China’s Quest for AI Dominance – And How It’s Going

Author: William Vorhies

Summary:  An update and observations about China’s plan to become the world leader in AI. 



Whether you get your news from Facebook or from the Wall Street Journal you can’t help having heard that China is out to displace the US as the world leader in AI.  Variously you may have heard that it’s already happened or soon inevitably will.

The twin questions of when they will succeed (is it inevitable) or whether they will succeed (if ever) is one I get all the time.  As a red-white-and-blue American I hope not.  As a world citizen of the tribe of data scientists I wonder why we can’t just all get along.  And as those divided feelings should presage, the current state of this struggle is about both competition and cooperation, and also about unintended consequences.


The Plan

If you’ve got a few minutes I recommend you read China’s plan for world domination of AI.  So no, since it’s available for anyone to read it doesn’t reveal any dark secrets or plots.  In fact it’s pretty well written.  So much so that you’d hope the US has one that’s equally well thought out.

You can find “Notice of the State Council Issuing the New Generation of Artificial Intelligence Development Plan – State Council Document [2017] No. 35” by following the link.  There’s even a condensed and illustrated English-language version for more popular consumption here.

Keep in mind this plan was released in July 2017 making it now three years old.  And like any long government-issued planning document, ours or theirs, counting real planning, rewrites, and revisions that makes it much older. 

So the state of AI and the state of the world that these planners experienced likely culminated in about 2016 around the time that the flurry of exciting discoveries and advancements in AI was closing and we were moving in earnest to implementation.

After several paragraphs about why AI will be the, or at least a major driver of economic success for the plannable future they announce these time based goals.

  1. The first step, by 2020, is to keep up the overall technology and application of artificial intelligence with the advanced level of the world. [AI} …[supports] the goal of the struggle of China’s entry into the ranks of innovative countries and the achievement of a moderately prosperous society.
  2. The second step: by 2025, we shall achieve a major breakthrough in artificial intelligence basic theory, and parts of the technology and application will be at a world leading level. We shall make artificial intelligence become the main driving force of China’s industrial upgrading and economic transformation, and make positive progress in the construction of an artificial intelligence society.
  3. The third step: by 2030, we shall make artificial intelligence theory, technology and application achieve the world’s leading level to be the major artificial intelligence innovation center of the world, intelligent economy, intelligent society achieve remarkable results, and lay an important foundation for China’s entry into the forefront of the innovative countries and economic powers.

So there you have it.  Don’t lose ground.  Start contributing to basic research and theory.  Make it the centerpiece of advancement.  And in 10 years, 2030, be the leader.

The plan also acknowledges China’s shortcomings from 2017 that pretty much cover the waterfront.

At the same time, we must also clearly see that the overall level of development of China’s artificial intelligence and that of the developed countries still have gaps. China lacks major original results and there are huge gaps in the basic theory, the core algorithm, key equipment, high-end chips, major products and systems, components, software and interface and so on; … and cutting-edge talent is far from meeting the demand…

The balance of the plan lays out some detail of the structure they intend to promote.  That includes

  • Emphasis on theory
  • University-to-commerce knowledge transfer
  • AI application across all possible domains from commerce to healthcare to government,
  • Home-grown infrastructure in the form of chips and analytic platforms
  • Developing a highly trained data science cadre and workforce
  • And popularizing AI to gain the support of the people.

All to be developed with the direction of the state controlled plan and the entrepreneurial efforts of private Chinese citizens.

There is the obligatory language about the dominance of government guidance and the “strengthening of military-civilian integration” for the defense and security concerns of China in all these developments.  There’s direct reference to utilizing relationships with the advanced research institutions of other countries.  There aren’t any dark secrets in here. 


So How’s This Working Out

Three years into the plan is a good point to start looking for successes and problems.  Certainly the rise of China’s tech unicorns is a notable success and who can get by without TikTok.  Yes the tech success of Alibaba, Tencent and the others was dramatically facilitated by having the models of their US doppelgangers to follow.  But is China relegated to being only a fast-follower?

If you read the 2018 comments by Lee Kai-Fu, (a former Apple, Microsoft and Google exec, now a venture capitalist) or any of the dozens of bloggers who have repeated his thoughts, China has already won.

China can outcompete the United States because AI has moved “from the age of discovery to the age of implementation, and from the age of expertise to the age of data.” For China, what matters now is “the power of data.”

By population, the number of relatively inexpensive newly trained data scientists willing to put in the hours to label data, and by societal norms much less constrained by data privacy, China has a clear edge in data.  But data is primarily useful for creating apps using current AI technology.

