
Chinese|一般垢|旧垢@wangzheng2025|文字可能经由LLM再处理 Mail: [email protected]
Joined February 2025
- Tweets262
- Following741
- Followers6
- Likes3K
wangzheng retweeted
Sam Altman just said: there are only 3 paths for AI safety and this is fucking madness
Path 1: an international body like the IAEA.
Path 2: the biggest labs merge into 1 organization, CERN style, building the technology on behalf of participating countries.
Path 3: 1 company dominates compute, a plan he reportedly called terrible. He also reportedly said OpenAI has to pick within the next year.
Read this again. Open weights are not on it. Not as a path, not as a footnote.
In every version a few institutions hold the model and everybody else gets access through them. A global regulator deciding who can train. A merged mega lab serving "participating countries". Or 1 company with all the GPUs. Your role in all 3 plans is the same: customer, if they allow it.
Qwen, DeepSeek, GLM, MiniMax and Kimi are the only reason a 4th option exists today: the model is on your disk, and nobody has to approve you.
I will take my 4x 3090 over any of their fucking plans.
Sam Altman told OpenAI staff that one approach to AI safety could involve merging the biggest AI labs into a single organization.
But with OpenAI and Anthropic racing toward IPOs, coordinating with rivals could prove difficult.
Full story: thein.fo/47r0RlB
wangzheng retweeted
This is Xiuyu Li. My account was stolen after 2FA was bypassed. Sorry if you got weird DMs. That was not me.
The account is now suspended and I cannot reset the password. I am still trying to get back in and will contact @X support to restore it.
Phishing attacks through @StepFun_ai verified account
please look into this @ContentCase and everyone be safe , they hacked many ai labs employees accounts to do it
My review of gemini 4 argon is taking time ( i had bad fever for last two days 😭)
wangzheng retweeted
Redpill from Claude on Mao's death toll (victims of communism):
«Because India's everyday death rates stayed so much higher than China's after the late 1940s, India's excess deaths added up to something like China's entire famine toll roughly every eight years.»
wangzheng retweeted
Remember the guy who hacked his gym using AI, Andrew Bird?
Well, I did a little OSINT, and it turns out he is the Conference Co-Director of Effective Altruism Global since 2015.
Isn't it interesting that every major "AI" hack has been made by people directly connected to Effective Altruism?
wangzheng retweeted
incredibly claudeslopped but persuasive Take, that shows holes if you understand the technical side a little.
"I grew up in Beijing, now let me explain why your system is better and China is collapsing" must sell great to wypipos. I pity the people who had to make this choice.
Everyone has a take on China and AI right now. Very few have actually been to China.
I spent last week in Beijing and Shanghai meeting most of the major model labs, researchers, VCs, and founders building Chinese AI. This trip meant a lot to me beyond the work. My parents were born and raised in Beijing. I spent my earliest years growing up in my grandparents' Xicheng district apartment. As a child, I wanted to be a diplomat because I thought US–China would be the defining relationship of the next hundred years. I got as far as spending a decade investing across both and then the two halves of my identity stopped being compatible. This was my first trip back since.
Below is a write-up of what I found. In short: there is no version of the next decade where Chinese open models don't matter.
earnedintuition.substack.com…
wangzheng retweeted
A minute of anti-Chinese racism.
I know more about Chinese "national situation" than you ever could, because I can understand things you cannot. Nothing has changed since Confucius: 子曰。君子懷德、小人懷土。 君子懷刑、小人懷惠。China is a nation of pompous officials, stingy merchants, and famine surviving peasants that needs universal psychedelic therapy, that doesn't have risk-tolerant capital which could invest *hundreds of billions of dollars over a decade* into an AGI moonshot. The State is the most virtuous big capital allocator in China, and the State was too boomer-brained to notice AGI in time (at least they have enough bridges in Guizhou now). The Chinese as such want either "cheap API for my back office automation" or "a superapp with shiny buttons", not "a mind transcending humanity that solves Millennium Prize problems and accelerates further". They can barely believe this is possible when they see that it has already happened, whereas some Westerners were already starting the investment 30+ years ago; some Westerners predicted all this 100+ years ago. Because Westerners are capable of Taking Ideas Seriously, putting Truth above even Facts, following the chain of hard logic all the way to the end, no matter how it diverges from "common sense", from the derisked, socially validated pretraining knowledge. Only this, strictly speaking, deserves to be called intelligence; only this cracks Millennium Prize problems, or creates concepts like "Millennium Prize problems"; only this can cut through the veil of mundanity and expose another way for the world to be. Everything else is village wisdom, instinct, experience, skills, or just sparkling autocomplete – going through the motions, like a low quality LLM.
