Pinned Tweet
Everyone in the Valley has a take on whether China is winning AI, robotics, and biotech. Almost nobody understands how the capital works
Notes from meeting China's tier-1 investors and founders on how their venture ecosystem actually works.
Bohan retweeted
I interviewed @neilmovva for @substratemag on latency-insensitive inference, alternative chips, and why massive datacenters face diseconomies of scale.
"We've always predicted that async inference is going to eat the world. That could easily be 95% or more of inference demand."
Neil's been at the frontier of inference for years, from NVIDIA to co-founding @sailresearchco. I learned a lot from our conversation - you can listen below.
Mr. @a16z please let me know when you start raising for the Chinese Dynamism fund!
I'd love to invest
I believe no one will go all in on China more than a16z can
Five bangers from Nathan Lambert:
1. China's compute gap:
"You talk to people [in China], and you're like, 'Oh, I heard an average OpenAI researcher has five thousand H100 equivalents that they can access at any time,' and like their jaw just falls to the floor. They're just like, 'I can't comprehend.'"
2. China's talent pool:
"The Chinese labs are very technically excellent... To say that China is anything but equal on talent, you're gonna have a hard time because they just have so many people being trained at such a high level."
3. OpenAI's superpower:
"[The frontier labs] have just taken the most ambitious bets the earliest in order to get there, and that's why OpenAI has done things like kickstart the reasoning model revolution."
"And I think that kind of breakthrough and that beautiful plot with the inference scaling laws is peak OpenAI. They're really good at getting this plot that just changes how everyone thinks about everything."
4. Underestimating the competition:
"The Chinese are so good at building models as well. I think there's a bit of cope where people are like, 'Only we are good at building models.' There are so many companies. There's, like, 20 to 30 companies that are building good models."
5. The AI risks that are already here:
"[The lack of preparation] is kind of the thing that worries me more than, say, RSI is coming and unlocking new risks. There are risks that are here and that are gonna continue to diffuse."
"Figuring out how to get society to prepare is very hard. I'm of the opinion that we need more eyes, we need more science, so that society feels informed."
@natolambert
Chinese labs currently operate at probably 1/10th the capacity of American labs if not less
They do whatever they can to get compute and Bytedance basically buys up all of Southeast Asian supply and willing to pay 2-4X American prices
Chinese researchers believe that if they can get to 1/3 American lab capacity then they can always maintain parity on capabilities but they're struggling to get that
Crazy how compute poor they are compared to some American lab researchers who tell me they have 30K GPUs allocated to their name they can do whatever with and they're bottlenecked on ideas to execute
The world of haves and have nots
Kudos to the Chinese for shipping so well under scarcity
Nathan Lambert of @trillium_labs says Chinese AI researchers' jaws drop to the floor when he tells them how much compute OpenAI has:
"The Chinese labs are very technically excellent. They are more compute-starved."
"Without a doubt, they have a comparable talent pool to draw on... To say that China is anything but equal on talent, you're gonna have a hard time because they just have so many people being trained at such a high level."
"What they don't have is compute."
"You talk to people there, and you're like, 'Oh, I heard an average OpenAI researcher has five thousand H100 equivalents that they can access at any time,' and like their jaw just falls to the floor. They're just like, 'I can't comprehend.'"
@natolambert @lishali88
Bohan retweeted
OOPS: CALIFORNIA POLITICIAN IN MISTAKEN CHINA WARNING
Dick Lucas, a man running to join the State Assembly in California, uploaded a video yesterday raising the alarm that China Airlines is right opposite Lockheed Martin, one of the world’s biggest weapons makers.
“They’re looking through that glass, they are spying on the guys and gals defending our country, we’re on it,” he warns.
This is hilarious. China Airlines is of course based in Taiwan, controlled by Washington through the hated ruling party called the DPP, set up in the late 1980s by the NED, the CIA’s soft-power spin off.
