DeepSeek and China govt's role, TikTok, birth data, healthcare, long-term capital, Operation Seagull 2 - Week in Review #40
The Soviet Union failed because they didn’t respect economics. But China is no Soviet Union.
Hi folks, welcome to Week in Review #40 of
.Some of you recently signed up for this newsletter after reading my article Is China a "warlike culture"? - Noah Smith is clueless about China (Part 4) and is probably waiting for the sequel. Rest assured, I have got you covered. I will publish it in a few days.
The main format of this newsletter, though, apart from in-depth essays from time to time, will be this kind of weekly review, in which I will review events happening in China that I believe will have a lasting impact beyond news cycles. Think of the long essays as building a theoretical framework to help you have a basic grounding in understanding China, and think of the weekly reviews as a way to apply those theoretical frameworks to ongoing events. I hope the existence of this newsletter will help China to start making some sense to you.
For this week, the events I will review are:
The game-changing DeepSeek-R1 and the Chinese government’s role in it (free)
More on “TikTok Refugees” (free)
China’s births grew in 2024 (paywalled)
The low-quality medicine scandal and the trilateral power struggle among China’s 3 different health-related ministries (paywalled)
More long-term capital in China’s stock market (paywalled)
“Operation Seagull 2”, the transnational law enforcement against telecom scams (paywalled)
Again, if you are already a paying subscriber of Baiguan, you may DM me to ask for free access or a refund if you have unwittingly double-spent.
If you think this is still too expensive, please ask your favorite authors to participate in my petition to make Substack more affordable for all the folks. To date, 33 Substackers, each with at least 100 subscribers, have joined this petition. We need 964 more signatures to make our voices heard.
#1 The game-changing DeepSeek-R1 and the Chinese government’s role in it
Jan 20 was supposedly the day that the whole world turned its attention to the inauguration ceremony of Trump’s second term. Yet, on the same day, DeepSeek, in its typical low-key style, dropped its R1 model, which excels at “deep thinking” and is comparable to GPT-o1.
However, unlike the very “close” OpenAI, DeepSeek-R2 is an open-source model, meaning anyone could download and install it in the local environment. What’s more, DeepSeek wrote highly transparent papers, even explaining in detail what trials and errors the team had gone through, a move that’s rarely seen nowadays because of obvious concerns about trade secrets.
It’s obvious that DeepSeek doesn’t care about protecting its trade secrets. It seems to exist purely to advance our technological frontiers. It is, perhaps, the real “Open” AI, so much so it got Marc Andreesen to comment: "We are living in a timeline where a [Chinese] company is keeping the original mission of OpenAI alive - truly open, frontier research that empowers all."
DeepSeek also stole the show of the big $500b “Stargate” announcement. One wonders, if a small Chinese company can achieve so much at so little cost, is a $500b price tag justifiable?
This news about DeepSeek also again confirms my long-time disapproval of US export controls. All that export controls can do is push China to innovate and possibly grow a new “technology tree” independently of the West. I also heard that DeepSeek’s team is entirely domestically educated, meaning no overseas study background. If true, this independent growth of technology tree could happen much faster than what’s normally expected.
It also showed an article I wrote in May 2024, titled Can China catch up with the US in AI? seems to have aged well. At a time when everyone was pessimistic about the future of China's AI, I argued that in terms of compute, algorithm, and data, there was no theoretical hard barrier that could bar China from catching up with the US in AI. I just didn't realize my thesis could be proven so soon, so easily.
And it’s not the only news about DeepSeek. On the same day as the R1 announcement, China’s Premier Li Qiang, our second-in-command, met some entrepreneurs, including Liang Wenfeng, founder of DeepSeek and High-flyer Quant.
I know several people from High-flyer, one of China’s largest domestic quant funds. But honestly, this is the very first time I actually get to see what Liang, a young founder & engineer, actually looks like.
Questions quickly arose about the Chinese government’s role in DeepSeek.
I posted something on X about this topic this week, and here is a refined version:
I am quite certain that DeepSeek went under the radar of the Chinese government for a while. There are two main reasons.
