Where you raise your children determines their future

KeithLee · June 16, 2026, 2:20am

This essay is inspired by Children’s zip codes change their brains, new study finds | Scientific American

Months ago, we had a disucssion on the long-term direction of the institution, not just for the AI research institute in Swiss, but for the entire GIAI ecosystem.

As the author of the Scientific American article says, children’s childhood environment largely shape how they think. The SA article extensively covers the US cases, but I already have abundant examples from South Korea with SIAI.

Many South Korean students at SIAI had suffered deeply from cognitive problem sets that European students overcome much easily. In Europe, I often feel pessimism that they are surpassed by Asian smart brains and I sometimes see through a bit of jealous on me for my AI expertises that they can’t have from earned from Europe.

But as the 75 Korean students have shown with British education-styled SIAI programs, when Korean students are confronted with applying textbook statistics/ML/AI knowledge to real world like case studies, except a few highly gifted talents, most others just have shown little to no readiness to such extensions.

There can be thousands of explanations, but I think the SA’s childhood exposure as the brain process formation is the most relevant explanation. We are ineed trained to choose an answer out of 5 options throughout K-12. And even math questions are not designed for later case studies, but just finding quickest possible one number answer. By the fact that most students go to private academy hours just to learn skills to earn higher scores in that 5-option questions, we might be able to learn faster for ready-made questions, but that training won’t help us writing a question itself, which is why the majority of 75 Korean students ended up with Fail in most courses in Machine Learning / Artificial Intelligence.

What surprised me is that after learning that most attendeees at SIAI ended up with Fail grades from ML/DL/RL like AI courses, outsiders mocked SIAI students making stupid choice to come to SIAI, not self-blaming Korea’s mid-directed education.

Their brains are designed in a way to think that getting a higher score from ready-made questions is the most efficient way of life, and those people who have tried to overcome Korean education’s upper-limit are seen as idiots or inefficient choices. For them, just going to a known school that promises them to get a job at Samsung / SK / LG like large corporations is more than enough. What they can do at the company and how to develop their skills are not in their minds. At all. Why? Getting into Seoul National University (Korea’s Harvard) is the most important thing for Korea’s education. Once you get in, you can easily graduate, and your life is all promising. Why bother learning new skills from European top school level education that will not be used at Samsung / SK / LG?

Most developed parts of the world have their own problems with the ‘Childen’s ZIP code’ issues, but the Korean experience tells me a lot about how to run a country, how to train talents, and how to choose which country for SIAI’s GSB to jump into.

Knowing all that now, will I ever try again in Korea for GSB? Nope.

For the same reason, for Japan, since they will only care about whether MARCH or not, I advise my team not even think about going to Japan with SIAI, at all.

China had the same problem with Japan and Korea until recently, but the Chinese leaders manage to break that inertia. If they continue, I am quite certain that they will have totally different class of talent groups. I envy them so much. In Europe, I have seen their envy to Japan/Korea frequently. You guys have modern system, so trains must run faster than our 150km/hour slow ones. Your road conditions must be way better than here. Your country does not have street crimes like ours. etc. etc. Now in these days, among tech-savvy groups, I begin to feel their envy to Chinese advances. I can see myself joining that envy group.

KeithLee · July 5, 2026, 10:45pm

For the development of an organization, be it a small company or a large country, there are two contending theories. One of which emphasizes individual’s potential which leads to democracy and freedom of choices, which is prevailing theme of growth and social structure in the western hemisphere.

The other school of thoughts emphasizes central control that achieves highest possible efficiency. This system is often adopted in East Asia and for growth economics topics, it often has been discussed as the most successful example case of centralized control’s maximum efficiency.

In Network theory, it is called Mesh Network vs. Star Network. Mesh has every node links to every other nodes. Star has only one single central node has connections to all others, while no other node has any connections with any other nodes. Like North Korean leader single-handedly controls the entire country.

For AI, we can say that both the US and Europe follow their tradiitonal Mesh Network social structure, and the US as the successful and Europe as the failing example.

In the East Asia, with Star Netowk, China is an example of success, while both Japan and Korea as Eruopean counterpart of the Star Network.

Whenever I teach the Network Theory, I always emphasize the fact that textbook’s notion of advantage/disadvantage does not really foretell a social system’s success and failure. For the Star, if the center is ultra efficient, the system will grow just as the efficiency that the center achieves. However, if the center fails, the entire system fails.

With the Mesh, if each node moves efficiently and if the total system does not have negative externalities (something like Hirshleifer effect), then the system achieves the maximum effiency possible. However, either at node level or at system level, if there grows inefficiency, it is going to erode the aggregate system.

For the childhood ZIP code problem, I would say that 2000 years old confucious-doctrine in East Asia was the ZIP code. Japan and Korea used well in the 20th century when they grew. But the mature system natively created inefficiencies here and there. Both countries so far have not been able to kill them successfully.

China, at least what we see from AI space, can be said that the ZIP code problem is broken. Since it is broken by the central control, I doubt the efficiency will stay long, like what we already have seen in Japan and Korea for the past 50 years, but at least at the moment, we can only say that the code is broken for their dream for global hegemony.

The Mesh version social infrastructure are often stable. It does not depend on a single central node, like an emperor or a king, but cleaning up messes for good really goes slow, which is why I doubt Europe will ever be able to get back to their 19th and 20th century status. The wane is inevitable.

For the US, despite the fact that I admit there are a series of inefficiencies here and there, the US private sector-driven AI hardware/software infrastructure is highly efficient and highly stable. There can be minor upgrades done by a few reforms, but Americans are doing almost their best. For the US be able to keep the hegemony, unfortunately, is not depending on what Americans do for the next 10-20 years, but what China can and will do. If China stays in the current course, central node driven ultra highest efficiency, it is just a matter of time. If the center collapses, even after the power dynamics flips, since it is not a robust system, it will work as a favor of the US.

In fact, this is what has been done for the past two millennia. When China was centralized and when the emperor could gather the country’s resources for maximum efficiency, it thrived. When the center corrupted or collapsed, it was not. Chinese now see the history lessons, which is why Xi Jinping’s power is not challenged. Keeping him at thrown is better than replacing him with another leader to build the social network again for the next 10 years.

Given that Chinese finally understands their true potential, it must be scary scenario to the US high offices as the power flip is seen as only the matter of time. However, never in Chinese history, a dynasty kept its maximum power over 50 years. The center concentrated node ironically is the weakest network, as only the center needs to be broken, for a new power to emerge.

How China will handle that political risk in the 21st century will be an interesting topic to observe. At least, never in Chinese history, had the country ever have kept peace without a powerful emperor. In other words, Chinese people do not have that ‘ZIP code’ for politically consistent system.