China Airlines office right across from Raytheon in El Segundo, not enough are saying
Readers added context they thought people might want to know
China Airlines is is the state-owned flag carrier of the Republic of China (Taiwan).
en.wikipedia.org/wiki/China_Air…
In the early 2000s, Taiwan, Hong Kong and Korea were viewed by the Mainland Chinese as clearly more developed, clean and advanced
And it was true
Their GDP per capita was higher than China's (and still is though not more than Tier 1 Chinese cities) and all the cultural consumption in China was importations from those more advanced soft power states. Taiwanese accents were sexy and Hong Kong music slapped
Crazy how the tables have turned in just 2 decades
As Mainland Chinese cities went through blitzscale growth, they now look like techno-futurist maximalism whereas Taipei and Hong Kong and Seoul look like Mainland cities stuck in a 80s/90s time capsule
There's been way less development in those cities than in China and they feel a bit run down and Chinese tourists now visit for nostalgia
In Taiwan now they watch Chinese reality TV and in Hong Kong they go across the border to Shenzhen for shopping, entertainment and medical care
Cue obligatory Shenzhen/Shanghai before and after pic
This is basically the internal OpenAI take now:
Robotics is solved
They just need to distill the big models into smaller system 1 robotics models
Unitree CEO Wang Xingxing recently said that GPT-7 could solve the challenges of embodied AI entirely and that he had made a mistake by not taking AI seriously enough.
Quote: “We’ve put it (GPT-6) on various robots and tested it across a range of tasks, and the results have actually been very good. Our company and other robotics companies are already running pilot deployments in factories. To achieve large-scale adoption, though, both generalization and efficiency still need to improve further. But I think these challenges could very well be resolved over the next few years. My faith in AI grows stronger every year. The biggest mistake I’ve made over the past decade or so has been not believing in AI enough. Deployment in commercial services may happen even faster than in industrial settings.”
In the valley now there's a lot of China bullishness and maybe the grass is always greener:
I know researchers at American labs who think China will definitely win and I know Chinese researchers who think they can never catch up
It's hard to understand the reality both because of the extreme narratives but also because we are in a narrow AI bubble watching China's narrow AI bubble
The majority of China's real economy is stagnant and the vast majority of their people are not experiencing the AI boom nor the economic growth (yet) but is that so different from America?
In SF you seemingly have 22 year old billionaire founders everywhere when the majority of Americans are feeling the squeeze of inflation and withering jobs. Lots of parallels
This article is probably one of the more nuanced takes on China's weaknesses and it calls for a coordinated Western response, which seems unlikely
But even without Western pressure, China's export maximalism may take down the world along with itself.
It's crazy how incremental growth in Chinese exports come at the expense of Western domestic industries, which weakens Western economies and means people have less money to spend on Chinese goods, so it's a vicious cycle
If you suppress the prices for manufactured goods globally you drive out global profits and China's foreign customers get poorer and demand shrinks
foreignaffairs.com/china/chi…
Wow. My favorite new fact
Founding father and mother of modern China, who overthrew the Qing Dynasty, spoke only English to each other
People underestimate how diverse China is internally, and it shouldn’t be a surprise that a country multiples bigger in land and population than Europe would have more linguistic and genetic variation than between Spain and Poland.
My grandparents are from a village in Zhejiang province just 5 hours drive from Shanghai yet they aren’t literate and don’t speak Mandarin so I’ve never actually talked to them directly before given our languages are unintelligible
The northern swath of China all the way from Sichuan in the southwest to Dongbei on the Russian/Korean border speak variant dialects of Mandarin (官话) so if people talk very slowly they maybe could understand each other.
The south, due to the rivers and mountains that cut up the geography is much more linguistically diverse and village to village is sometimes an unintelligible language. Even though Cantonese is written with mostly the same characters I even sometimes have trouble reading it.
Another fun fact is that Cantonese is hypothesized to be closer to how ancient Chinese was pronounced and so sometimes ancient poetry’s rhythm and rhymes make more sense when read in Cantonese
Interesting historical fact I learned about today: Sun Yat-sen and his wife Soong Ching-ling, both legendary founding figures of modern China (Soong Ching-ling was Honorary Chairman of the PRC and Sun Yat-sen's portrait still goes up in Tiananmen Square on National Day), only spoke English to each other!
