China demographics thread.

[FairAndUnbiased]

Please stop misrepresenting me. My claim is very specific: It is about science/engineering breakthroughs at the top 0.01% level, the kind which win Fields Medals, Nobels etc.

China can win the whole semiconductor tech race, including EUV, 2 nm chips etc. without needing to make a fundamental science breakthrough. Because it is applied science and engineering. Doesn't mean that it's not important, but just looking at a different segment.

I think this thing arose because of migration trends (which are directly relevant to demographics). Anyways yes, back to demographics, and we can discuss these again when the issue arises.

There's 3 Nobels per year. There's more than 3 scientific breakthroughs per year. Thus it's clear to say that Nobels don't represent all scientific breakthroughs. And some Nobels turned out to be duds or harmful, like lobotomy.

At the end of the day it's a human selected prize from a self anointed committee of white western scientists. They're experts but who can claim they're free of bias?

Koreans have 0 Nobels, is SK terrible? the only Nobels won by Indians are under the British Empire. colonial India was better for Indians than independence?

The Transformers paper "All You Need is Attention" (2017) is a landmark paper that proposed the modern Transformers architecture, from a team of eight scientists at Google. The lead author is Indian. Six of the other seven authors are white men, some from Eastern Europe who immigrated to the U.S. to work at Google. All eight authors left Google to launch their own startups in the West.

It is based on an earlier paper by 3 authors, one Korean and two Europeans, from 2014 that proposed an attention mechanism. The paper was published in the U.S.

ImageNet and AlexNet were seminal breakthroughs in AI. ImageNet was from a Chinese American woman who immigrated to the US. AlexNet was from a Canadian team - as you mentioned. Word2Vec was a Czech guy who later immigrated to the US.

The major breakthroughs in Deep Learning in the late 2010s are mostly from the U.S. This is notable since many news articles claimed that China was publishing as many or more AI papers in the late 2010s. Even the ResNet paper was from Microsoft. China should reflect why it domestically is unable to come up with major breakthroughs in Deep Learning.

ImageNet is a manually labeled dataset... I mean how big of an innovation is it to come up with the idea of paying guys minimum wage to label data?

 

[gadgetcool5]

delete

 

[Eventine]

The Transformers paper "All You Need is Attention" (2017) is a landmark paper that proposed the modern Transformers architecture, from a team of eight scientists at Google. The lead author is Indian. Six of the other seven authors are white men, some from Eastern Europe who immigrated to the U.S. to work at Google. All eight authors left Google to launch their own startups in the West.

It is based on an earlier paper by 3 authors, one Korean and two Europeans, from 2014 that proposed an attention mechanism. The paper was published in the U.S.

ImageNet and AlexNet were seminal breakthroughs in AI. ImageNet was from a Chinese American woman who immigrated to the US. AlexNet was from a Canadian team - as you mentioned. Word2Vec was a Czech guy who later immigrated to the US.

The major breakthroughs in Deep Learning in the late 2010s are mostly from the U.S. This is notable since many news articles claimed that China was publishing as many or more AI papers in the late 2010s. Even the ResNet paper was from Microsoft. China should reflect why it domestically is unable to come up with major breakthroughs in Deep Learning.

No one outside of the US with its near infinite supply of capital - especially back in the 2010s - was interested in sponsoring basic theoretical research during the middle of the AI winter. Back then DARPA was practically speaking the only source of grants for most theoretical AI research, so it’s no wonder these breakthroughs all came from the US.

China and even Chinese international students as a talent pool did not really get into AI until it began to be commercialized. Culturally China breeds a practical mentality and financially China was/is much weaker than the US so it does not surprise me it is not able to keep up on the basic research side. This is one of the benefits of near infinite capital from being the reserve currency and investment sink for the elites of the rest of the world; a role China is not likely to ever take.

 

[gadgetcool5]

No one outside of the US with its near infinite supply of capital - especially back in the 2010s - was interested in sponsoring basic theoretical research during the middle of the AI winter. Back then DARPA was practically speaking the only source of grants for most theoretical AI research, so it’s no wonder these breakthroughs all came from the US.

China and even Chinese international students as a talent pool did not really get into AI until it began to be commercialized. Culturally China breeds a practical mentality and financially China was/is much weaker than the US so it does not surprise me it is not able to keep up on the basic research side. This is one of the benefits of near infinite capital from being the reserve currency and investment sink for the elites of the rest of the world; a role China is not likely to ever take.

China should certainly try to internationalize the yuan more, but it does not take infinite capital to invest in basic research. The U.S. spends about $100 billion a year on basic research, which is about 2.5% of China's annual budget. And since the cost of certain inputs are lower in China, China can benefit from some advantages on the financial side.

U.S. elites are irrevocably committed to the technological containment of the PRC, and is committed to cutting off technology transfer, so unless the PRC wants to become a vassal state of the U.S. in the long term like Japan, it has no choice but to cut out the short-term mentality of trying to copy Western tech, and learn to compete with the U.S. and its empire in terms of generating indigenous innovation.

 

[FriedRiceNSpice]

No one outside of the US with its near infinite supply of capital - especially back in the 2010s - was interested in sponsoring basic theoretical research during the middle of the AI winter. Back then DARPA was practically speaking the only source of grants for most theoretical AI research, so it’s no wonder these breakthroughs all came from the US.

China and even Chinese international students as a talent pool did not really get into AI until it began to be commercialized. Culturally China breeds a practical mentality and financially China was/is much weaker than the US so it does not surprise me it is not able to keep up on the basic research side. This is one of the benefits of near infinite capital from being the reserve currency and investment sink for the elites of the rest of the world; a role China is not likely to ever take.