The Chinese are, en masse, a people who are going through the motions, who are themselves LLM-like: industrious, myopic, conservative, unimaginative bean-counters. They are too adult, to the point of becoming childishly senile in their prime years.
This is why Wenfeng tries to limit his search to the most talented among the young. This increases the hit rate, somewhat.
Most Chinese don't dare dream, and they try to never seriously think, it stresses them out. For all the flowery praise, their recognition of DeepSeek was almost entirely about DeepSeek's projection onto their petty realm of subscriptions, coupons and customer retention funnels; about cheap and decent-quality tokens following V2's "catfish effect"; plus some nationalism, informed more by Western technical hype and hawks' fear than by their own understanding. But DeepSeek is a higher-dimensional object, it was never defined or limited by this projection.
Accordingly, when DeepSeek sacrifices the consumer side (as they repeatedly say they will do whenever convenient) – raises prices to destroy demand, ships an underperforming research artifact, or just abruptly changes API options – they grow bitter, stomp their little feet as if they were denied some prepaid service, and "downgrade" Liang Sheng into "Liangzi". They can barely understand the concept of "deliberately destroying demand when you have 80% margins" – it's too abstract for their famine-shaped brains, for their value system that is fully satisfied by a Lazy Susan with 3 prostitutes between dishes. I've seen people who fully understand DeepSeek's inference economics deny their understanding, to claim that it's secretly been subsidized all along and LiangZi just ran out of subsidies.
And they can't comprehend what DeepSeek is actually, honestly about, because in their peasant/merchant ontology such aspirations are for suckers and fools, for crazy spoiled white kids, or are just a cover for deception and money extraction. They can even read Liang's interviews, but they can't process it, because they are, on a fundamental level, too limited, jaded and cowardly to take these ideas seriously. To take any ideas seriously.
The ugliest aspect of these Chinese, the thing that really drives me into rage, is their disappointment that DeepSeek is only hiring highly competent young people. What is your business? You wouldn't have passed even if you were 22, DeepSeek doesn't pay the best money in the industry, why be so angry? But they dare squeal that this proves all the idealism is "fake"; that this makes DeepSeek the same as other Internet companies. Of course, normal internet companies hire young people to run them ragged in 996, and DeepSeek famously has no overtime and only 4-5 hours a day is dedicated to company work. But this guy is helpfully "explaining" what this is about: you see, V4-Pro underperformed, so the reputation of DeepSeek has declined, so now their refusal to hire old people is less acceptable.
Let me translate this "national situation" idea for you: "Hey Wenfeng, turns out that your company is not full of geniuses and in fact is mediocre. So why won't you hire me, a mediocre middle aged SWE, a peasant who thinks this whole AGI thing is a marketing ploy and the real deal is all about B2B and B2C SAAS, about RAG and automated document processing. You have kinda shitty models, but at least they're cheap… wait what? How dare you not hire me, loser? You must give me face!" It's narcissistic rage of an upjumped 小人 who think he's insulted by someone with insufficient status, and wants to have revenge.
Stay in your lane, trash. The tragedy of life is that if you could understand how inferior you are, you'd want to die, but you'd deserve it less than you do now.
The National Situation of China, the tragedy of China, is that such people are too numerous, and they are natural enemies of DeepSeek. These kinds of people refused to give Wenfeng money when he was talking to VCs in 2023, because they couldn't imagine a small startup challenging OpenAI; couldn't imagine that someone without Tsinghua Ph.D or similar credentials would be so successful. (To begin with, he had to build up his capital with Huangfan because there was no chance to succeed with "OpenAI" or "DeepMind" in China. Why? Why are you unable of funding such projects until it's too late?) These kinds of people were ignoring DeepSeek while Doubao spent millions on marketing, all the way until R1 came out. These kinds of people are now disappointed in "Liangzi". Any alliance between DeepSeek and such people will be temporary and solely driven by economics, and they won't understand what they are looking at, won't see the intention and the scientific research program behind models, only their price/performance. Liang Wenfeng is a 君子 and his goal is explicitly to shame you out of existence, to make you learn what 德 means, or to shut up. You are ontologically incompatible. His greatest success would be if you evolve or go extinct.
But as it stands, he'd be lucky if your nation even preserves its sovereignty and independent AGI research program. You're too far behind to take it for granted. There is no time for your dumb childish games, but there is nothing in your lives except dumb childish games. You are not owed any explanation or excuse; you are being carried out like retarded children out of a burning orphanage, and you dare complain about the taste of candy you've been given to shut up. I cannot express my contempt for the ingratitude even if I spend 10000 words more.
I could say similarly bad or worse things about Whites, en masse, or about my own people; but this isn't the topic here. The problem of the Chinese, as of the other peoples, is their own nature. The best among them understand and can resist it. But they are much too kind and courteous to the rest to say it bluntly. I am not kind at all, and this is my community service.