For Kai-Fu’s assertion to be true you would need to assume that innovation at the fundamental level of AI is over.  DNNs are as good as it’s going to get.  And this is probably the wrong answer.

Yes we’ve been making a lot of advancement in applications, particularly with DNNs and traditional ML.  But there have been ‘AI winters’ before when innovation slowed, only to reappear. 

Reinforcement learning is one candidate.  And the deep thinkers and pioneers in AI are becoming increasingly unhappy with the limitations of neural nets and the restraints imposed by their compute and data hungry needs.  US and worldwide research into new basic AI innovations hasn’t slowed down but evolutionary improvements to neural nets have been more common than revolutionary alternatives.


How Is China Doing at Advancing Fundamental AI Research

There is some excellent analysis by Senior Fellow Dieter Ernst at The Centre for International Governance Innovation (CIGI) a Canada-based think tank.  Ernst has a long series of academic credits and has performed original interview-based research in China around AI and chips for some years.

Ernst views China’s AI efforts as fragmented and specifically notes a surprising disconnect between university researchers and those who could benefit from commercializing it.

Here we need to pause and give credit where credit is due in the US to DARPA which has driven both fundamental research funding and particularly commercialization of AI and other technologies.  Despite the stated linkage between China’s defense establishment and its commercial world, their DARPA equivalent is either missing or ineffective in promoting home-grown research into commercial AI innovations.

Yes China has made remarkable strides in papers presented at international AI conferences and in the number of patent applications produced.  In fact China has adopted the paper and patent count as a KPI of progress.  But observers see that these are mainly on the application side and not advancing the fundamental research of AI.

Separately you may have read that publish-or-perish in China’s academic community has taken on a new dimension with major scandals occurring over the last four years. 

In 2017 the New York Times reported that “since 2012, the country has retracted more scientific papers because of faked peer reviews than all other countries and territories put together”.

And this month the Wall Street Journal reported “Internationally peer-reviewed journals published more than 100 scientific research papers from China-based authors that appear to have reused identical sets of images, raising questions about the proliferation of problematic science…”.


The First Unintended Consequence

Of course China should not be expected to get all the details right from the get-go on such comprehensive plan.  There are plenty of unintended consequences to go around.  The first relates to one reason why the dynamic ‘private sector’ AI companies in China are so slow to engage in or adapt original research when clearly their M&A activity is nearly as hot as it is in the US.

Ernst points out that unlike the US, China’s stock exchange requires companies to be profitable for at least three years before they are allowed to IPO.  So AI startups, racing for the cash-out are penalized for heavy R&D spending which slows that path to profitability.  The easy way out is to focus on AI applications where China excels.


The Other Unintended Consequence

The second unintended consequence belongs to the US.  After years of failing to hold China accountable for outright IP theft, or their sending 350,000 students each year to benefit directly from the research of US universities, we now have the digital decoupling.

Long before 2018 the US restricted advanced technologies that could be exported to China, chief among them software and chips.  But starting in 2018 as shown by this chart from Ernst’s report the US began to even more dramatically restrict AI technologies including chips. 

From China’s perspective this restriction demanded that they develop their own capabilities.  China had long had very high levels of expertise in producing commodity chips at low cost.  That’s exactly what gave Huawei the edge to dominate the 5G market.  Now however, although starting from behind, the Chinese will develop their own advanced chip and AI language capabilities. 

This is an area where the US had long been perceived to have a natural advantage, but our actions, while understandable, are creating a formidable competitor in replacement for substantial previous customer.


And One Last Competitive Strength China Brings

Finally, there is the issue of 5G.  There is no question that China holds a lead in the adoption of 5G.  With respect to AI, the issue with 5G is simple.  As one of Ernst’s interviewees pointed out “If data can’t move where it is needed, it’s useless.”

This plays to China’s strength in applications and will give them a leg up in IoT and edge compute, if they are able to seize it.  The US infrastructure is behind here but undoubtedly will catch up.  But how long will it take?


Wrapping Up

So no, China has not won.  And no, the US is not invulnerable.  But the US does still hold a lead in several key areas. 

It has been argued that China’s hybrid model of top-down capital allocation modified to some extent by free market companies will ultimately be slower and less efficient that they US free market model.  Increasingly the US government is becoming an active player in this fight through regulation.  We should hope that investment and regulatory encouragement (not congressional investigations of big tech) will be its contribution.


Other articles by Bill Vorhies.

About the author:  Bill is Contributing Editor for Data Science Central.  Bill is also President & Chief Data Scientist at Data-Magnum and has practiced as a data scientist since 2001.  His articles have been read more than 2.1 million times. or


Go to Source