China Is Using America’s Own Plan to Dominate the Future of Artificial Intelligence
China Is Using America’s Own Plan to Dominate the Future
of Artificial Intelligence
The Chinese are massively investing in AI research and
tech, while the Trump administration is cutting federal programs wholesale.
BY GREGORY ALLEN, ELSA B. KANIA SEPTEMBER 8, 2017
In late 2016, the Obama administration published three
reports that shared an extraordinary conclusion: advances in machine learning,
a technology that allows systems to learn and improve without explicit
programming, are enabling a revolution in Artificial Intelligence (AI). As AI
systems become increasingly capable of not only routine tasks like driving a
car, but also complicated ones like designing car engines, AI technology will
be the driving force behind transformations across both the economy and
national security.
If imitation is the sincerest form of flattery, then the
authors of the Obama administration reports have received a heartfelt
compliment from their Chinese counterparts. With China’s July 2017 Next
Generation Artificial Intelligence Development Plan, the country has announced
that it, too, sees AI as the transformative technology that will underpin
future economic and military power. China’s plan calls for exceeding all other
nations in AI by 2030, but the checklist for China’s ambitious agenda is
strikingly similar to the policies prescribed by the Obama administration’s
reports. Greatly increase long-term investment in AI research and development
(R&D)? Check. Promote collaboration on AI between the private tech sector
and the government? Check. Develop a pipeline of top AI talent? Check. Invest
to mitigate AI’s potential risks and societal disruptions? Check. The
similarities go beyond such high-level objectives, even including many specific
policy details and recommendations.
A reader of both documents would be forgiven for
concluding that the Chinese strategy’s authors had copies of the White House
reports on their desks as they wrote. Unlike a college term paper, however,
there are no extra points for originality in governance. All that matters is to
have the right policies and to implement them effectively. Unfortunately, there
is plenty of reason to worry that China may have an edge in implementing
policies practically cribbed from the United States.
The Obama administration sought to increase support for
AI R&D, since annual federal funding for all computer science and
mathematics R&D is less than half of what Google alone spends. Rather than
boosting those levels, the Trump administration’s budget calls for cutting AI
research at the National Science Foundation by 10 percent, to a mere $175
million. China, meanwhile, has demonstrated a willingness to spend jaw-dropping
sums on technology when there are strategic national interests at play. For
instance, in 2014, the Chinese government announced a 1 trillion RMB ($150
billion) investment fund to turn the Chinese semiconductor industry into a
global powerhouse. Those figures are not merely hype. By 2017, the government
had already made a third of that sum available. Similarly, China’s new plan
foreshadows massive increases in funding for AI R&D.
One might think that U.S. government support for AI
R&D is not terribly important, so long as U.S. companies lead the way.
Perhaps that’s true in sectors like health care, but the U.S. companies with
the best AI technology are often considerably less willing to invest in
national security applications. When Google acquired DeepMind, arguably the
world’s leading AI R&D organization, DeepMind required Google to prohibit
the use of their research for military and government surveillance purposes.
When Google acquired a leading military robotics developer, Google announced
that the firm would no longer accept military contracts. And Google is far more
cooperative with the national security community than most tech companies.
Chinese firms are less reluctant to make these
distinctions. As China pursues a strategy of “military-civil fusion” in AI, it
wields a range of policy mechanisms to incentivize industry cooperation. This
summer, Tsinghua University — considered “China’s MIT” — announced plans to
establish a Military-Civil Fusion lab that will provide a platform for dual-use
advances in AI. Also this year, China established its first national
deep-learning laboratory under the leadership of Baidu — often called “China’s
Google” – and in partnership with the Chinese Academy of Sciences, Tsinghua University,
and Beihang University. (Beihang University also happens to be a leader in the
development of Chinese military aerospace technologies, including autonomous
systems and robotics.)
No Chinese company has more enthusiastically embraced AI
technology than Baidu. Despite lagging its U.S. and Chinese competitors in the
mobile era, Baidu is reinventing itself around AI. Under the leadership of AI
pioneer Andrew Ng, who until this March served as Baidu’s chief scientist and
head of AI R&D, Baidu’s 1,300-person AI team achieved impressive results,
developing better-than-human speech recognition software a year before Western
companies and building out Baidu’s well-respected machine-learning
cloud-infrastructure team. Ng’s departure was a significant setback, but Baidu
remains focused on AI. Qi Lu, a former Microsoft executive and a leading AI
specialist, has joined Baidu as chief operating officer.
