3 ms·

I am not sure what he is talking about. I work on AI/Machine Learning and I can't think of any major paper that came out of China.

wombatmobile 6y ago
Papers are one metric. In the case of ML, as with other technologies such as web infrastructure, there comes a point where the state of research plateaus for a while and the main progress is in commercial implementations. So, for example, autonomous vehicle technology now is more a matter of implementations than theory. This is where China has an advantage over the US by virtue of the willingness of its government to make large scale infrastructure investments. The US once led in such endeavours e.g. Eisenhower's Federal Aid Highway Act of 1956 that paid for 90 percent of the cost of construction of Interstate Highways. Now, in the government domain, the leader is China. China intends for self-driving cars to propel smart megacity https://asia.nikkei.com/Economy/China-intends-for-self-driving-cars-to-propel-smart-megacity https://asia.nikkei.com/Economy/China-intends-for-self-drivi...
visarga 6y ago
That doesn't matter so much because they can apply all the SOTA results just as well as the rest of us, and maybe better because they have fewer ethical concerns.
- ma2rten 6y ago
  Maybe that's true, but that is not what he is saying: "In the race for publishing papers China has now caught up."
  - visarga 6y ago
    Most of those papers are just applications of already existent ideas to different problems.
    - unishark 6y ago
      Also tiny incremental advances that are barely worth publishing, assuming they are even replicable. It seems they pay their faculty based on publication quantity.
  - ck_one 6y ago
    http://cvpr20.janruettinger.com/statistics http://cvpr20.janruettinger.com/statistics => caught up is an understatement
arvinsim 6y ago
I am ignorant on this field so please indulge me with this question: Isn't in China's best interest not to make public any innovations that they can leverage against other countries?
- ma2rten 6y ago
  Maybe if it was some kind of military technology. But Schmidt's quote is about publications and the article talks about civilian companies.
YeGoblynQueenne 6y ago
Out of curiosity, can you please define "work[ing] on AI/Machine Learning"? On your actual point (sorry for the digression) many machine learning papers have at least 2/3s of authors with Chinese-sounding names, though affiliated with US institutions (or companies). Maybe no major papers have come out of "China", but they sure have come out of the work of Chinese citizens. China has increased its research output and impact in recent years. The AI Index report has a plot: https://hai.stanford.edu/sites/default/files/ai_index_2019_report.pdf https://hai.stanford.edu/sites/default/files/ai_index_2019_r... You'll have to go to page 19 to the pdf for the section titled "Published papers: Citation impact by region" which shows Field-Waited Impact by region and over time- sorry, I can't see a way to link directly to the page. Other plots in the sam report also paint a similar picture: China is revving up its research engine on AI. I see this as a missed opportunity. Most Chinese researchers are and were trained in US and UK institutions. Unfortunately, instead of making sure to forge strong bonds of friendship and collaboration that would hold after the return of those students and researchers to China, the US at least has chosen to see them as a threat, so the effect is that they've basically been training their competition. The UK is a bit more on the fence about it for now.

wombatmobile 6y ago

Papers are one metric. In the case of ML, as with other technologies such as web infrastructure, there comes a point where the state of research plateaus for a while and the main progress is in commercial implementations. So, for example, autonomous vehicle technology now is more a matter of implementations than theory. This is where China has an advantage over the US by virtue of the willingness of its government to make large scale infrastructure investments. The US once led in such endeavours e.g. Eisenhower's Federal Aid Highway Act of 1956 that paid for 90 percent of the cost of construction of Interstate Highways. Now, in the government domain, the leader is China. China intends for self-driving cars to propel smart megacity https://asia.nikkei.com/Economy/China-intends-for-self-driving-cars-to-propel-smart-megacity https://asia.nikkei.com/Economy/China-intends-for-self-drivi...

visarga 6y ago

That doesn't matter so much because they can apply all the SOTA results just as well as the rest of us, and maybe better because they have fewer ethical concerns.

ma2rten 6y ago
Maybe that's true, but that is not what he is saying: "In the race for publishing papers China has now caught up."
- visarga 6y ago
  Most of those papers are just applications of already existent ideas to different problems.
  - unishark 6y ago
    Also tiny incremental advances that are barely worth publishing, assuming they are even replicable. It seems they pay their faculty based on publication quantity.
- ck_one 6y ago
  http://cvpr20.janruettinger.com/statistics http://cvpr20.janruettinger.com/statistics => caught up is an understatement

arvinsim 6y ago

I am ignorant on this field so please indulge me with this question: Isn't in China's best interest not to make public any innovations that they can leverage against other countries?

ma2rten 6y ago
Maybe if it was some kind of military technology. But Schmidt's quote is about publications and the article talks about civilian companies.

YeGoblynQueenne 6y ago

Out of curiosity, can you please define "work[ing] on AI/Machine Learning"? On your actual point (sorry for the digression) many machine learning papers have at least 2/3s of authors with Chinese-sounding names, though affiliated with US institutions (or companies). Maybe no major papers have come out of "China", but they sure have come out of the work of Chinese citizens. China has increased its research output and impact in recent years. The AI Index report has a plot: https://hai.stanford.edu/sites/default/files/ai_index_2019_report.pdf https://hai.stanford.edu/sites/default/files/ai_index_2019_r... You'll have to go to page 19 to the pdf for the section titled "Published papers: Citation impact by region" which shows Field-Waited Impact by region and over time- sorry, I can't see a way to link directly to the page. Other plots in the sam report also paint a similar picture: China is revving up its research engine on AI. I see this as a missed opportunity. Most Chinese researchers are and were trained in US and UK institutions. Unfortunately, instead of making sure to forge strong bonds of friendship and collaboration that would hold after the return of those students and researchers to China, the US at least has chosen to see them as a threat, so the effect is that they've basically been training their competition. The UK is a bit more on the fence about it for now.