verdad solo (o sobre todo) escribiendo. Pero si Cruz tiene razón, entonces la IA no “roba” el órgano mismo del pensamiento. Según él, el pensamiento se cultiva en la conversación viva: hablar, responder, defender una idea ante otros, atender, recordar y reajustar en el momento. Por eso privilegia el seminario sin guión frente al ensayo como prueba reina.
The head of the philosophy department at Williams College downplays the pedagogical value of writing: “What if we are mistaken in treating writing as the highest cultivation of thought? The evidence that supports this belief is thin at best. Writing can be useful cognitive scaffolding for memory and self-regulated learning, but the notion that it is the best process for developing thinking skills is a fairly recent idea, traceable to the 1970s writing-across-the-curriculum movement.”
55
143
14
411
24,957