First, Quant funds have historically had a bad reputation in China. Whenever there is a market crash, quants are automatically blamed (Week in Review #12, on “Quant Quake”). So, it is unimaginable that a quant fund could persuade the government that it would be the future darling of AI until it proves itself.
Secondly, I think many people have a distorted view of what government can achieve in China. Most of our success stories happen not because of government intervention but precisely because there is a lack of intervention (a.k.a meddling). The government didn’t mandate Wechat, the government didn’t mandate Alipay, and the government didn’t mandate Black Myth Wukong. And here, the government didn’t mandate DeepSeek. They just happened. Each time the government gets itself into a business too early and too energetically, though, it usually has the opposite effect due to the obvious inefficiency of the state machine.
Chip-making is a great example. State-sponsored programs were directed at chipmaking a few decades ago, but they failed miserably. It was only until US sanctions created a real market for domestic chips that we had a chance.
EV was a good example, too. None of the best-performing EV makers (BYD, LiAuto, Huawei, Xiaomi) are state-owned. While the government did have a sound industry policy that gave strong financial support in the beginning, they made sure they gradually tapered off the support so only the strongest survive. For each Chinese EV name you hear about today, there are at least 30 dead ones that no one could remember. It’s the man-made hunger game that was responsible for EV’s success, not state support. (I wrote about this great EV hunger game in Is China defeating Tesla? - Noah Smith is clueless about China (Part 3))
The government’s job is to ensure fairness, justice, and equality, not to be innovative. Its main job in China, as in any country, is to provide a safe environment, manage systemic risks, and build basic infrastructure so that innovations can happen spontaneously on top of it. For example, Alipay may be a successful business invention, but without earlier efforts by the People’s Bank of China to realize free bank-to-bank transactions, Alipay wouldn’t have been economically viable. Likewise, if China had not pushed to connect every rural village with the Internet, our e-commerce industry would not have developed so rapidly and comprehensively.
The government in China also sets the overall direction of the society, with general guidelines and state-sponsored programs to help shepherd social and economic forces into these directions. There are also many subsidy programs and state-led investments in key sectors, especially when a strategic industry is fledgling.
However, state support is not the single biggest factor that contributes to the success of the most successful ones. The bigger factors are, as in any economy in the world, enough organic market demand & capable management. If there is strong demand and very capable management, then yes, state support will “如虎添翼add wings to a tiger". But, if there is none of the above but just state subsidies, you will only get a bloated, parasitic organization that exists to suck blood out of state coffers. I imagine if you do a regression of the rate of success over the number of subsidies, you might not have any quantifiable relations, and I suspect you will even have a slightly negative correlation
So, if anyone thinks “subsidies” are the secret source of business success stories in China, they have an out-of-touch idea of how business works. If anyone thinks our state people are smart business people, they probably haven’t met any “state person” before. If you ask these bureaucrats if their business skills are good enough to qualify as a good business manager, they themselves will also tell you, quite frankly, no.
So next time you see a good success story out of China, resort to good old common sense rather than an imagined stereotype of how a “state” works in China.
Basic laws of economics apply in this country, as in any country. The Soviet Union failed because they didn’t respect economics. But China is no Soviet Union.
#2 More on “TikTok Refugees”
I am glad to see the “TikTok Refugees” phenomenon (Week in Review #39) continue without any negative surprises.
On the one hand, there are signs that the Xiaohongshu platform is embracing those “refugees”, seeing it more as a growth opportunity rather than a challenge for censorship. For instance, as
reported here, it has introduced a translation feature onto the platform. Also, it has opened an official X account, openly communicating with the English-speaking audience. Although of , a strategic communication expert, found some obvious areas for improvement in their messages. They should seriously just hire Ivy.And despite a temporary shut-down of TikTok just prior to the inauguration, it went back to service after Trump guaranteed (with questionable legality) that TikTok still had some time to negotiate a way out. If TikTok continues to survive, eventually most of the “refugees” will return there. The net benefit for Xiaohongshu/RedNote, in the end, will be an epic international marketing campaign that they spent 0 dollars on.
#3 China’s births grew in 2024
The much-anticipated birth data was out in 2024, and the worst hasn’t happened.