Overall, Chinese tech companies are not as far behind as
many in the West assume. This year, nearly all of the top-performing teams in
the ImageNet Challenge, an influential AI competition, were from China.
iFLYTEK, a Chinese AI startup, currently ranks second in the reading
comprehension tasks measured by the Stanford Question Answering Dataset
leaderboard, besting teams from Microsoft and Google, among others. While not
yet the overall leaders in AI technology, Chinese researchers have proven their
ability to catch up quickly and produce genuine innovation.
As China and other nations increasingly approach parity
with the U.S. in key military technologies such as precision guided munitions
and stealth aircraft, the Department of Defense hopes that AI will enable it to
maintain military technological supremacy well into the 21st century. That’s a
worthwhile goal, but achieving it will require that the United States not
merely have a strategy, but also implement it better and faster than other
countries. The Defense Department has sought to deepen ties with Silicon Valley
to bring cutting-edge AI tech to the Pentagon, but even Defense Secretary James
Mattis acknowledges that thus far the Pentagon has been challenged in these
efforts.
Here too, China is embracing and implementing America’s
strategy. Ng led Baidu’s AI efforts from the firm’s Silicon Valley campus.
Since his departure, Baidu has announced plans to add a second site in the
area. Tencent, the Chinese social media giant that both Facebook and Snapchat
have confessed to imitating, has also moved in.
Chinese companies believe that by rotating Chinese staff
to Silicon Valley and American staff to Chinese campuses, they can accelerate
the timeline for reaching parity with the United States in AI technology and
depth of talent. At present, however, the size and experience of China’s AI
workforce remains dwarfed by that of the United States. Half of the top 10
employers of AI talent in China are U.S. firms, including IBM, Intel, and
Microsoft, who may thus be integral to the development of China’s human capital
in AI.
China’s copying of America’s homework could be a positive
development if it deepens our understanding of the risks brought about by the
rise of AI and how to mitigate them. The 2016 White House report addressed the
workforce challenges of AI head on. Larry Summers, who served as Obama’s first
director of the National Economic
Council, predicts that the rise of AI in the economy could bring about “a third
of men between the ages of 25 and 54 not working by the end of this half
century.” Such an unemployment rate would be higher than that endured by
Germany during the Great Depression when the Nazi Party came to power.
China’s AI plan likewise shows a clear understanding that
increased automation is likely to reduce demand for labor across many sectors,
putting societal stability (and perhaps the legitimacy and survival of the
Communist Party) at risk. That’s why China’s strategy includes a call for a new
regulatory focus on the ethical, security, and governance challenges of AI,
including workforce displacement. Like the Obama plan, the Chinese plan also
calls for government action to mitigate the economic pain and social
instability of worker displacement. China’s government-led transition of
hundreds of millions of laborers from agriculture to manufacturing shows that
it has experience in this area, but for now its AI-related efforts remain at
the planning stage. With official GDP growth rates still above 6 percent,
China’s near-term problem is actually a labor shortage.
Unfortunately, the United States is no longer attempting
to plan for these challenges. The current U.S.
Treasury Secretary, Steve Mnuchin, has stated that AI workforce issues are “not
even on our radar screen.” The White House Office of Science and Technology
Policy (OSTP), which was instrumental in leading AI policy work during the
Obama administration, has been essentially depleted of staff, with more than 70
out of 100 positions unfilled. The current administration has effectively
deprived itself of critical expertise and insights on AI.
Without a recognition of the historic challenges of
artificial intelligence — or a focus on preparing workers, soldiers, and
diplomats for the future — will the United States be able to withstand AI’s
disruptions or maintain competitiveness? Though China’s success in implementing
its highly ambitious AI strategy remains to be seen, China’s actions are a
clear indication of governmental commitment to this agenda at the highest
levels. At this point, the United States is failing to implement the sensible
and ambitious AI policy that the Obama administration left behind. Meanwhile,
China is using the Obama playbook to stride boldly forward into the AI
revolution.
Comments
Post a Comment