Until recent years, China was too far behind across a wide spectrum of technology that it made little sense to invest too much in basic research, there was simply too much catching up to do. With limited capital available for investment, it was far more prudent to focus on low hanging fruit and prioritize on catching up in fields that offered the maximum return on investment. Only once you have an ample supply of capital and a solid industrial-technological foundation does it make sense to allocate more resources to the cutting edge and emerging fields.

 

[tokenanalyst]

No one outside of the US with its near infinite supply of capital - especially back in the 2010s - was interested in sponsoring basic theoretical research during the middle of the AI winter. Back then DARPA was practically speaking the only source of grants for most theoretical AI research, so it’s no wonder these breakthroughs all came from the US.

China and even Chinese international students as a talent pool did not really get into AI until it began to be commercialized. Culturally China breeds a practical mentality and financially China was/is much weaker than the US so it does not surprise me it is not able to keep up on the basic research side. This is one of the benefits of near infinite capital from being the reserve currency and investment sink for the elites of the rest of the world; a role China is not likely to ever take.

China was the 5th publisher of AI papers in 2000s and the 1st in 2010, looks like they have been researching AI for a long time. I can bet money that a good chunk of those papers in the EU and US have Chinese names as authors.

 

[FairAndUnbiased]

Until recent years, China was too far behind across a wide spectrum of technology that it made little sense to invest too much in basic research, there was simply too much catching up to do. With limited capital available for investment, it was far more prudent to focus on low hanging fruit and prioritize on catching up in fields that offered the maximum return on investment. Only once you have an ample supply of capital and a solid industrial-technological foundation does it make sense to allocate more resources to the cutting edge and emerging fields.

Even 20-30 years ago China invested a fuck ton of money in leading edge applied research like battery, AI, IC, dpace, etc.

Unless you mean even further back like 1980s.

 

[fatzergling]

No one outside of the US with its near infinite supply of capital - especially back in the 2010s - was interested in sponsoring basic theoretical research during the middle of the AI winter. Back then DARPA was practically speaking the only source of grants for most theoretical AI research, so it’s no wonder these breakthroughs all came from the US.

China and even Chinese international students as a talent pool did not really get into AI until it began to be commercialized. Culturally China breeds a practical mentality and financially China was/is much weaker than the US so it does not surprise me it is not able to keep up on the basic research side. This is one of the benefits of near infinite capital from being the reserve currency and investment sink for the elites of the rest of the world; a role China is not likely to ever take.

AI/ML was considered a lost cause until the 2010's at the very earliest. The field suffered because none of it's algorithms scaled and had a bad reputation for overpromising. As a result, only the US had the time and energy to keep ML academia afloat before then.

 

[tokenanalyst]

Even 20-30 years ago China invested a fuck ton of money in leading edge applied research like battery, AI, IC, dpace, etc.

Unless you mean even further back like 1980s.

Even further there was research in many areas including semiconductor, an impressive task give the development conditions at the time.

The first transistor in 1956:

"It is understood that the collaboration group was led by comrades Wang Shouwu, Wu Erzhen, Wu Xijiu, Cheng Zhongzhi and other comrades. They worked hard for more than three months and after more than 30 arduous explorations and experimental productions, they finally successfully developed our country's invention in November 1956 . The first semiconductor germanium alloy transistor opened a new era in China's semiconductor industry."

And the first IC in 1964:

"Later, my country's first silicon device micro-assembled integrated circuit was born. In 1961 , Wang Shoujue, who was working at the Institute of Physics of the Chinese Academy of Sciences, learned about the invention of silicon planar devices and solid circuits in the United States. He resolutely decided to terminate the ongoing development of silicon mesa tubes that had achieved certain results, and concentrated the efforts of the semiconductor research laboratory to develop silicon planar technology, and At the end of 1963 , the development tasks of five silicon planar devices (low reverse current diodes, PNPN high-sensitivity switching devices, high-speed switching transistors and two high-frequency transistors) were completed .

On this basis, in 1964 , Wang Shoujue led a team to successfully develop China's first silicon transistor micro-assembled integrated circuit. This is a type of resistor-capacitor coupling gate circuit commonly used in third-generation electronic computers. It is an electronic circuit composed of a total of 19 components including 6 transistors, 7 resistors and 6 capacitors made on a silicon chip . The circuit is packaged into a tube shell smaller than a watermelon seed."



I think is cutting edge research in pretty harsh times.​

 

[Eventine]

China was the 5th publisher of AI papers in 2000s and the 1st in 2010, looks like they have been researching AI for a long time. I can bet money that a good chunk of those papers in the EU and US have Chinese names as authors.

View attachment 128661

I'm not seeing a race towards dominance here; in fact those numbers show China publishes fewer research AI papers today as a % of the total compared to 2010. Further, the vast majority of China's papers are in areas like computer vision, where practicality is driving the research. Basic research - the sort that leads to break throughs like Deep Learning - still mostly take place in the West and especially the US.

I'm not saying this to down play China's achievements as China is still obviously ahead of everyone else not in the West and many of the authors of Western papers are Chinese or Indian. But the Western choke hold on basic research persists and it has much to do with the ability to just throw government and investor money at it.

AI/ML was considered a lost cause until the 2010's at the very earliest. The field suffered because none of it's algorithms scaled and had a bad reputation for overpromising. As a result, only the US had the time and energy to keep ML academia afloat before then.

Exactly so. The US's ultimate power comes from finance. It can afford to keep research areas alive that are of no commercial value, and did for decades through the various AI winters. China and the rest of East Asia chose - and is still choosing - to focus on practical, applied research in industries like manufacturing. This is a conscious trade off and it has consequences, one of which is that the West is incredibly strong in basic research, while East Asia is incredibly strong in advanced manufacturing.