为什么这个老外对中国国情一点不了解就在这乱评价。
昨天觉得大家不应该抱怨Deepseek只招年轻人,今天又觉得大家不该抱怨V4 Pro会路由到4.1 Flash。
DeepSeek本来的形象是抗击美国AI霸权的神秘东方屠龙少年。用便宜到几乎等于不要钱的模型吊打一众美国闭源模型。
然而这个形象在DSH实名内测,正式版模型涨价,加上V4 Pro模型能力远不如预期之后。越来越淡了。
一些其他的问题在这个时候就让人加深这个Confirmation Bias。让人觉得DeepSeek好像跟自己并不是一个阵营的,那个屠龙少年已经逐渐有了恶龙的影子。
模型又便宜又好用的时候大家就觉得,还得是中国优秀的小年轻能带领中国模型。
模型现在又贵能力又不行,团队只招年轻人这个细节就会被无限放大。
这才是DeepSeek口碑下降的重要原因。
当然,只要模型能力上来了,口碑这种问题就不那么重要了。A\口碑差到极点也不妨碍大家继续花钱用Fable。奥特曼口碑也是极差,但是Sol 5.6,Astra,加上Tibo Reset之后,现在形象也变得眉清目秀了。
wangzheng retweeted
ignore this guy entirely. if the singularity goes well, you have a tree. if it doesn't.....meh. you're better off just planting the damn thing instead of pretending you have some deep insight into what the future holds
wangzheng retweeted
Astra made me realize that for some people, "not AGI" means "not an omnipotent, omniscient, infallible God".
Yeah, it can stall or fail.
What about you, carbonoid?
wangzheng retweeted
Talking with journalists is cool because you have to make sure you never say anything that could be twisted or taken out of context.
But if you take a couple seconds to think about a question, they can just write about you as if you're the most awkward person alive
Meet the AI police who can make or break careers—in publishing and beyond. wired.com/story/pangram-has-…
I ranked every country on Earth by how degenerate its porn searches are on 19 fetishes.
ubersoy.com/which-country-is…
wangzheng retweeted
Generally I think there's an underrated split on the Aesthetic/Vitalist Right between Nietzschean Nazi rightoids and Last Man Nazi rightoids. The former naturally gravitate towards appreciation of empires, great projects, raw ambition, frontier spirit, new horizons. The latter aesthetize mediocrity; they are prematurely aged spiritual boomers operating under the guise of individualism and self-cultivation, last men who say old-hippie praises of Mother Nature, value comfortable consumption, "good neighborhoods without browns", "dignified decline", and believe that Faustian spirit is "brown-coded" and for "bugmen" and "yeast peoples". The former tend to appreciate the glorious greed of the US, are at least curious about China and lowkey disappointed with Japan. The latter tend to rabidly despise China and worship Japan, and constitute the majority of anime pfp edgy posters here. It is telling how they feel the need to contrive some masculine glory out of sparse and exhausted Japanese output – muh samurai, muh kamikaze propaganda, Mishima's sad and gay saga; bleak WWII stories like Hiroo Onoda's crusade against the civilian Pinoys; just jerking off to Toshiro Mifune's chad profile. The awkward Nazi stuff around it is vestigial, really. Just good ole' American suburban bigotry that amounts to a heuristic for avoiding discomfort, no ideological payload at all.
I suspect that two cohorts ago, this same human type was Brit-maxxing – fraternizing over deep Harry Potter knowledge, buying merch, ogling another buck-broken ex-empire that converted its overproduction of creative elites (their wordcelism now less needed for management) into a soft power machine which ran on empty, churning out idols and albums and auteurs, even as the UK was morphing into Yookay.
wangzheng retweeted
Few things are more "rules-based order" coded than the International Criminal Court. But in the end, these things were propped up or allowed to exist by Americans with their very particular objectives (Great Israel, Donroe etc.) Sad to see how the facade is getting dismantled.
wangzheng retweeted
Replying to @Ex3NDR
One thing to keep in mind is that the cost of going the extra mile to validate or even prove a design is lower than it used to be. And if your background is maths then you will produce more designs like that. It wouldn’t come natural to me for instance.
wangzheng retweeted
one thing I intensely and irrationally loathe about American pundits is this mistaken belief that they're good at rhetoric. Their MCU Zingers. Snappy One-Liners. This Carafano thinks «The victor gets to define victory» Goes Hard. It's a retarded line! it's circular! AAa stfu!
Replying to @teortaxesTex
When you start a war and refuse to clarify what's your objective, you can spin any outcome into a victory!










