This is because she spoke Shanghainese and he spoke Cantonese, which are mutually unintelligible.
However they both spoke perfect English because they were both educated in English-speaking schools: Soong Ching-ling at Wesleyan College in Georgia, and Sun Yat-sen in Hawaii and in then-British Hong Kong.
In fact, funnily enough, when Soong Ching-ling - on her deathbed in 1981 - was told of the Politburo’s decision to admit her to the Communist Party, it was announced to her first in Mandarin (which she never mastered) and then again in English so she would understand (thepaper.cn/newsDetail_forwa…).
This all highlights how recent China’s linguistic unity is: it’s actually one of the country’s major achievements.
China has had "one script" since the Qin dynasty, but "one pronunciation" is a very recent achievement - people from different regions often couldn’t understand each other when they spoke until the past few decades.
To the point where even China’s "Father of the Nation" and "Mother of the Nation" couldn’t speak the national language to each other!
Back when Scale was DoD’s biggest AI vendor he was hawking on China all day, given the reality that he unfortunately looks Chinese
Now, just like Zuck, his whole talk track has shifted, understandably. When your whole team is FOBs you can’t be shitting on their homeland all day
Meta AI Chief @alexandr_wang on China: “I've personally changed my own opinion on this matter.
In the past I was very worried about China. I had read these old speeches that were given by PRC officials in the past that talked about how they viewed AI as this opportunity to leapfrog the West.
But as time has gone on, I think we’re painting them to be too much of a boogeyman when they're going to see the same things that we see in terms of this is powerful technology. There's going to be a lot of shared goals in making sure that AI goes well.”
Full HUGE* Conversation out now.
Just paid respects at Lee Kwan Yew’s memorial tree
Singapore is running just as efficiently if not more than he planned and the technocracy still reign strong
He insisted that no monuments be made for him and the only memorial to him is this one tree his son planted, to the founding father of this nation
A true GOAT
Hosting our second Cypress Society lecture on 10/21 with Os Guinness, social critic and descendant of THE Guinness brewer
He’ll be examining the roots of the American Republic on its 250th birthday
The Cypress Society is a nonprofit based in SF aimed at reviving the tradition of the public lecture in the humanities. We welcome donations
luma.com/9w7acrdc
Bohan retweeted
Comfy API is live
Point it at a workflow JSON and it handles the rest:
→ Resolves your custom nodes, models, LoRAs, and Python deps
→ Pins it all, ComfyUI version included, into a versioned Build
→ Cuts an immutable release, so what you test is what ships
→ Deploys as an autoscaling endpoint on RTX PRO 6000, H100, H200, or B200
Click the link below to deploy your first workflow and learn more with our blog ⬇️
Secretary of War Hegseth claims he wants the military to focus on “lethality” but revealed preferences show that the national security strategy is Looksmaxxing because nothing scares the Chinese more than elite waist-to-height ratios and clean shaves
Religion is resurging in America and even among the Godless technologists of Silicon Valley, but I'm kinda surprised it's happening in Western Europe and especially France, one of the most secular countries where over 80% claim to be Catholic but no one believes in God
wsj.com/world/europe/in-fran…
JB coming to town
If you're an SF techbro and would like to learn the aesthetics and wisdom of the aristocratic classics, you should sign up!
I’m hosting a happy hour @8vc SF on Nietzsche’s Untimely Meditations this coming thursday October 1st.
It is an overlooked work of his through the writing of which Nietzsche discovered his true calling. I read this work in August in Switzerland and have been so moved by it — specifically its critiques of modernity, science, technology and instructions for escaping the mimetic trap.
I find it especially relevant to the AI transition we find ourselves in and an invaluable guide in finding meaning at a time when everything seems to be shifting under our feet.
Apply below to attend:
forms.gle/uMEQNrxLzoYuk7es6
